(updated 5.4.2021)

5.4.2021 2:21pm - For readers asking what "fusion" is, I would recommend: However, for a quick and punchy distinction between "fusion" and the PO framework of generalizing results by contriving ignorability assumptions, here is a gem:

5.4.2021 12:26pm - Thanks go to ASF @AmericanSephard for reminding us of #WorldPressFreedomDay and of the legacy of our son Daniel who contributed his best to the idea of Press-freedom.

5.4.2021 10:52am - (Replying to @hangingnoodles and @curiouswavefn) I am inclined to agree with @fchollet , but ML will not be the epicenter of this transition, unless it learns the principles of "deep understanding".

5.4.2021 6:02am - (1/2) Readers ask for my comments on a new "fusion" paper that does not look like ordinary "fusion". Indeed, the misappropriation of the word "fusion" is unfortunate, for it would create "confusion" on top of "fusion". My second comment: The sentence 1/2
5.4.2021 6:02am - (2/2) (Replying to @yudapearl) "the literature on these graphical approaches has not yet included estimators for data fusion parameters" is misleading. First, such estimators have been produced. Second, it creates the impression that the graphical approach limits the use of powerful estimators - surreal. 2/2

5.4.2021 4:35am - (Replying to @AndersHuitfeldt) And this keeps them going for more than half century? I would get awfully bored and switch to Flat Earth Society -- in constant need of creative ideas.

5.4.2021 4:30am - (Replying to @LoganLender @newzionists and 5 others) Thank you for a thoughtful Tweet. I am sure many of us have asked ourselves same questions and have written about them here and there. It would be great to compile all questions and answers under one cover. A great project for @newzionists

5.4.2021 4:27am - (Replying to @AndersHuitfeldt) I am not looking for graphs, I am looking for principles of inference. Are there such principles in the Cochrane society?

5.4.2021 2:07am - Mistaken identity. I confused William Cochran with Archie Cochrane, sorry. But, can anyone tell me if an organism like me, deeply immersed in the art causal inference, has a chance of understanding the principles behind "The Cochrane handbook". Any intersection at all?

5.4.2021 12:00am - (Replying to @gottfriedmath) The concept "Cochrane Tools" is new to me. I wrote a paper on Cochrane's contributions, see, but have not encountered "Cochrane Tools". Any pointer?

5.3.2021 5:59pm - I wish I could join this discussion if only someone can tell me what the "RCT/Cochrane combo" is, and "evidence type" is.

5.3.2021 1:53pm - (Replying to @McApple08392419) HMM... I am inclined to believe its hand coded model, not discovery.

5.3.2021 1:17pm - uses all the right buzzwords and all the right arguments on why we need causal inference. But they are secretive as to whether they use causal models, what kind, and how they represent them. We should give them another year to get out of the closet.

5.3.2021 12:46am - Happy birthday Theodor Herzl, I once wrote an oped about the unexpected results of your First Zionist Congress. Nowadays, you can enjoy the young greetings coming from the New Zionist Congress @newzionists

5.2.2021 11:52pm - Sharing a thoughtful interview with great many quotes we can use in our little corner of causal inference fighting for its scientific role in a world ruled by institutional bulldozers eg. "the alchemy of expertise and high moral posturing"

5.2.2021 7:44pm - Holocaust Memorial ‘Stumbling Blocks’ Defaced With Anti-Israel Graffiti in Cologne, Germany

5.2.2021 7:31pm - For those among us who understand that the key to peace in the Middle East rests in Palestinian schoolbooks, this news comes as light at the end of a long tunnel, twisted by clerics, politicians, "peace making" profiteers and State-Dpt wishfuls.

5.2.2021 6:46pm - The sentiments I am hearing from my family in Israel: (1) Shock over the awful way these families have lost their loved ones. (2) Anger at a government that let clans run hazardous rituals in some parts of the country, while resisting Gvt safety control.

5.2.2021 5:07am - (Replying to @VC31415) We need to add U--->Y,and remove A--->Y, else we get an M-structure, where X is not a confounder.

5.2.2021 4:13am - (Replying to @yudapearl and @VC31415) Or, how about D<--U--->X--->Y, D--->Y now X is a de-confounder, and not an ancestor of D.

5.2.2021 4:05am - (Replying to @VC31415) How about this: D---->Y<----X Now X qualifies, but it is not an ancestor of D.

5.1.2021 10:20pm - (Replying to @ZArchimboldi) Not thus far, and I am open to ideas.

5.1.2021 9:02pm - It is for this reason that I regard counterfactuals as a signature of creativity. There is no counterfactual without rule-breaking.

5.1.2021 8:57pm - Wishing all Christian Orthodox readers on this channel a Holy Saturday. Below, the #HolyFire ceremony in Jerusalem, today.

5.1.2021 8:43pm - According to Heckman it's all in Haavelmo, see .

5.1.2021 1:58pm - Another Quora question that keeps coming up: "How does the Rubin causal model differ from graph theoretic approaches like Pearl's do-calculus?" Strangely, those who know both can answer it, and those who don't keep asking, eg.,

5.1.2021 11:10am - This is a useful resource list for causality, some of the items are new to me. [ Wome important ones are missing] Thanks

5.1.2021 10:57am - "What Are The Differences Between Econometrics, Statistics, And #MachineLearning?" While I am sure NOT every economist/stat/ML researcher agrees with my Quora taxonomy, the many "how true!" awakenings in those fields were worth the effort.

4.30.2021 11:57pm - Another insightful piece by Einat Wilf, especially for readers who are sick listening to the color-prismed mentality of politicians, analysts and reporters.

4.30.2021 6:36pm - (1/2) HRW and Causality. The man who founded Human Right Watch (HRW) quit - he could not stand the corruption. Yesterday, I watched how Ken Roth violates causal logic in his obsession to criminalize Israel: He criticizes Hamas for doing x and Israel for doing y, never mentioning 1/2
4.30.2021 6:36pm - (Replying to @yudapearl) how x and y are causally related, i.e., that y responds to x and x responds to higher forces, none related to y. A good homework for causality 101. A sneaky trick for politicians. An illuminating lesson for a world gone mad, screaming "Apartheid!!" (eg

4.30.2021 11:12am - Iran hit with four-year judo ban for ordering athletes to avoid Israelis via @timesofisrael
4.30.2021 11:12am - Replying to @yudapearl How long before UN follows? Anyone knows of another organization in which one member threatens another with extermination and all other members accept it as normative?

4.30.2021 1:37am - (Replying to @kazorral) You have asked to be corrected if wrong, so let me try: Israel is NOT a result of European antisemitism but of the natural aspiration of any people to reclaim its normalcy. Both Palestinians and Israelis are paying the price of an ideology that opposes that natural aspiration.

4.30.2021 12:55am - I would say: Any such account is unavoidably hallucinatory, idiotic of even worse, Zionophobic, implying that the occupation and its ugliness is Israel's choice, pursed for fun, sports or to prolong a conflict that the other side is trying so hard to end.

4.29.2021 9:27pm - 44 people crushed to death, dozens hurt at mass Lag B’Omer event in Mt. Meron via @timesofisrael
4.29.2021 9:27pm - (Replying to @yudapearl) Initial sights from the terrible disaster that took place in Israel, during the Lag-Baomer celebration at Har Miron, near Tzfat.

4.29.2021 8:44pm - Continuing our discussion of Personalized Decision Making, a new post clarifies the distinction between personalized and covariate-specific decisions. The former concerns a specific individual while the latter, a subpopulation resembling that individual.

4.29.2021 8:18pm - (Replying to @GenomicsCRT) Confessing to a similar weakness, I still come back to the Book of Why to enjoy how convincing I sounded three years ago.

4.29.2021 3:32pm - Thanks for noticing the peculiar absence of the C-word. And glad we have a word today, called "cause", to measure the progress DL is making towards becoming a "science".

4.29.2021 3:07am - As kids, we started gathering scrap wood two weeks before the holiday, then, when the bonfires were lit, the whole neighborhood came together, toddlers and grandparents, to commemorate the (failed) Bar-Kochva revolt against the Roman empire, 132-36 AD.

4.29.2021 1:44am - (1/ ) Remember our discussion about inverting Wright's Rule? I conjectured that it is not known to economists and SEM researchers. Well, I was right. After posting an inquiry on SEMNET, it turned out that none of the papers who deals with SEM identification 1/
4.29.2021 1:44am - (2/3) (Replying to @yudapearl) indicates that the authors had any clue of how to invert Wright's Rule by path tracing - with one exception. I hate to embarrass authors of SEM books with this shortsightedness (they do not take it well) so I will just whisper to instructors of SEM: If you want to surprise 2/3
4.29.2021 1:44am - (3/3) (Replying to @yudapearl) your students with a powerful method of writing down a (partial) regression coefficient for every path coefficient, peak at Rex Kline' book, 4th ed. page 179-180; your students will thank you forever. 3/3

4.28.2021 8:05pm - How to harness population data for personalized decision making was discussed several time on this education channel. Our latest thinking on this issue is now posted here: I hope you see its potentials for personalized medicine and precision marketing.

4.28.2021 10:58am - (Replying to @prem_k) Don't sell business people too short. They understand quite well that if they wait for others (eg. scientists) to invest in next generation technology it might take 3 generations, not one.

4.28.2021 10:34am - First time I see this quote by Clayton Christensen. Now that it has the business community behind it, I think Causality has a future.

4.28.2021 9:30am - (Replying to @TheLeanAcademic and @tdietterich) The "moon" is "deep understanding" of the world, or at least some chunk of it, as defined by the ability to answer question at all 3-levels of the hierarchy.

4.28.2021 9:25am - (Replying to @roydanroy) By "statistical learning" I mean Rung 1 exercise, ie., associations, prediction and retrodiction. RL is above it, for it has the primitive of "action". I have commented several times that RL is level 1.5 on the Ladder, as explained by Elias here

4.27.2021 11:03pm - (Replying to @kncukier) Agreeing to everything you are saying, so why is it futile?

4.27.2021 10:09pm - (Replying to @sameer_) No put down intended. Re-examining priorities was intended. My physics book defines "inertia" as motion that persists long after the reason for it has ceased to exist, even after new reasons have arisen to change direction.

4.27.2021 9:47pm - (Replying to @tdietterich) From my perspective, it is wasteful to study why deep learning works so purely at levels of understanding we aim to achieve. Plain geometry works beautifully in 2-dimension, but when we wish to go to 3-dim we do not spend more resources studying plain geometry.

4.27.2021 9:19pm - (Replying to @roydanroy) You are right. The "waiting" was wastefully spent in statistical learning. We finally understand what it takes, and its time to take advantage of it.

4.27.2021 9:13pm - (Replying to @EliSennesh and @tdietterich) However you describe it internally, "deep learning" is still fitting data by complex functions.

4.27.2021 9:09pm - (Replying to @tdietterich) Differing on argument 2: The principles by which DL optimizes fitting of data have nothing to do with Deep Understanding (as defined here ttps:// and are not generalizable to Rung 2 and 3.

4.27.2021 9:03pm - (Replying to @tdietterich) I beg to differ. To reach the moon we better study space rockets, not birds, though birds were by far the best technology we had (for a long time) for rising above the ground. A theory of what birds (ie DL) are missing is available, and it tells us: its a waste to study birds.

4.27.2021 8:54pm - I have proposed a provisional definition of "deep understanding" = The ability to answer questions from all three rungs of the Ladder of Causation. I justified it here:

4.27.2021 8:02pm - To clarify, this was not intended as a pun but as a serious question: If we are interested in acquiring Deep Understanding, why struggle to understand Deep Learning? Shouldn't we first understand the ingredients necessary for deep understanding, then see if DL can acquire them?

4.27.2021 5:01pm - (Replying to @HL327) Your conclusion mentions "efficacy', which is a causal notion. So somewhere there ought to be a causal assumption supporting it.

4.27.2021 2:44am - (Replying to @DavidRalin) artificial general intelligence

4.26.2021 10:21pm - Congratulations go to three AI colleagues who were just elected Members of the National Academy of Sciences: Michael Kearn, Yann LeCun and Anna Karlin,Center%2C%20Los%20Gatos%2C%20Calif.

4.26.2021 9:32pm - Highly recommended for AGI thinkers.

4.26.2021 9:21pm - Hard to understand why we should struggle to understand Deep Learning instead of learning Deep Understanding.

4.26.2021 9:05pm - What's the difference between a Zionophobe and an anti-Semite? The latter has the decency of not demanding legitimacy.

4.26.2021 5:01am - Here is a paper introducing causal inference in Stat 101. Has anyone tried to follow their recommendations?

4.26.2021 12:31am - If you interview 5 DL gurus and ask them how to snap out of the DL quicksand, you are not likely to hear: "Reach for a branch or person's hand to pull yourself out." In other words: Data Science is a 2-body problem - connecting data to reality. See:

4.25.2021 8:19pm - (Replying to @akaus001) Thanks for reminding me of this fierce battle of Haifa which was instrumental in ending the corrupt, misfunctioning and brutal Ottoman rule in Palestine, 1918. True, the British mandate was not a picnic but, compared to the Ottomans they were angels.

4.25.2021 11:59am - First day with NO Covid-19 deaths! I think it is a remarkable achievement and a hopeful sign for all of us. It is also remarkable that so highly divided country, with essentially no functioning government, can nevertheless unite and overcome a danger, when push comes to shove.

4.25.2021 11:50am - srael has recorded no new daily Covid-19 deaths for the first time in 10 months!

4.25.2021 11:14am - When it comes to insight and perspective, there is none like Bernard Henry Levy. Read his column in Tablet:

4.25.2021 11:06am - It is comforting to see 20,000 French citizens still having a strong sense of decency in their blood. Will Sarah Halimi become the George Floyd of France?

4.25.2021 10:56am - (Replying to @newzionists) For a moment I read it as: "Technicolor Democrat", which could be a beautiful description of a Democrat who changes colors to fit the fashionable (a few names come up, but none matches Bernie Sanders)

4.25.2021 3:23am - As we mention the Armenian genocide by Ottoman rulers, and Erdogan's ambitions to revitalize the "glory" of the Ottoman empire, here is a little known chapter on the liberation of Be'er Sheva in 1917 by Australian and New Zealand's troops. Today this city celebrates #AnzacDay

4.25.2021 12:52am - The escalation Zoo: "Two smaller militant groups" have managed to fire 36 rockets on Israel, Hamas wants all dummies to believe that they have "no interest in escalation" and, watch my lips, public opinion makers will warn Israel against acting "disproportionately"

4.25.2021 12:32am - (Replying to @akaus001) Thanks for the article. I was aware of Erdogan's Caliphate ambitions and his meddling with the Temple Mount in Jerusalem, but was not aware of his adventures in South East Asia and his courtship of Indian Muslims. One of the most dangerous man to watch, next to Khomeini.

4.24.2021 11:53pm - (Replying to @95thoughts) and @LauraDeming This is a welcome endorsement for #bookofwhy and another support of my theory that statistics is a science lacking intuition; whatever intuition it carries comes from its causal component which, unfortunately, statisticians have managed to suppress.

4.24.2021 3:53pm - If Erdogan wants to convince us that the events of 1915 do not reflect a pathology in Turkish culture or education he should prove it in his actions vis-a-vis the Kurdish people in North Syria and vis-a-vis the Jewish people in Israel.

4.24.2021 1:21pm - I join @DavidHarrisAJC and my Armenian colleagues and friends in remembering the 106th anniv. of #ArmenianGenocide. I was 9 yrs old when I first read about it, from the book "The Forty Days of Mosa Dagh" by Franz Werfel (1933). It never left my mind and heart.

4.24.2021 12:30pm - (Replying to @wait_sasha and @chrisalbon) Why would anyone try to troll me? If they want my inner truth, all they have to do is to read the last chapter of Causality:

4.24.2021 6:15am - (Replying to @wait_sasha and @chrisalbon) What's wrong with commenting?

4.23.2021 10:46pm - (1/ ) Remember Wright's Rule? (ie that the correlation between any two variables is given by the sum of products of path coefficients along unblocked paths, see Today I had the thrill of my life when my nephew (16) called: I am stuck on page 83 of #Bookofwhy 1/
4.23.2021 10:46pm - (2/ ) (Replying to @yudapearl) Where did you get this formula? He asked. "Its Wright's Rule" I said. "That's cool!" he said "How did he derive it?" Instead of showing him, I said: "Today we know how to invert it, and write down each coefficient in terms of the correlations" "Wow, that's super cool!!" 2/
4.23.2021 10:46pm - (3/3) (Replying to @yudapearl) He made my day, though he did not ask me who derived it because, come to think about it, it is really "super cool." It takes the fresh mind of a 16 yr old to spot a nugget in things we take for granted. [For the inversion formula, see p. 84 of Primer] 3/3
4.23.2021 11:52pm - (Replying to @yudapearl) A reader reminded me that the inversion formula is more explicit in Theorem 2 of, and also applies to models with unobservables. As far as I know, the formula is still not known in the SEM literature (some economist may correct me) which is astonishing.

4.23.2021 9:57pm - Glad the JDA is re-labeled the "Van Leer Definition". I called it the "thoughtless 200". And to Michael Waltzer who begs for a "distance" between Zionophobia and anti-Semitism I say: Double the distance! Of course! the former is so much uglier in its heinous quest for legitimacy.

4.23.2021 6:50pm - Put differently, the failure is in not recognizing that causal inference is not a patch that can be added on to statistics-ML education like, say, steepest descent or logistic regression, or support vector machines etc, but requires a new language, new grammar and new calculus.

4.23.2021 6:36pm - (Replying to @chrisalbon) Even if you do not finish this book you can find the answer already in Chapter 1: Causal modeling is not taught in stat 101 is because stat instructors are denied a mathematical language to ask causal questions and math tools to answer them. It's a case of failed leadership.

4.23.2021 11:11am - (Replying to @stu_frost @psb_dc and @nigewillson) I agree on causality being a key to moving forward, but I am unclear on why business problems are more sensitive to this need than, say, public health problems.

4.23.2021 2:00am - (Replying to @SimonFlyvbjerg and @mitomaths) Is it really a matter of "direct vs. indirect" approaches? or a matter of feasibility? Can statistical tools answer the causal question "How does x affect y?" I doubt it. The Ladder of Causation says: No way!

4.23.2021 12:16am - Protest rallies will take place on Sunday, in Paris and across the US. 15 yrs ago I wrote a eulogy for Ilan Halimi [no relative] which explains why France refuses to prosecutes its anti-Semites, and why the French press is complicit with the refusal.,7340,L-3229504,00.html

4.22.2021 11:08pm - (Replying to @artistexyz @VC31415 and @causalinf) I see nothing wrong with it. What makes you hesitate?

4.22.2021 3:56am - (Replying to @artstop) Do not knock her. Sarsour could have been Bernie Sander's running mate, had he won the primaries. Glad he told us who his bed fellows are before the election.

4.22.2021 3:38am - (Replying to @voidmstr) No kidding!! Kumi Ori Ki Ba Orech U'chvod Adonai Alayich Nigla.....

4.22.2021 3:35am - We should keep this paper handy next time an economist asks: "Show me one "real life" study where DAGs could have made an improvement".

4.22.2021 12:21am - Why "my" historical infancy? Good question. Because my Bar Mitzva reading was from Book of Isaiah (60:1): "Arise, Jerusalem, your light has come!" So, I feel a personal attachment to the Book of Isaiah in the Dead Sea Scroll, and to the two scribes who wrote it 2,200 years ago.

4.22.2021 12:06am - (Replying to @gottfriedmath) Glad the words "causal inference" are becoming a selling point for companies line Netflix, but reading the text, I see "We use A/B tests to introduce new product features" to be the closest they get to causal inference. I hope they read #Bookofwhy and learn all seven tools.

4.21.2021 11:06pm - AI analysis shows two scribes wrote one of the Dead Sea Scrolls
4.21.2021 11:06pm - (Replying to @yudapearl) Who would imagine that AI would some day unravel the mysteries of my historical infancy?

4.21.2021 5:37pm - (Replying to @akaus001 and @genomixgmailcom) No, No. Occaio-Cortez is forgivable. She had the courage to admit she knows nothing about Israel, but her voting base wants her to say something nasty, so she learned to utter O-CCU-PA-PA-PA-TION, it sounds so sophisticated.

4.21.2021 4:34pm - (Replying to @genomixgmailcom) I thought about Rashida Tlaib, but decided she is too busy accusing Israel for animal cruelty and child pornography to be interested in Iranian women and UN credibility.

4.21.2021 3:54pm - We are eagerly waiting to hear the wisdom of leaders of the women's rights movement, say Linda Sarsour or Congresswoman Ilahn Omar; they surely have something to say to women under the Ayatollahs and to women who still lookup to the UN for protection of universal values

4.21.2021 10:28am - Zoom Registration For Upcoming Webinar With PFLP’s Leila Khaled Now Removed From Platform
4.21.2021 10:28am - (Replying to @yudapearl) Evidently, Universities have outsourced their moral compass to Zoom and other social media platforms. We will soon be witnessing Zoom's executives deciding course material.

4.20.2021 8:48pm - University leaders do speak out forcibly against invited merchants of hatred but only when the hatred is peddled against protected groups. Palestinian terrorists eg. #LeilaKhaled aim their poison craftily against Jewish students, letting @UCPrezDrake off the moral compass hook.

4.20.2021 1:41am - This is amazing, it's me and my second cousin Theodor Bickel! Where did you get it, Dennis? One of my most cherished memory is singing "We were here!" with Theo; I sang in Hebrew "Anachnu Po!" and he did in Yiddish "Mir Zeinen Do". He passed away six years ago and I miss him.

4.20.2021 12:35am - (1/2) A few years back, at the end of a lecture on Schindler List, a group of us had dinner in an Italian restaurant . Someone started humming the Hymn of the Jewish Partisans and soon the restaurant owner and all his guests joined us full voice singing: 1/2
4.20.2021 12:35am - (2/2) (Replying to @yudapearl) The owner asked: What's this song? It sounds familiar!! We told him. He asked: So why are you singing it today? Ans. Because 70 years ago, hundreds of fighters in the ghettoes and the woods wanted us to know one important thing: "We were here!" We are affirming their wishes. 2/2

4.19.2021 10:32pm - (Replying to @artistexyz @VC31415 and @causalinf) A causal DAG is a parsimonious encoding of a set C of constraints on the distributions resulting from all possible interventions. m() should measure the degree to which C is violated but, note, we need interventional data to detect the causal portion of m().

4.19.2021 11:26am - Data science is like a pit of quicksand. Rescue advice: Lean back, reach for a branch or person's hand to pull yourself out.

4.19.2021 10:39am - It's missing, true, but it is forgivable. Anyone who starts AI with a course on statistical learning, or machine learning, will find it very hard to snap out of the curve-fitting bubble and ask some basic questions about interventions and explanations.

4.19.2021 12:30am - (Replying to @roydanroy and @ceobillionaire) Cryptic ... ????

4.19.2021 12:29am - A Palestinian leader may speak "self determination" and "equal right" but would never utter the words "in their homeland"; this would be a political suicide. Watch his J-Street speech and see if I am not right.

4.18.2021 4:00pm - There is a new paper dealing thoroughly with Cyclic SCM However, for instruction purposes, I would just go over the cyclic examples on Causality page 215-17, and demonstrate how interventions and counterfactuals are computed.

4.18.2021 10:54am - I was tempted to forward you proposal to our Social Science Department, but I'm not sure they are ready for the 21st Century. BTW, I would delete the "DAGs", since we do not wish to exclude cyclic SCMs.

4.18.2021 10:45am - Kabul's last synagogue! Sad.

4.17.2021 10:10pm - All it takes is one Zionophobic faculty member to turn a reputable university into a platform for Hamas and a nightmare for Jewish students. Bright @SFSU students who are seeking refugee status in a more civilized university will receive personal guidance and advice.

4.17.2021 9:29pm - (Replying to @MimeeXu and @PooyanJamshidi) I'm relieved. The Preface was more scary than it should have been. Good, now I will not be accused of over-selling. Thanks.

4.17.2021 3:52pm - Agreeing with an old Tweet below. For statisticians to admit that AI has resolved Simpson's Paradox is to admit a century of blindness. The late Dennis Lindley was the only statistician I know who acknowledged both, the resolution and the blindness.

4.17.2021 3:38pm - Glad the paper is finally out, gatekeeprs or no gatekeepers, AI or not AI. It will be added to my Simpson's archives

4.17.2021 12:37pm - (Replying to @ildiazm @f2harrell and 8 others) Am I right to assume that by "personalized" you mean Conditional Causal Effect? Many use these interchangeably, but I think a distinction is warranted.

4.17.2021 11:36am - A forward looking UAE? I can hear my Palestinian colleagues protesting: "And what about your honor? i.e., our vow to insure that a land, once Muslim remains Muslim?" I see hordes of BDS activists screaming: "And what about Palestinians' inalienable rights to eradicate Israel?"

4.17.2021 3:41am - Our discussion of going from population data to individual cases will not be complete without presenting a LITMUS TEST by which to evaluate the many proposals currently loading the statistical and ML literature: Try your favorite method on these data

4.17.2021 3:17am - Here is a new article saying, in essence, let's forget causal analysis and make the correct decision on each individual Why not? We (presumably) have "machine learning methods for individual causal effect estimation." But do we?

4.17.2021 2:27am - Note that @CarlosCinelli explanation, as well as the analysis here DOES go from population averages to individual cases, but does so CORRECTLY, using counterfactual logic to combine experimental and observational studies.

4.17.2021 2:12am - (Replying to @mendel_random and @adamkvonende) I've enjoyed, and have learned from the history of the "personalized treatment" aspiration, and was waiting for the punch line: "Now we know better ... Now we know when and how it can be achieved or approximated..." That is why I was expecting a causal model for guidance.

4.16.2021 6:00pm - (Replying to @f2harrell) But I am also arguing that we do have the methodology to re-do individualized medicine correctly, on solid scientific grounds.

4.16.2021 5:56pm - (Replying to @f2harrell and @peter_jemley) I tried to digest but could not figure out what quantity is the target of the study and what is being improved by combining EHR and RCT. Do these quantities have notational signatures?

4.16.2021 3:09pm - This is indeed what I am saying. See I further say that those who are not using counterfactual inference, are wasting tax payer's money when they promise "personalized medicine".

4.16.2021 3:04pm - (Replying to @BHilbush) I should have said "personalized medicine as it is practiced today" is on shaky ground, not the one that should be practiced, based on counterfactual logic. See

4.16.2021 11:02am - I have argued before (eg that the whole enterprise of "individualized medicine" is on shaky grounds. The priests of this industry need be re-educated, but educators are busy building another "Data-Science Center" on campus, with the help of NSF and NIH.

4.16.2021 10:35am - I also like @CarlosCinelli explanation, and it is amazing how hard it is for people to accept the fact that single case explanations cannot be based on population averages. This is probably due to historical resistance in the PO sub-cultures to the Rung 2 vs. Rung 3 distinction.

4.16.2021 10:15am - Readers who found our "Crash Course in Good and Bad Controls" to be useful, are likely to find this revised version to be even more so: It has been groomed under the critical eye of two reviewers, one from a (non-extinct) DAG-skeptic culture.

4.16.2021 1:23am - The UCLA group on History of Science presents: HPASS 29 April 2021, 4pm; Marina Banchetti (Florida Atlantic University) talk on: An emergentist conception of chemical properties Meeting ID: 953 9617 1279 A telling story on how Alchemy was put to rest.

4.16.2021 12:50am - A eulogy I wrote for Ilan Halimi 15 years ago,7340,L-3229504,00.html brings up additional factors on Why France Refuses to Prosecutes its anti-Semites, and why the French press is complicit with the refusal.

4.15.2021 10:21am - Thank you @MomentMagazine for retweeting my piece on 1948, the year when history held its breath and then started a new chapter. Please share my poetic inspiration with your readers:, describing The Chase that has been going on since, relentlessly.

4.15.2021 4:55am - This text reads fresh each time I read it, especially for the 100th time. I feel like sending it to the Ayatollahs; are they on Twitter?

4.14.2021 9:55pm - Singing is an essential part of their practice, thanks to visionary medical training.

4.14.2021 7:57pm - My humble offering to Israel, on her 73rd birthday. Mirdaf (The Chase) by Yaron London "A land whose long history reads chase after chase, Two thousand pages plus one,... Yes, she is fearful, but as if not concerned, Will wait for the end of the chase"

4.14.2021 6:34pm - It is inspiring to see philosophers of economics continue their century-old struggle to understand what economic models are about. This article explores the "modular" vs. "ceteris paribus" theories, and has much to benefit from Structural Causal Models.

4.14.2021 4:21pm - Highly recommended to all readers. @noatishby is one of the best guides to Israel's tapestry, especially for the ideologically perplexed.

4.14.2021 3:57pm - (1/ ) Happy Birthday Israel! Sharing @EinatWilf pride and gratitude for having been born into a generation of Jews who knew their place, their country, and have earned their own defense when I was 11 years old. An unending dream #YomHaatzmaut #IndependenceDay. I am also grateful 1/
4.14.2021 3:57pm - (Replying to @yudapearl) for being able to contribute, at this last phase of my life, to the fight against the deligitimizers of Israel, especially those who craftily come up with "scholarly" proposals how/why Israelis should be peacefully lured back into a state of statelessness.

4.14.2021 3:32pm - (Replying to @JaapAbbring @VC31415 and @RoyalEconSoc) Is there a link available to Imbens review of causal inference?

4.14.2021 3:29pm - Please nominate your favorite paper for 2021 AIJ Classic Paper Award, deadline may 14.

4.14.2021 8:21am - Text over Drama -- New reading of Israel's Declaration of Independence.

4.13.2021 3:33pm - (1/ ) "Estimation" is a term loaded with technical interpretations. Face-2 can better be described as "guessing". Say I obtain records of 1,000 students, each with measurements of height (X) and weight (Y). I am told that, in the future, only height will be given to me, & I'll need 1/
4.13.2021 3:33pm - (Replying to @yudapearl) guess the weight the best I can. Now, given my rusty memory, I can't remember 1,000 pairs of numbers, but I'm willing to compromise the accuracy of my guesses. What do I do? I choose b such that y=bx+eps is the best fit line, and I remember only b. No E(Y|x), no cov, just Gauss

4.13.2021 9:18am - This is a better link to 1948, the year that everything changed: In memory of our neighbor's son, Danny Baal-Koreh, who went to fight five armies and came back in a coffin. May his memory be a blessing to all who treasure peace and freedom.

4.13.2021 5:00am - (1/ ) At sundown, today, siren sounds will mark Israel's Memorial Day, commemorating her fallen soldiers and victims of terror. The least I can do in their memory is to share personal experience, as I did here: Standing out, I will never forget the first 1/
4.13.2021 5:00am - (Replying to @yudapearl) military funeral I witnessed, saluting our neighbor's son, Danny Baal-Koreh, and how the unbearable pain of void has given birth to a sense of hope: Behold, I noticed, the rifles carried by his army comrades were real rifles -- we can nos defend ourselves!

4.13.2021 3:12am - (Replying to @olcan and @roydanroy) Face (2) does not require conditional mean, just fitting.

4.13.2021 2:23am - (Replying to @roydanroy) Why do we need to "recover" the best fitting line if we can compute it? Is there any collection of points for which a "best fitting" line does not exist? Curious.

4.12.2021 11:57pm - And for causally-enlightened readers, here are two more faces of the regression-like equation y=bx+u : 3. E(Y|do(x)) =bx + c (Rung-2) 4. Y(x) = bx + u (Rung-3) Homework: Show that (4) implies (3) and (3) implies (2), but not the other way around.

4.12.2021 9:17pm - Linear regression has two faces, that should be taught side by side: 1. A claim about nature: E(Y|x)= bx + eps 2. A strategy for estimation: The best (min squared error) linear approximation of y . The former is empirically refutable, the latter is beyond refutation.

4.12.2021 11:20am - A more detailed obituary of Jack Minker, from the University of Maryland. He will be missed by many.

4.11.2021 9:43pm - (Replying to @HolgerCevallosV Philosophical details that have caused as much damage as mistaken philosophical theories.

4.11.2021 8:25pm - This would have another advantage: removing the confusion between regression and structural equations. Writing y ~ Normal(Xb, sigma) (or P(y,x), as Tom suggests) would not lure researchers into thinking the X in some way "causes" Y -- the confusion of the century.

4.11.2021 6:20pm - A sad day to AI. I am informed of the passing of Jack Minker, one of the founders of deductive databases, a friend, a fierce-less leader, and great human being.

4.11.2021 5:37pm - It's hard to match David Harris's essay on Israel's 73rd birthday in both its insightful historical perspective and its coverage of Israel's miracles and accomplishments. The most I can do is add a very personal touch: "What Israel means to me?"

4.11.2021 3:27pm - This is a correct take-away. Though the richness can be seen already in simple examples, as in Causality pp.35-6 where two different SCMs, both compatible with the same CBN, yield two vastly different probabilities of counterfactuals.

4.11.2021 12:37pm - Good question. The 2011 paper describes the process by which the vast space of functions in any SCM can be transformed into a finite space for functions, parametrized as missing variables in a CBN.

4.11.2021 9:38am - I remember how, in 1961, my colleagues at RCA Laboratories (Princeton, NJ) argued and argued whether Israel had the right to capture Eichmann and try him in Jerusalem. My aunt testified at that trial. History, so it seems, has taken her side.

4.11.2021 6:42am - (Replying to @PHuenermund and @JaapAbbring) I think what Paul alludes to is that absence of complete identification criteria has not been of major concern to economists. The class of non-cyclic problems, for example, in which we do have such criteria, was abandoned to the mercy of strategies that offer no such criteria.

4.11.2021 12:08am - I am a few days late in congratulating colleagues Alfred Vaino Aho and Jeffrey David Ullman, recipients of the 2020 ACM A.M. Turing Award for fundamental algorithms and theory underlying programming languages. Hearty congratulations Alfred and Jeff!

4.10.2021 6:13pm - (Replying to @danijarh @osazuwa and @robinc) SCM's are not DAGs. They allow cycles, as long as each variable (say each of X and Y) are assigned one function which describes how each depends on all other variables -- a society of responders. See solutions to counterfactual problems in cyclic SCM, Causality p. 215-217.

4.10.2021 5:32pm - @eliasbareinboim called my attention to the fact that the original papers cited below may not have mentioned the titles CBN and SCM explicitly. For formal definitions of CBN, see: Causality pages 23-24,, & (for Semi-Markov models)

4.10.2021 5:15pm - This paper seems to have solved the cycle problem in SCM's. It would take me weeks to absorb all the results, but I strongly recommend it to readers who are concerned with feedback, cycles, equilibrium etc. esp. economists who blame arrow-phobia on cycles

4.10.2021 2:18pm - More and more readers are asking me about the differences between SMC's and CBN's. I would recommend the following 1993 papers: where the distinctions were first articulated, and this 2011 paper for summary.

4.10.2021 10:33am - I got it. You are doing "response type" analysis (as in in the context of CBN, and derive bounds on probabilities of counterfactuals. I have no objection, of course, the bounds are tight. Except I would call it counterfactual analysis, instead of CBN.

4.10.2021 10:11am - (Replying to @calimagna and @AndrewYNg) In my vocabulary data-centric = non-causal, and model-centric = causal. (The first is a theorem, not opinion). That is why we need to clarify semantics.

4.10.2021 9:55am - (Replying to @crude2refined and @ykilcher) I agree the the time has come to rein, classify and cure "bias" with mathematics, (and we have the necessary mathematics today) else it will rein us.

4.10.2021 12:33am - (Replying to @SophieBays and @ch_nira @SofieBays I have never heard of PhD's in Bayesian Networks. What university offers this degree? Most importantly, please share with us some of the thoughts/problems about causation that you think are unique to industry.

4.9.2021 12:35pm - Clarifying: the semantic clash is about what the meanings are of the words "data centric" vs. "model centric". If @AndrewYNg meets my challenge, readers will be able to reconcile the two interpretations, and take a position on where AGI is heading.

4.9.2021 10:19am - The writeup by @Analyticsindian represents semantic clash between two non-communicating echo-chambers. I challenge @AnrewYNg to recommend my writings to his followers (eg, & and, in return, I hereby recommend his to my followers.

4.9.2021 3:27am - Here is what Holocaust Museums fail to understand: The relevance of the Holocaust cannot be based solely on depressing lessons of the past; the defiant and miraculous emergence of Israel from its ashes must be an organic part of any Holocaust narrative:

4.9.2021 2:55am - I don't usually follow the conversation of business people and what make them see things different than others, but this post tells got me thinking: If they resonate with the Ladder of Causation, perhaps they are doing other things right.

4.9.2021 2:27am - WHY did the Holocaust Happen? Here is a scholarly analysis of the question, which is perhaps more terrifying, captivating and memorable than most emotional expressions evoked in Holocaust Remembrance Day.

4.8.2021 7:33pm - The future of Holocaust remembrance via @JNS_org
4.8.2021 7:33pm - (Replying to @yudapearl) Sobering words by Rabbi Meir Lau, Chair of Yad Vashem, about #ItStartedWithWords and how the dehumanizing words we hear all around us have the potency of ending up with consequences even the authors of those words have not imagined.

4.8.2021 6:15pm - (Replying to @zaffama and @dakami) I don't get it. We agreed to start with ONE BN, X--->Y. Why go to a SET. Perhaps you mean a set of SCM's compatible with X-->Y. If so, the set is parametrized by functions, not by P(u). I need refresh my memory, going back to Definition 2 of

4.8.2021 2:17pm - Be they Rubin given, or Neyman given, but how do you start a research problem? Do you assume something is known about those "givens", (say ignorability?) or do you derive them from something more fundamental? (say, how variables affect each others?)

4.8.2021 2:15pm - (Replying to @zaffama and @dakami) I don't get it. We agreed to start with ONE BN, X--->Y. Why go to a SET. Perhaps you mean a set of SCM's compatible with X-->Y. If so, the set is parametrized by functions, not by P(u). I need refresh my memory, going back to Definition 2 of

4.8.2021 2:08pm - (Replying to @zaffama and @dakami) I confess to ignorance of "credal networks". Can you discuss how they compute bounds on PN = P(Y(X=1)=1| X=0, Y=0) from this CBN X--->Y.

4.8.2021 1:46pm - Watch: Israel Marks Holocaust Remembrance Day via @JOL
4.8.2021 1:46pm - (Replying to @yudapearl) Its the day when I remember my grandparents, aunts and uncles, to whom I say again and again: things have changed! Your lives have spawned a new life, a new generation of Israelis who will fulfil your dreams in freedom and dignity: You will be proud of them!

4.8.2021 1:05pm - (Replying to @artistexyz) The query was P(Y(1)=1|Y=0, X=0), not tilde(X).

4.8.2021 1:39am - (Replying to @dakami) Here the problem is not hardness but non-uniqueness, which persist over "real world" problems.

4.8.2021 12:04pm - (Replying to @zaffama and @dakami) This is indeed what current works in counterfactual reasoning are focusing on: narrow the bounds with whatever information you have available. See, and

4.8.2021 1:21am - PO takes counterfactuals as God given and uses them to compute other quantities (eg causal effects) but it does not compute counterfactuals from any model or reality, nor from CBN.

4.8.2021 1:07am - (Replying to @dakami) It's true for the asymptotic probability PN so, surely, it does not get better for finite sample.

4.7.2021 10:42pm - Many readers try to get counterfactuals from Causal BN, but it is impossible. Let's examine this one X--->Y. It gives us P(x,y) and P(y|do(x)) and P(x|do(y)). Now, try to compute the counterfactual PN = P(Y(X=1)=1| X=0, Y=0) See Causality p. 35-36 and

4.7.2021 8:06pm - (Replying to @gottfriedmath) I do not understand the question. CBN only covers rungs 1 and 2, not 3, because it is defined by interventional distributions, not by deterministic functions like SCM. We need determinism to define counterfactuals.

4.7.2021 7:06pm - Those who read my 2019 commencement address know that I prefer the word "Zionophobia" over "anti-Semitism." I nevertheless support #IHRA after a personal experience with EDI officers on 3 campuses; they won't move by any other means, except the legal.

4.7.2021 12:21am - (Replying to @garyzhubc) "Operationalize" is a better word than "generalizes". Lewis's reliance of "similarity" among worlds leaves the question of representation unsettled. Same can be said about Mill's methods; someone should sit down and operationalize or interpret each of his method. A nice project.

4.6.2021 11:34pm - Speaking of advertising for a job or a postdoc, here is one from @eliasbareinboim I do not know what could be more "innovative" than research on Causal AI. If I was qualified I would go for it.

4.6.2021 11:25pm - (Replying to @FreakX19 @omaclaren and 8 others) Feeback loops are accommodated in SCMs.

4.6.2021 10:58pm - (Replying to @omaclaren) You are right about continuous time models. But I would put it that way: Because of their discrete nature, DAGs and SCM do not add a valuable inferential tool on top of ordinary analysis of SDE, unless you allow for discrete time approximations.

4.6.2021 9:54pm - (Replying to @omaclaren) DAGs and SCMs can be seen as causal tools on top of standard statistics. So, everything you can do with statistics you can continue to do with the addition of DAGs and SCM. There is nothing doable in plain geometry that cannot be done in 3-dim geometry.

4.6.2021 9:06pm - Agree. I still wish to alert readers of this educational chanel to watch out for statements such as "DAGs cannot do XYZ"; chances are they do not represent an understanding of DAGs or SCM .

4.6.2021 9:01pm - My position about #datacentricAI is expressed here: I just havn't realized that some folks are proud to belong to a "data-centric" ideology; I bet they han't seen the Ladder of Causation and its implied limitations:

4.6.2021 9:39pm - For the life of me, how can DAGs NOT allow anything that statistics does? DAG take one view of SCM (ie., the oracle of all wisdom) and statistics takes another view of SCM. How can one view interfere with what you choose to do with the other?

4.6.2021 8:51pm - (Replying to @matloff) You said nothing "wrong," but I can't figure out how researchers who care about causal issues can find DAGs to be of limited value. Unless, of course, they are PO disciples, in which case it is not the standard errors that shape their preferences but other forces.

4.6.2021 8:31pm - (Replying to @matloff) I assume you mean "In narrow stat", where one cares about standard errors of statistical parameters and not about causal parameters and other things that really matter. "In our school, we would consider your microscope of limited value" said the blindman to the biologist.

4.6.2021 8:14pm - More Than 350 Academics Sign Letter Supporting IHRA via @jewishjournal
4.6.2021 8:14pm - (Replying to @yudapearl) I have signed this letter after realizing that adapting the IHRA definition of antisemitism is a necessary instrument to jolt college administrators out of their stubborn and deliberate refusal to curb Zionophobic hostilities on US Campuses. I hope you can join me.

4.6.2021 5:23pm - US Lawmakers Reintroduce Bill to Require State Department Review of Palestinian Curriculum
4.6.2021 5:23pm - (Replying to @yudapearl) My! My! I can't believe my eyes! After decades of fumbling with dozens of imaginary "roadblocks to peace" Lawmakers decide to take a look at the huge elephant in the room --- Palestinian textbooks -- and why Israel is a temporary phenomenon, soon to be erased from the earth.

4.6.2021 10:35am - "PhD Position on Causal Inference & Machine Learning" I was struck by this AD from TU Delft, The Netherlands, for putting CI first and ML second : Evidently, the faculty at TU Delft understand where the future of ML lies. If qualified, I would apply.

4.6.2021 4:04am - (1/ ) As predicted in "Simpson't Paradox (SP): The riddle that would not die", here comes another paper contesting the centrality of causal thinking in explaining the surprize element of SP My comment: Surely some level of surprise may 1/
4.6.2021 4:04am - (2/ ) (Replying to @yudapearl) accompany reversals in purely predictive situations. But the persistent fascination of decades of philosophers and statisticians by the phenomenon can only be explained when the reversal amounts to violation of deeply entrenched logic, i.e., the logic of causation, not merely 2/
4.6.2021 4:04am - (3/3) (Replying to @yudapearl) a departure from mental heuristics about calculating proportions. The impossibility of a drug that is "good for men, good for women and bad for a person" is a theorem in the logic of causation, not in the logic of proportions, nor in any other logic.
4.5.2021 11:09pm - (1/3) While I have been hoping to see every statistics department offer a course in causal inference, @Lester_Domes is ahead of me. He hopes to see UG courses in statistics start with the principles of causal inference as they emerge from the data-generation process. I once labeled 1/3
4.5.2021 11:09pm - (2/3) (Replying to @yudapearl) this process "The Oracle of all Oracles" (see, as contrasted with the "joint probability function" which is currently the crowned oracle in statistics education. I am afraid none of these dreams would be realized until machine-learning (ML), which 2/3
4.5.2021 11:09pm - (3/3) (Replying to @yudapearl) currently dominates all educational agenda, adopts the data-generation process as its oracle.

4.5.2021 7:21pm - (Replying to @jeanqasaur and @redteamwrangler) It worked! I am totally verified. It feels like being vaccinated; a band-aid on your shoulder and an illusion of being accepted -- both harmless.

4.5.2021 5:22pm - (Replying to @artistexyz @Perperuna2 and @DrBobGoldberg) Because I adopt some aspects of Bayesianism which are unique to Bayesianism, like reliance on prior knowledge. I cannot generalize to pregnancy -- never experienced it, no prior knowledge.

4.5.2021 11:55am - (Replying to @radagaisus) This requires a definition of "the processes that explain the relationship" which is usually murky, and it still does not define the scope of "causal inference" which includes identification and estimation.

4.4.2021 2:25am - (Replying to @awhillas) Not really. The relationships shown in the diagram describe features of reality, not of the data, and whether we can measure the confounders or not does not change this fact.

4.4.2021 10:41pm - (1/ ) I am with you all the way. Unfortunately, the more I read ML papers the more I come to realize that this little distinction is not in the vocabulary of my ML colleagues; reality as a source of data simply does not exists, only the data exist, like the shadows in Plato Cave. 1/
4.4.2021 10:41pm - (Replying to @yudapearl) This is also the greatest barrier of communication between ML folks and the research community at large, for which a notational distinction exists between the data-generation model (eg. y := ax + u) and its distributional manifestation (eg. cov(x,y) =c). Time to remove barriers.

4.4.2021 9:08pm - With one important twist: The relationships shown are NOT in the data but in reality, of which the data is merely a blurry shadow.

4.4.2021 8:44pm - On the other hand, @cmMcConnaughy , the past decade has brought us an explosion of new understanding of external validity and the basis for generalization, see #Bookofwhy, so I am not as alarmed as you are. I am confident this new understanding will percolates to the right folks.

4.4.2021 3:28pm - Glad to hear that Karim Chalak will be joining the department of economics at the University of Montreal. We should expect to see fresh wind in the sails of causality in economics!

4.4.2021 3:12pm - (Replying to @jacula_modyun) On the contrary. CI is definitely a winner in the sense of providing an understanding of the hurdles that researchers faced through the ages. The anti-Whig movement prohibits the description of those hurdles in terms of what we know today, and that's IMO incomplete & boring.

4.4.2021 10:10am - I like the title "Whiggish Historian". It gives you a license to challenge traditional historians: "He who does not understand causal inference today should not write about the history of statistics or econometrics or philosophy of science, bc these are all chapters of CI"

4.4.2021 7:00am - Happy Birthday, Tel Aviv! I have always seen you as a metropolis, but my grandpa met you when you were still a baby, like the picture on the left. You just never get old !!!

4.3.2021 11:17pm - (Replying to @EinatWilf) That reminds me of how Peter Jennings (ABC) asked Hanan Ashrawi: "Do you recognize Israel's right to exist?" and she answered: "Arafat recognized Israel in 1988". It is always "So and so did, yesterday", never "yes I do, today".

4.3.2021 9:30pm - Normalize Punching Harder. Because Zionophobes still think they are invincible, given their moral "Uber Alles" supremacy.

4.3.2021 11:04am - For me, the strongest argument is the proverbial phrase: "Correlation is not causation", which means: the latter is reality, the former is appearance, often misleading.

4.3.2021 10:54am - One of the most urgent and least understood question of our time...

4.3.2021 1:58am - (Replying to @intrinsic_motiv @roydanroy and @tdietterich) If #Bookofwhy appeals to (ex)-theoretical physicists I know it has merits which, one day, will also be recognized by machine learning folks. Thanks for letting me know.

4.2.2021 7:37pm - If has been exactly a year ago that we first heard of the acquittal of our son's murderers: They are still under some sort of "detention" awaiting Supreme Court hearing of our appeals, while the Gods of justice refuse to believe their own eyes.

4.1.2021 9:24pm - (Replying to @artistexyz and @akelleh) It is always possible to write Rung-2 quantities like P(y|do(x)) in terms of counterfactuals (Rung 3) but not the other way around. Some beautiful relationships are derived here:

4.1.2021 9:16pm - (Replying to @artistexyz and @akelleh) A/B and RCT give you P(y|do(x)) for all y and x. Some like to summarize it in the difference E[y|do(x=1)]-E[y|do(x=0)], others like the ratio E[y|do(x=1]/E[y|do(x=0)], or whatever. The choice is arbitrary and secondary.

4.1.2021 2:48pm - (2/2) Glad to retweet this, because an identical question came up just last week: "How does Rubin Causal Model differ from Pearl Causal Model?" and will continue to haunt the novice until the priests of the former agree to solve one toy problem using the two frameworks. They won't. 2/2

4.1.2021 2:35pm - (1/ ) A useful analogy is our vision system, which takes data from two sensors (our eyes) and constructs a 3-dimensional depth perception. The principle is also called "triangulation" by many philosophers (eg Deaton and Cartwright who craved for it and 1/
4.1.2021 2:35pm - (Replying to @yudapearl) weren't able to operationalize it for obvious reasons: they did not have the logic of 3-dim geometry (read: causality). The name "triangulation" came from the Greek geometers who were able to estimate the distance of an enemy ship by taking two measurements on the shore. Try it.

4.1.2021 2:17pm - (Replying to @mryap) One correction: Daniel's story took place around 586 BCE, the year when the First Temple in Jerusalem was destroyed by Nebuchadnezer, the King of Babylon. That is, about 2,500 years before R.A. Fisher. Albeit, it was not randomized.

4.1.2021 2:06pm - (Replying to @PogrebnyakE @prem_k and 3 others) "Data Fusion" is what it says. You take data from here, and data from there and, because the sources are different, they provide two different views of reality so, as a result, you get more than the sum of the two components, provided that you know how to combine them properly.

4.1.2021 6:29am - (Replying to @nyarlathotepesq @yskout and 4 others) "assuming away" is a cultural phenomenon, unrelated to the information available from the testing.

4.1.2021 6:27am - Very useful introduction to "data fusion", for the uninitiated, doing away with problem-specific nomenclature, and putting all tasks involving multiple sources of information (experimental, observational, selection-biased, population subsets, etc) under one umbrella & one engine.

4.1.2021 5:39am - (Replying to @ashtroid22 @pablogerbas and @EpiEllie) Last I checked, Hayes was still in the pre-causal mediation era. Is it updated now?

4.1.2021 5:25am - (Replying to @tdietterich @masud99r and @PogrebnyakE) The information assumed about the target is a qualitative assessment of what mechanisms may account for disparities with the source. (And, indirectly, what mechanism are assumed to remain invariant).

4.1.2021 5:17am - A/B testing gives us P(y|do(x)), so all the pedagogical papers on data fusion, that discuss how to incorporate interventional distributions in the analysis essentially describe the role of A/B testing in CI. Examples:,

3.31.2021 10:31pm - (Replying to @mphielipp and @jsameijeiras) I would consider model-based inference to be level 3, even if it's not RL, and even if the model is not learned, provided the model is rich enough to allow the generation of explanations (eg, Causes of Effects,

3.31.2021 3:03pm - (Replying to @intrinsic_motiv @roydanroy and @tdietterich) Nice proverb. I once used: [Physicists] continued to write equations in the office and talk cause–effect in the cafeteria But that was because they could not write equations for cause-effects, even if they wanted to.

3.31.2021 7:54am - Interesting article, giving me a chance to read what some of my colleagues think about AGI and to try to understand it. So far, some success, but not entirely. I wish they spoke in equations.

3.31.2021 2:26am - (Replying to @HenningStrandin and @neuro_data) In Cartwright time, philosophers were invoking temporal information in attempting to prolong the hope of reductionism. Today, we know that temporal information can't help, as shown by the M-bias model, so the impossibility of reductionism (to probability) became a theorem.

3.30.2021 11:52pm - (Replying to @neuro_data) In his Philosophical Magazine article of 1950 (first pub. 1949) he says: "chess is generally considered to require "thinking"... ; a solution of this problem will force us either to admit the possibility of mechanized thinking or to further restrict our concept of " thinking. "

3.30.2021 11:35pm - (Replying to @GaneshNatesh) Thanks for posting. I was not aware of this one.

3.30.2021 11:31pm - (Replying to @neuro_data) My recollection: At the end of his Scientific American article "A Chess-Playing Machine". But my copy is in the office, which is still under COVID siege.

3.30.2021 11:27pm - Thank you @AsraNomani for summoning the courage to fight for commonsense in education. As a grandfather of two teenage boys in Berkeley, California, I am extremely worried of what they are fed in school. I join you in

3.30.2021 11:13pm - (Replying to @neuro_data @Lizstuartdc and 3 others) So why shy away from all the fun adventures embraced by the word "causality" and limit it to "causal effects"?

3.30.2021 11:02pm - (Replying to @JonSchw73589513 and @vishakh) Sure. The ideas I borrowed from GSC are well documented in Causality (2000) and credited in @Bookofwhy (see page 244). Maudlin imagined some 1980's works which presumably could have saved me years of labor and which were NOT cited or credited. I invite readers to name any.

3.30.2021 10:37pm - (Replying to @Lizstuartdc @jmmaronge and 3 others) Gee, you are missing all the fun with so many other aspects of causality: mediation, responsibility, External Validity, data fusion, explanations, causes of effects, missing data, etc. In fact the entire landscape of Rung-3. I would reconsider.

3.30.2021 10:31pm - Indeed, Shannon wrote the first Chess Playing program in 1948 and posed the question "Can Machine think" twoyears before Turing (1950). However, unlike Turing, he hesitated to say: Yes.

3.30.2021 10:14pm - (1/ ) @vishakh , I like your review of @Bookofwhy, but I am afraid you fell into the same trap that Tim Maudlin did, assuming that I was not familiar with the Pittsburgh team of 1980s. I challenged him to name ONE idea from philosophy that would have saved me ONE hour of work, and 1/
3.30.2021 10:14pm - (Replying to @yudapearl) I believe I convinced him that his claim was premature. I think it is wrong to assume that if one does not recite the opinions of all philosophers from Aristotle to Lewis one must have missed something relevant and cannot add to philosophical wisdom. Reading Hitchcock's piece in
3.30.2021 10:14pm - (Replying to @yudapearl) the Stanford Encyclopedia shows that philosophers are benefitting substantially from the new causal modeling framework and, again, I can't see where I could have benefitted in writing #Bookofwhy. Your take on my debate with Gelman is fun and accurate.

3.30.2021 8:39pm (1/ ) I don't think any of them "switched". According to Thomas Kuhn, scientists just swallow their professional pride and let their students "switch". But listen to what Pat Suppes says about Causality 2000: "Without assuming much beyond elementary probability theory, Judea Pearl's 1/
3.30.2021 8:39pm (2/2) (Replying to @yudapearl) book provides an attractive tour of recent work, in which he has played a central role, on causal models and causal reasoning. Due to his efforts, and that of a few others, a Renaissance in thinking and using causal concepts is taking place." No "switch", yet "open mind". 2/2

3.30.2021 7:26pm - I know you are not supposed to laugh to "white supremacy' jokes, but this one blew my mind, perhaps because it is so real. Or surreal? Or, perhaps because my students at UCLA were publicly accused of "white supremacy". I laughed non-stop for 3 minutes.

3.30.2021 5:25pm - (Replying to @suchisaria @tdietterich and 2 others) Great papers, proving the point about searching under the lamp post vs. searching for your wallet.

3.30.2021 5:20pm - (Replying to @tdietterich and @roydanroy) True, this limitation is not a limitation of the "account" but of nature. Namely, under certain conditions of reality domain shift can be overcome, and under others it can't. Funny thing, some ML folks prefer not to know those conditions perhaps because it spoils the data-mining.

3.30.2021 3:36pm - (Replying to @roydanroy and @tdietterich) That's not how I would put it. I'd say: *subject to assumptions, some of them are checkable, that are NECESSARY for a solution to exist, namely NECESSARY for ANY clever ML algorithm to achieve what we hope it to achieve.

3.30.2021 1:09pm - (Replying to @tdietterich) You are absolutely right, trasportability could be another port, since "domain adaptation" and "transfer learning" have been in the minds of all ML folks. However, the necessity of causality to solving these problems is not as obvious as in A/B, so ML will continue to struggle.

3.30.2021 11:31am - You've found the one secret link to the hearts of #DataScience folks: -- A/B testing -- without which they could go through life as if the world is made of data, data and, again, data. Now they might be prepared for "explanation" and other goodies. Thanks. There is hope.

3.30.2021 3:52am - This new Alan Turing note is a tremendous boost of visibility and relevance to every computer scientist.

3.30.2021 3:33am - (Replying to @ruescasd) Mixtures of probability and causal models I classify as "causal" (as for example any structural causal model). The reason: you can't express the mixture in probability language alone.

3.30.2021 3:02am - Continuing this thread, the most troubling impediment for ML folks today is their unawareness of the fact that the 2nd obstacle no longer exists; we now have a language to express and analyze causal relationships. Radical data-centricity is not longer justified.

3.30.2021 2:33am - The primary attraction of probabilities over causality was "trust what you see, not what you think". This was later amplified by "trust what you can express and analyze, not what you can't". Pearson, Reichenbach and Suppes, were lured by the latter. Both work on ML folks today.

3.29.2021 11:40pm - Apropos. Here is a new paper received that, though not advocating reductionism, insists on: "We can distinguish four major conceptions of causation, emphasizing either regularity, probability, counterfactuals, or mechanisms." Why beat two dead horses?

3.29.2021 10:44pm - (1/ ) I do not know of any thinking organism who still believes that such reduction is possible. I think Nancy Cartwright was the last one to entertain it, before quitting in exhaustion: "no causes in, no causes out". I know of some economists and statisticians in hiding who 1/
3.29.2021 10:44pm - (Replying to @yudapearl) occasionally express hope that the world will return to the good old days of "its all in the joint density function", but none others.

3.29.2021 6:58pm - As a contributing author, I am sharing the final program of the 2021 AI-Stat conference: and the final list of papers: I see only 11 papers with "causal" in their title. Good sign, next year we will see 111.

3.29.2021 6:41pm - These suggestions of fighting off the theatrics of "Israel Apartheid Week" are fine, but they lack one weapon: HUMOR. Imagine the comedy that Charlie Chaplin ("The Great Dictator") would have created out of the Palestinian Apartheid Factory. Enlist Woody Alan and Mel Brooks.

3.29.2021 6:44am - I would start with these 3 position papers:

3.29.2021 6:37am - I strongly support this new organization, which aims to rethink the way Zionist students should reclaim their campuses

3.29.2021 6:17am - Mohamed bin Zayed University of Artificial Intelligence and Weizmann Institute of Science establish joint AI Program
3.29.2021 6:17am - (Replying to @yudapearl) Here is another way of re-starting AGI research on the right track.

3.29.2021 1:23am - Why Passover is a Universal Holiday via @jewishjournal
3.29.2021 1:23am - (1/2) (Replying to @yudapearl) And while we stress the universal message of Passover, let's not forget the Jewish aspect of the story. 3,000 years after the Exodus from Egypt, and 73 years after the establishment of Israel, the world is still debating whether the descendants 1/2
3.29.2021 1:23am - (Replying to @yudapearl) of those liberated slaves, those persistent narrators of the story of liberation, and those passengers on the Exodus-1947 voyage deserve a homeland of their own. This debate is perhaps the greatest indictment of the world's capacity to cope with oppression and injustice.

3.29.2021 12:23am - (1/2) "Whoever quotes a statement in the name of the one who said it brings redemption to the world." (Mishna: Avot 6:6). I must therefore credit my co-author, Dana Mackenzie for this insightful quote on R. A. Fisher, which is also relevant to our discussion of "explainable AI". 1/2
3.29.2021 12:23am - (Replying to @yudapearl) It is better to get a probabilistic answer to the right question, e.g., "was this accident caused by a reckless driver?", than a certain answer to the wrong question, e.g., "does reckless driving cause accidents?". That's the science of "Causes of Effects"

3.28.2021 1:04am - @EpiEllie Appropos, I just gave a talk at UCLA on "What is Causal Inference" As you can see, (1) it's not just about "causal effects" and (2) It is very close to what philosophers mean by "inference." Enjoy, and feel free to share with Epi-folks.

3.27.2021 3:40am - (Replying to @jacobinmag) Of course Anti-Zionism Does't Equal Antisemitism. It is a more dangerous and morally reprehensible form of racism, masquerading under the cover of "social justice", while denying one people what it claims for others. I wouldn't honor it with the title "Antisemitism."

3.27.2021 12:00am - Here is the best Holiday Greeting I received for Passover, which begins Saturday at sundown: A medley of all the Passover songs we sang since kindergarten, for 3,333 years, performed by the best musicians you can imagine. Happy Passover to all, and "Next year in Jerusalem!"

3.26.2021 7:32pm - (Replying to @AndroSabashvili) Try

3.26.2021 7:05pm - (Replying to @Syl4HealthData) Gee, where did you get that Seder plate? I've got one just like this one. Bought in Jerusalem, possibly 1976. Don't read @Bookofwhy before the 4-questions, you'll have no questions left to ask. Happy Passsover.

3.25.2021 5:49pm - (Replying to @bksalimi and @_aditya_lahiri) The original papers on "probabilities of causation" do not assume any causal model. They assumed only data from (1) experimental and (2) observational studies. See The bounds can be made tighter w/ additional assumptions.

3.25.2021 12:51pm - Please share with us how the folks in the SIGMOND-2021 Conference react to your argument. I am afraid convincing them that a wedded approach could be "futile" is as futile as showing them an approach that works.

3.25.2021 4:28am - (1/ ) Students of causal inference who were intrigued by the cleverness of the Inverse Probability Weight (IPW) estimator (eg, probably asked whether a similar trick can be played on the front-door formula, or some other causal-effect estimand produced by 1/
3.25.2021 4:28am - (Replying to @yudapearl) the do-calculus. The answer, surprisingly and pleasingly is YES. This recent paper shows that EVERY identifiable causal effect can be estimated by a "Weighted Empirical Risk Minimization" method, a fancy name for IPW-like estimation. Worth keeping in mind.

3.25.2021 12:54am - To our Greek followers: Happy Anniversary. We share memories about the Ottoman Turks as well as WWII. ps. I also love your music, from Nana Mouskouri to Mikis Theodorakis.

3.24.2021 11:39pm - (1/ ) Another paper deserving our attention is "Explaining Black-Box Algorithms Using Probabilistic Contrastive Counterfactuals" by Galhotra etal: which argues that existing works in explainable AI are futile attempts to capture the notions of probability of 1/
3.24.2021 11:39pm - (Replying to @yudapearl) sufficiency and necessity (see & #Bookofwhy) w/o the language of counterfactuals. The paper generalizes Tian's bounds from binary to arbitrary variables and should, hopefully, shake members of SIGMOD-2021 to take a fresh look at "explainability".

3.24.2021 6:40pm - (1/ ) Ed Simpson would be happy today, seeing his paradox returning to its causal womb after a century+ of agonized confusions by statisticians, probabilists and causality-avoiding philosophers. Those who follow my arguments why the paradox unveils the logic that governs our mind 1/
3.24.2021 6:40pm - (2/ ) (Replying to @yudapearl) (see #Bookofwhy and will appreciate the detailed account of Sprenger & Weinberger of the torturous journey Simpson's paradox has taken since its discovery (1899) and why I consider it settled. This will not stop statisticians (Meng, Spanos etc.) from 2/
3.24.2021 6:40pm - (3/3) (Replying to @yudapearl) denying any causal interpretation to Simpson's paradox, but it will, I hope, position their denial in the proper historical perspective. 3/3

3.23.2021 11:25pm - It seems that psychologists and ML folks are bemoaning the same road-block: "The Generalizability Crisis." See and This is understandable -- they have not acquired a language for causation. But it's 2021, why haven't they?.

3.23.2021 9:57pm - This is perhaps a good place to witness the mentality of my son's murderers: Upset by one grievance or another, steal the life of a precious human being, then dance to the tune of "what-about-X" and "what-about-Y". Someone is sure to join you, perhaps even on Twitter.

3.23.2021 5:21pm - Just thinking aloud. The Danish delegate to the UNHRC comes home for the weekend and, at dinner time, his daughter asks: "Dad, did you vote for this?" "Vote for what?" "This resolution!" --- I would bury my face in the spaghettis.

3.23.2021 11:06am - (Replying to @kwbroman and @robwilliamsiii) "Looking at data to see what it seems to say." This is what I meant by "pre-scientific data analysis". No put-down, just a personal confession that I don't know how to teach this art, be it to students or to robots. If I could, I would call it AI.

3.23.2021 10:53am - The journal "Observational Studies" is putting together a special issue on Leo Breiman's influential paper "Statistical modeling: Two Cultures." (2001). My contribution to this issue is posted here:

3.23.2021 1:18am - (1/ ) Thanks for sharing your SERious EPI podcast. All I can think about is El Capitan smiling at you reading transportability papers, nodding his fatherly head and saying: "It's all true, I checked the equations." It was also rewarding for me to learn how the seeds of rudimentary 1/
3.23.2021 1:18am - (2/2) (Replying to @yudapearl) ideas got baked in your mind and amplified to become the rich research agenda that you are currently pursuing. Thanks. 2/2

3.22.2021 11:08pm - (1/ ) As promised, I'm posting our AI-STAT paper "Exploiting Equality Constraints in Causal Inference" Motivation: When a parameter is zero, the graph becomes sparser and identification is easier. What about if it's not zero but a given constant? This was 1/
3.22.2021 11:08pm - (2/2) (Replying to @yudapearl) treated here: Now, what if you do not know the value of a parameter and all you know is that it is EQUAL to another? See the paper posted. 2/2

3.22.2021 9:01pm - (Replying to @erikbryn) Too cryptic. Can you explicate the context of "two states"?

3.22.2021 8:45pm - Abbas advisers push for new strategy: 'Soft sovereignty'; State is 'distant' dream
3.22.2021 8:45pm - (Replying to @yudapearl) For a moment I was excited to read about "Palestinian new strategy" hoping it entails "new thinking" about the conflict and ways to end it. Sadly, the authors talk about a "new strategy", not saying what "END" the "strategy" is to serve. Evidently they have not met an Israeli.

3.22.2021 7:55pm - (Replying to @artistexyz and @VC31415) Are you saying that score-based structure learning is fundamentally more informed than constraint-based? What is the source of this added information? I thought the two differ only in how fast they converge toward the (same) partially directed graph. Ready to be corrected.

3.22.2021 4:12pm - (Replying to @Nux1971) It works for me with:

3.22.2021 12:02pm - For those who requested the color slides of my UCLA lecture "What is Causal Inference - A Logical Perspective" Happy to share: Enjoy the colors and, if possible, the logic too.

3.22.2021 2:02am - Theoretically, having a dual citizenship, I could vote in the Israeli elections, two days from now. But this would require me to be physically in Israel, quite a hardship. The video below tells us what's at stake.

3.21.2021 11:42pm - (Replying to @MariaGlymour @UCSF_Epibiostat and 2 others) Where can we watch @Megtron9 show? Reading @Bookofwhy in Yosemite sounds like Rubaiyat: "Here with a Loaf of Bread Beneath the Bough, A Flask of Wine, a Book of Verse, and Thou, Beside me, Singing in the Wilderness, And Wilderness is Paradise Enow.

3.21.2021 11:26pm - Universities, funding agencies, and research foundations used to see their role as ensuring that someone on the team would be "skating where the puck is going, not where it has been." These institutions, in my observations, are now being led by players skating where it has been.

3.21.2021 10:56pm - (Replying to @carpioecon) It will be posted in a day or two. Thanks for your interest.

3.21.2021 4:56pm - This week, to my amazement, the most astounding discovery was not in AI, but in archeology. The ancient scrolls discovered in the Judean desert blew my mind And, with history breathing on my neck, I recited the text of Zacharia:
3.21.2021 4:56pm - (Replying to @yudapearl) "Once again shall old men and old women sit in the streets of Jerusalem leaning on a cane because of their great age: And the streets of the city shall be full of boys and girls, playing in the streets." We had it memorized in 5th grade, in Hebrew, it all came back to me

3.21.2021 8:12am - (Replying to @RaulMachadoG) I do not regard statistical labels on data to be "interpretation". E.g., if I say: "X has mean = 0.34 and variance 0.78" have I provided an "interpretation" or just "summary"? Some statisticians, admittedly, believe that only statistical summaries amount to "interpretation".

3.20.2021 10:46pm - Anyone interested in a panoramic view of AI-Statistics? Take a look at 2021-AISTAT program & scan the 1,850 titles as fast as you can. Our humble contribution: #663: "Exploiting Equality Constraints in Causal Inference", something you always wondered about

3.20.2021 10:26pm - You have expressed beautifully the nature of the "learning-first" addiction that I tried to cure here: What I would still like ML folks to appreciate is how much of the cited limitations is a mathematical impossibility, not just temporary difficulty.

3.20.2021 6:36pm - (1/ ) Post my reply to @MFordFuture , I went back and reflected on the interview we did 3 yrs ago: That causality is a necessary component of GAI has become axiomatic, same with the idea that it needs its own representation & tools. What has not changed at all 1/
3.20.2021 6:36pm - (2/ ) (Replying to @yudapearl) is the myth that it can be added on, as a patch to traditional ML methods - a mistake. Watch our HAI centers mushrooming abound, w/ barely lip-service to causality. Watch our funding agencies re-pouring resources into unsolvable problems. Watch our educational institutions...2/
3.20.2021 6:36pm - (3/3) (Replying to @yudapearl) In short, this is the time for philosophers of science to observe, investigate and analyze, in real time, the inertial forces that keep science at bay. Why study phlogiston and epicycles if you can watch those forces in action, while the actors are still alive?

3.20.2021 8:23am - No one doubts that pure symbols and pure data are both a thing of the past. The interesting puzzle for philosophers of science is why it takes so long for our esteemed universities (& HAI Centers & NSF & DARPA & Sloan) to start teaching/understanding the principles of symbiosis.

3.20.2021 2:57pm - (Replying to @MFordFuture and @VentureBeat) Some call it "traction", others call it "awakening"; what's important to realize, it would entail a painful paradigm shift, as noted here I invite philosophers and historians of science to observe the dynamics of a gigantic Kuhnian shift in the making.

3.20.2021 8:12am - (Replying to @RaulMachadoG) I do not regard statistical labels on data to be "interpretation". E.g., if I say: "X has mean = 0.34 and variance 0.78" have I provided an "interpretation" or just "summary"? Some statisticians, admittedly, believe that only statistical summaries amount to "interpretation".

3.20.2021 7:11am - (Replying to @DavidWLevine1 and @Grady_Booch) Beg to differ. We do have a clear scientific roadmap to causality; what remains is a political struggle to liberate ML education from its data-centric addiction.

3.19.2021 7:04pm - The reason todays #AI developers do not mention causality is that it's become hard to fake. Much to the credit of modern causality researchers, simple litmus tests are now available for detecting attempted faking.

3.18.2021 11:22pm - It is a great honor for me to serve on the advisory board of The New Zionist Congress @newzionists . Working with a group of self-emancipated students to reclaim their rightful space on US campuses is an opportunity I find irresistible. Join!

3.18.2021 1:16am - Nice summary of an important article, though I have my own answer to the question: Why machine learning struggles with causality? Ans: No causes in, no causes out. (Conservation of causal energy - you can't drive with an empty gas tank).

3.18.2021 12:53am - Poetry in inequality, Congrats to the lovely Chernoffs!

3.18.2021 12:50am - Congrats to Laci and Avi !!!

3.17.2021 6:08pm - You don't need evangelists when math and commonsense are on your side. I am just a lens grinder, like Baruch Spinoza (1632-1677); once people open their eyes in a well lighted room, all they need is good eyeglasses, and an occasional lens grinder, when things get really foggy.

3.17.2021 4:27pm - Would love to see how these tools are used and perceived through the lens of a modern disciple of Haavelmo, Marschak and Koopman.

3.17.2021 3:45pm - Heroic congratulations to Bianca de Stavola @bldestavola on receiving the 2021 Bradford Hill Medal from the Royal Statistical Society: She is the only honorary decorated with the crown of "causal inference," which reads: The RSS is making great progress.

3.17.2021 3:18pm - It's hard for me to believe that a concept on which we form a general consensus (eg. sufficient cause) should be a "challenge" to define in terms of the only representation (SEM) known to enjoy a general consensus. There should be a trick there.

3.17.2021 12:38pm - (Replying to @DrBobGoldberg) Great example of "Counterfactuals in Poetry". This Passover we will be verifying each line against a causal diagram, and redefine "minimal sufficient set", while singing "Dayenu".

3.17.2021 12:19pm - Who can afford to miss a tutorial on "Fairness" spoken in the language of cause and effect?

3.17.2021 3:38am - In order to make this proposal more accessible to SEM scholars, the INUS condition would need to be expressed directly in SEM notation. Such a translation would be of great value to people concerned with explainability and with Causes of Effects.

3.17.2021 12:39am - @_KarenHao Your words remind me that I am an Asian American too, albeit from Western Asia, who is fortunate not to have experienced the kind of hate you are describing. My Israeli students do, and they are fighting back, like you, with pride and more:

3.16.2021 10:41pm - The reason I can't endorse Mackie, as does, is that none of his endorsers has addressed the weaknesses pointed out in Causality pp.313-15 and, more importantly, none would tell us how the knowledge needed to apply INNUS is represented.

3.16.2021 11:38am - (Replying to @balazskegl) Except we do it with a model of the world in mind, and NN, as they are practiced today have no such model "in mind".

3.16.2021 11:33am - (Replying to @JProtzko @Kane_WMC_Lab and @AndrewRAConway) This is also the answer I would give, except "Lord's paradox" is now "Lord's solution": Express the quantity you wish estimated in causal language and go after it, with whatever causal assumptions you can defend.

3.16.2021 7:27am - (Replying to @bnielson01 @DavidDeutschOxf and @DouglasCarswell) The word "evolutionary" may appear only sporatically (eg. @drfeifei talk at Montreal) but the intellectual paradigm behind the ANN arguments (ie, that knowledge resides in the data) is definitely "evolutionary."

3.16.2021 7:10am - This may be a more reliable link More importantly, how can anyone talk about "explanation", or "explainability", without the analysis of "Causes of Effects"?

3.16.2021 6:05am - Glad your club is discussing "Causes of Effects vs. Effects of Causes", which is one of the most neglected topics in the literature (eg. only 45 Google citations). People who talk about "responsibility" and "personalized medicine" w/o CoE are kidding us.

3.15.2021 10:34pm - Will The New Zionist Congress save American Jewry from itself?

3.15.2021 10:04pm - The New Zionist Congress now brings to US Campuses what the First Zionist Congress brought to ugly Europe of 1897: An "auto-emancipated Jew" (see -- an idea that has contributed to the world more than all its predators put together. We salute your voyage!

3.14.2021 6:02pm - Hurray! Today is PI-day, rejoice! Why is it that I feel such a joy on PI-day? Is it because I befriended Archimedes at age 9? Or because it unites all of humanity into one noble comradeship? Or b/c it reminds us that causal-DL is like squaring a circle?

3.14.2021 3:10pm - (Replying to @data4sci) Yes. My lecture: "What is Causal Inference" is now posted on Youtube:

3.14.2021 3:00pm - (Replying to @feishaAI and @bschoelkopf) True, but for every y=f(x)+Gaussian there is a model x=g(y,u) that is statistically indistinguishable from the first.

3.14.2021 5:39am - (1/2) Thanks for brightening my day. I just finished reading @sejnowski 's "The unreasonable effectiveness of deep learning in artificial intelligence" and was struck by the absence of the word "causal" in the article, not to speak of the Ladder. 1/2
3.14.2021 5:39am - (Replying to @yudapearl) This means that the possibility that the Deep Learning enterprise is bound to a flat 2-dimensional subspace of intelligent behavior does not occur to DL researchers. I was delighted therefore to learn that your course starts with the ladder and its implications. ML has a future.

3.13.2021 7:51pm - (Replying to @ColmanHumphrey) No parody, just honesty.

3.13.2021 6:14pm - (Replying to @Chris_Said) Gee, I wish I could resonate with this explanation, but I have hard time seeing the forest in the many trees being simulated.

3.13.2021 3:29pm - Once in a while someone posts a gem of a tweet and, rushing to click the 'like' button I discover that it is mine -- what a disappointment. The one below is such a gem.

3.13.2021 1:21pm - (1/ ) Unfortunately, what you call "Vanilla Simpson's Paradox" is still being presented in statistics textbooks as paradoxical, and showing that it can hold in the data is celebrated as a victory of statistical thinking. Rarely would a statistician address the Real Simpson's Paradox 1/
3.13.2021 1:21pm - (2/ ) (Replying to @yudapearl) and we know why -- the RSP is causal, and statisticians hate to admit that (1) there are basic limitations to statistical thinking and (2) they have not acquired the tools to deal with these limitations. See, Liu & Meng's and Spanos's attempts to avoid RSP, as described in 2/
3.13.2021 1:21pm - (Replying to @yudapearl) I think RSP should be discussed in the first week of every statistics 101 class and, come to think about it, in every ML class. How else can ML students learn that Data Science is more than just Data? Need ML go through a century of torment, like Stat?

3.13.2021 2:11am - Speaking about the design of large scale experimental studies. Or was it observational?

3.13.2021 1:22am - (Replying to @michaelgmadden) Guess what the salary is of UCLA's Vice Chancellor of EDI (Equity-Diversity-Inclusion). Just take a reasonable guess before reading this:

3.12.2021 10:07pm - Plus, it's a money-making enterprise for the "trainers". In the universities, the Equity-Diversity-Inclusion office is most lavishly budgeted and staffed, busy to create divisions to justify its existence and expansion. We are soon to hire more EDI officers than professors.

3.12.2021 9:43pm - My slides are available here:

3.12.2021 12:00pm - Many principled persons, employed by unprincipled employers are now facing this dilemma. That is why I am appealing to those can, to speak out for those who can't, before no one can.

3.12.2021 11:51am - (1/n) My last lecture received more "likes" than even my first tweet in June, 2018 Some complained that the message behind the exercise-cholesterol example was not clear. I concur; the punch line is that if we stratify on another variable, say: 1/n
3.12.2021 11:51am - (2/2) (Replying to @yudapearl) "color of owned car" we can get the same scatter plot, even though the color-specific trends are wrong. Conclusion: It is the story that makes trends right or wrong, not the data. This is discussed in greater length here and 2/2

3.12.2021 10:05am - (Replying to @BrianSJ3) No question that we need different types of "Why's" to run a causal conversation, and your taxonomy is a start (I could add a few more types) but I do not believe Aristotle taxonomy should be our starting point; it was not choreographed for automated implementation.

3.12.2021 9:46am - (Replying to @BrianSJ3) I couldn't find Aristotle's Four Causes to be very relevant to current generation of AI systems. But perhaps you can enlighten us in this direction?

3.12.2021 8:51am - (Replying to @McApple08392419) Yes. And I would add statistics/econometrics/ML.

3.12.2021 2:05am - It is refreshing to find young folks digging into the origin of ideas and exploring the logic of Zionism from the source material, untainted by modern colors and distortions. A grain of honesty to a fake-laden world.

3.12.2021 1:10am - Not exactly. "Blaming" does not help anyone. What the talk tried to summarize is: (1) We can identify which assumption may be responsible for the failure and submit it for further tests or refinement and (2) Replacing assumptions with data will not do us any good. No alternative!

3.11.2021 10:22pm - (Replying to @gkbytes and @VeredShwartz) Pleading innocent. I did not coin the term but still wondering whether an AI system can process causal conversation without semantics of cause and effect relations.

3.11.2021 4:56pm - Good news! My lecture on "What is Causal Inference" is already available on youtube, free for all: Enjoy! And, if there are questions, do not hesitate, preferably after reading #Bookofwhy, for the obvious reason that the answer is very likely to be there.

3.11.2021 3:39pm - Agree. Lord's Paradox highlights clearly many of the questions about ways of comparing two disparate populations, and it irritates the hell out of traditional statisticians. Why? Because the question is causal, and statisticians are embarrassed by lacking the tools to handle it.

3.11.2021 4:56am - (Replying to @schulzb589) My understanding is that every function can be "causal" or "non-causal"; the shape of the function does not determine the context in which it is used.

3.11.2021 4:17am - No way! We will continue to teach about Newton and Archimedes and Galileo, be they "white", "purple", "woke" or squalk, same as I was taught by my "white" teachers (see "They put a human face behind every theorem and every discovery." In this way,
3.11.2021 4:17am - (Replying to @yudapearl) children learn that science is not a book of facts and recipes, but a struggle of the human mind to unveil the mysteries of nature. It's going to be Isaac Newton, woke or no-woke!

3.11.2021 3:11am - And while some economists speculate on why formal solutions for "external validity and "data fusion" are "unlikely to be true", others show them to be both true and useful, in a language that every economist should find convincing.

3.11.2021 1:20am - Hilarious to see "reversed regression", that shook up social science in the 1980's make a comeback to 2019 NYT. "Men earn a higher salary than equally qualified women, and simultaneously, men are more qualified than women doing equally paying job" See

3.11.2021 12:58am - "Power in America now comes from speaking woke", unless we resist. It is our obligation, we, whose jobs and social standing is less vulnerable to the howlings of the woke vice squads, to speak out for those who can't.

3.10.2021 11:18pm - Another commonsensical paper has arrived: "We cannot simply accessorize our ML-based predictions with causal assumptions ...ML algorithms must be carefully integrated within a formal framework for causal and statistical inference." I hope ML folks listen.

3.10.2021 10:58pm - (Replying to @1MiBn) The formal solution may or may not account for the practical difficulty concerned. If it does not, one has to show that it does not, and for that one needs to understand the formal solution. The author admits he does not. So why not be positive and encourage readers to study it?

3.10.2021 8:59pm - I especially love Rubin's: “[Graphs are] based on an unprincipled and confused theoretical perspective.” and “to avoid conditioning on some observed covariates,... is nonscientific ad hockery.” His disciples continue in his track, e.g. Imben's "Comparison"

3.10.2021 6:04pm - Some economists will never reconcile to the idea that the "external validity" and "data fusion" problems can be given a "formal solution". They state that such "grand claims...seem unlikely to be true". Really? I'm eager to be shown why; shall we?

3.10.2021 1:44pm - ‘Saturday Night Live’ exhibits Zionophobia via @JNS_org
3.10.2021 1:44pm - (Replying to @yudapearl) Glad to see the term "Zionophobia" getting traction. So: "Dear Michael Che @NBC , Relax, You are not being charged with anti-Semitism, God forbid, you are being accused of a worse offense: Zionophobia, one you cannot deny, dare you?

3.10.2021 12:56pm - I will notify readers on the availability of the recording.

3.10.2021 12:45pm - (Replying to @RaulMachadoG) When I was a member of a Kibbutz, I met Marxists that were more humanists than my Rabbis. So that doesn't explain why some leaders of BLM would turn away from humanity to become bigots like Marc Lemont Hill.

3.10.2021 11:19am - (Replying to @RikTalo) We would need to know what "environment" is it that MuZero model and how. Is it the data-generating environment? or the data-fitter's environment?

3.10.2021 6:26am - Can not world knowledge come from data analysis? A question that we tried to answer here: But there is a more pressing question: Suppose you get some world knowledge from a scientist, would you know how to used it?

3.10.2021 5:35am - Compare to this paper on "interpretability" that, likewise, decries the fundamental ineptness of data-centric systems, yet does not offer an alternative.

3.10.2021 5:18am - Glad to finally see a paper on "explainability" that explains why explainability cannot be achieved through data-centric thinking, but requires world knowledge about what is to be explained.

3.10.2021 4:50am - I was asked to sign this petition concerning the BDS resolution at UCLA which I did. At the time, I urge organizations such as Alums for Campus Fairness to pay attention to how University Administrators operate, as I describe on:

3.10.2021 12:52am - (Replying to @robwilliamsii) Tried to access, but still password illiterate.

3.10.2021 2:30am - From Durban to The Hague: 20 Years of NGO Lawfare via @jewishjournal
3.10.2021 2:30am - (Replying to @yudapearl) When we hear the term NGO, we usually think of a group of creative volunteers acting for noble causes that governments tend to neglect. Things have changed the past two decades. Some NGO's became propaganda mercenaries in the service for shady regimes, as this article reveals.

3.9.2021 2:14pm - News from the expert: " The Zoom limit 500 people (including students in the class)". So, if you try by 4:05 pm, there is good chance you will get in.

3.9.2021 1:12pm - I got a general agreement from the audience at my SER talk, when I said: "When a statistician says 'we have to think hard' what he/she means is 'I havn't got a clue of how to do it'" This paper is laden with 'think hard' warnings.

3.8.2021 9:10pm - I was just told by the experts that the Zoom limit for this webinar is 300. Sorry. It behooves me to ask non-UCLA people to wait till 4:05 pm, and register if capacity permits. I guarantee that it will be recorded and released for unlimited watching in about a week time. Thanks.

3.8.2021 6:26pm - (1/2) We have to sharpen our understanding of what it means "DAGs solve ATE identification for non-parametric SEM". It means: If there is a way of estimating ATE from the data that depends ONLY on the assumptions in the DAG (but not on the shape of the structural equations, or the 1/2
3.8.2021 6:26pm - (2/2) (Replying to @yudapearl) the error distributions) then the do-calculus will find it and will tell us HOW to estimate it. If not, the calculus will tell us: "No way, the estimation strategy depends on the underlying functions, not on the structure alone." In your problem, it'd say: "Sorry, it depends"2/2

3.8.2021 1:37pm - Invitation. Tomorrow, 3/9/21 4pm PST, I will be presenting a tutorial on Causal Inference for UCLA students and faculty. See To join the webinar, please click the link below It'll be recorded. See you there for a gentle view of CI.

3.8.2021 7:56am - Here is an idea of an action that may restore sanity to higher education: A credible guideline for students, ranking universities on a racism scale 1-10, funneling talented students away from racist universities, thus threatening their academic ranking. Our last remaining weapon

3.8.2021 1:40am - (Replying to @lewbel) Don't you have to assume linearity? Recall, you are talking to a non-parametric mind, spoiled incurably by non-parametric models

3.8.2021 1:33am - So why is it that I can watch this clips for the 51st time? Is it because I did not have such sweet memories from basic training (1953)? Or because each of these soldiers could have been my grandson? Mazal Tov!

3.8.2021 12:22am - Our faithful calendar informs us that the March 16 is the deadline for nominations for the 2021 AMS Causality in Statistics Education Award. Details are here:,%245%2C000%20cash%20prize%20each%20year. Welcome are new ways for helping statistics instructors introduce causal inference in class.

3.7.2021 4:13pm - I hope you get some economists to attend, Morpheus, to free their minds. And don't forget to show them the difference between Rung-2 and Rung-3, they are the ones who have to go through it. #causaldiagrams

3.7.2021 3:35pm - And to think that I spent a good part of my army service in this place; can you bit it? I almost turned Christian from overdose of miracles.

3.7.2021 3:16pm - DAGs solve the ATE identification problem for non-parametric triangular SEM (see paper by Paul and Elias). I am fairly sure in your question, as stated, P(U1, U2, U3) is not identified, nor is causal effect of Y1 on Y2. I think some assumptions are missing. #EconTwitter

3.7.2021 1:40pm - (Replying to @AJCGlobal and @EinatWilf) Curious, but it is inaccessible. Please provide a link.

3.7.2021 11:30am - A coalition of Jewish and Asian voters managed to defeat California proposition 16 (affirmative action) in the last election. The same can be done to the Ethnic Studies Model Curriculum.

3.6.2021 10:09pm - Even if they aren't "successful" there's still something to be learned from every ethnic group, because ethnic heritage is a natural filter of valuable collective experience. As the Mishna says: "He is wise who learns from every person" (Avot 4:1). Our educators lost that wisdom.

3.6.2021 6:50pm - (Replying to @wanderandroam) I can only speak for myself. If I hadn't been blindly convinced that probability has the answer to all scientific questions I wouldn't have developed Bayesian networks. What it took to wake me up was not "evidence" but, perhaps, playfulness driven by the grand dream of AGI.

3.6.2021 3:57am - (Replying to @jamespayor) It is forgivable for a Law that was formulated in the pre-causal era to have imperfections. No point trying to patch it.

3.6.2021 1:53am - With all due respect to Goodhart, the Law is not true for "ANY observed statistical regularity". He probably meant "non causal regularity". The regularity between Aspirin and Headache has not collapsed though I've used it quite a bit for control purposes.

3.6.2021 1:19am - I've never heard of Goodhart's law so, naturally, my curiosity is say up. I'm still not sure, but glad nevertheless that causality within Reinforcement Learning is beginning to receive mainstream attention and that the relationships between the two are cast in SCM language.

3.5.2021 11:58pm - No funerals among friends. Thomas Kuhn even praised scientists for clinging blindly to their textbooks; only by blind conviction that truth is on your side can hard things get accomplished.

3.5.2021 11:43pm - My letter concerning the controversial Ethnic Study Curriculum proposed for California, and may become a model curriculum nationwide.

3.5.2021 10:01pm - First, the attention to #Bookofwhy exceeds my expectation, see Reviews: Second, the resistance to it in some quarters (eg ML) is also expected: The clinging of scientists to their textbooks, is one of the strongest forces known in subatomic physics.

3.5.2021 10:45am - Amazing how quickly can a noble movement like BLM be hijacked by bigots like Marc Lamont Hill and turn into "Israeli lives Don't Matter". How cheaply he "dismantles the Zionist Project" and the lives of 8 million human beings who awe their lives to this project, as if it was some
3.5.2021 10:45am - (Replying to @yudapearl) some city's garbage disposal project, not a culmination of a people's heritage. History is laden with noble movements turned Hill-like ugly.

3.5.2021 7:00am - (Replying to @PHuenermund @ccaballeroh10 and @VC31415) I would answer it: Because reality has a cycle, and SCM represents reality.

3.4.2021 10:14pm - Hurray! I am now Twitter "verified". I've found my true identity and feel like born again. All due to the help of @drfeifei and her friend Michael Montano. As God told Moses (from the burning bush) "I am who I am", which probably meant: "nevermind Twitter, just open your eyes."

3.4.2021 9:31pm - As a kid, I read many stories of Till Eulenspiegel, and I came to believe that I knew dutch, but this review taught me I don't, despite the many familiar terms. Thanks for sharing? How do you say "ladder" in Dutch?

3.4.2021 1:35am - (1/ ) Thanks for sharing this profound experience. It reminds me somewhat of how my mother, in Israel of 1939-44, labored to shield me of any information about the Holocaust, so that I will grow up NORMAL, free of any Jewish hang ups. I did, but I guess I still have some, 1/
3.4.2021 1:35am - (Replying to @yudapearl) e.g., I did not get the point about the Ham and Eggs; I thought only Jews and Muslims have a problem with pork.

3.4.2021 12:35am - (1/2) Not sure I can agree to all steps, but I am eager to understand what make some students think that way. How about just asking for each measurable variable A: what is the source of variations in A? Build a structural model and then ask: (1) What do you want to estimate? 1/2
3.4.2021 12:35am - (2/2) (Replying to @yudapearl) And (2) Can we, or how do we, estimate it with regression? See eg. and 2/2

3.2.2021 10:19pm - Looking forward to participate in this panel, hoping to understand how universities can permit hateful climate to metastasize unabated for over 2 decades, then speak about "discourse that is respectful to all members of our community." Beyond me.

3.2.2021 10:00pm - (Replying to @chad_oda @animesh_garg and @GaryMarcus) Over-stretched this week. Will try next one.

3.1.2021 5:10am - For inquiring readers: Yes, my interview is now available on line, and can be watched at http://standwithus.TV, top of the page. Blaming my accent, the subtitles could stand some editing.

3.1.2021 1:35am - Would love borrow and apply it to a couple of research communities around me: There is no bond as strong, durable, and sacred as the one between a researcher and his/her grad school textbooks.

2.28.2021 10:48pm - (Replying to @jslez) I'll ask around.

2.28.2021 9:32pm - (Replying to @memosisland) I was suspicious of this circularity and, for that reason, I tried to distinguish formally between "inductive bias" and "cheating".

2.28.2021 9:27pm - I remember the days when EU used to lecture to Israel about human rights and other sacred principles and, funny, some Israeli NGOs actually listened.

2.28.2021 9:03pm - (Replying to @elderofziyon) Is this the US @StateDept ? Verified by Twitter?

2.28.2021 10:41am - (Replying to @learnfromerror and @joftius) Kindle happened to block me from your book yesterday. It should be fixed soon. But if I havn't taken any data on the price of beans in China, nada, won't my independence assumption be subjective? Here we are, u send me to your book and I am sending you to China, no search needed.

2.28.2021 10:34am - (Replying to @joftius and @ruescasd) The line between "we know" and "we assume" need further explication. My grandma "knew" that unicorns exist and me "knowing" they don't is based on pure hearsay and some hunch that my hear-sayer was credible. That is why I am hesitate to conclude that hunches are not knowledge.

2.28.2021 9:51am - (Replying to @learnfromerror and @joftius) Weakness?? So, if I want to find out the weather in LA, my ability to assume (using priors) that it is statistically independent on the price of beans in China is a "weakness", while forbidding me to make this apriori assumption is a "strength". Hard to swallow.

2.28.2021 9:43am - (Replying to @docmrichards) My theory: Homosapiens are born with a template that turns into a causal model with the help of (1) hearsay and (2) sense data. Chimps are born with a primitive template and are deprived of verbal "hearsay". ML will be sped up if given a rich template and how to tune it.

2.28.2021 9:31am - (Replying to @ruescasd and @joftius) "Bias" has a negative connotation when referred to whimsical assumptions made, for example, to simplify computation, rather then convey knowledge. "Bias" is extremely valuable when used to convey knowledge, however dubious its source. For example, that unicorns do not exist.

2.28.2021 12:28am - (Replying to @joftius and @learnfromerror) Personally, I prefer priors that represent "bias" on priors that are chosen by route (eg. Dirichlett prior on some obscured parameter). The former are potentially defensible.

2.27.2021 11:51pm - (Replying to @learnfromerror and @joftius) You. Deborah Mayo @learnfromerror Not clear to me if you advocate abandonment or embracement of subjective beliefs in statistical analysis. Some consider subjectivism a weakness of Bayesianism, others (eg me) consider it a virtue. Where are you?

2.27.2021 10:55pm - (Replying to @learnfromerror and @joftius) Not clear to me if you advocate abandonment or embracement of subjective beliefs.

2.27.2021 10:45pm - Any chemists among our readers? A review of #Bookofwhy has just come to my attention: It says it's a "must for any serious student of philosophy of science, and should be required reading for any first-year undergraduate statistics class." No contest!!!

2.27.2021 8:57pm - Purim, Tehran 2021 !! Amazing! 2,600 years after Queen Esther risked entering the King's court to plead for her people, her story is read in this court, not less risky. What's more amazing, I understand every word of it, and have not set foot in Tehran for at least 2,000 years.

2.27.2021 7:43pm - (Replying to @robwilliamsiii @moorejh) felt truly divine revelation upon discovering statistics; imagine where he would be upon discovering causation, or upon reading "contesting the soul of data science"

2.27.2021 7:24pm - (Replying to @RaulMachadoG and @jslez) Clarice Weinberg was the first to show epidemiologists that controlling for a descendant of a mediator can be as bad as controlling for the mediator itself, see #Bookofwhy page 125-126.

2.27.2021 6:02pm - (Replying to @massbless) It's a two way street: If you have no language to express causal relations you will spend your career immersed in data, for which you have a language. Econ had structural equations which stat could not swallow, b/c they went beyond probability:,

2.27.2021 5:45pm - (Replying to @artistexyz) The main reasons PO was accepted by more statisticians than CI are: (1) It came out first (2) It gave the Illusion that everything is just standard probabilities and (3) It absolved researchers from the responsibility of understanding their assumptions.

2.27.2021 12:59pm - Thomas Kuhn was even more charitable. He doesn’t blame scientists for not jettisoning the paradigm that has been the basis of their life’s work; on the contrary, their firm belief that it can solve all their problems "is what makes normal or puzzle solving science possible.”

2.27.2021 12:50pm - (Replying to @McApple08392419) Estimation is indeed the last stage of the Inference Engine and I would include Deep Learning in providing effective estimators. Put differently, after you are assured that a function exists, unleash the best "curve fitting" method to approximate that function.

2.27.2021 7:56am - (Replying to @teppofelin) Great paper with many quotes and examples. I did not realize the data-centricity debate is raging in Genome Biology as well.

2.27.2021 6:15am - (Replying to @teppofelin) Great paper with many quotes and examples. I did not realize the data-centricity debate is raging in Genome Biology as well.

2.27.2021 5:09am - (Replying to @eagerbo and @jslez) This is indeed what #Bookofwhy is saying: Statisticians have been thinking and thinking about causality but have not committed that "thinking" to a formal language, because the only language they had was probability. Do you see any other language in Fisher, Neder, Senn, Mayo, ??

2.27.2021 4:05am - (Replying to @massbless) On the contrary. Statisticians never did causality poorly, they just avoided it, so as to do it right. The #Bookofwhy gives a historical account of a century of avoidance. (Summarized in & further discussed in

2.27.2021 3:27am - (1/n) Language-clinging is stronger than religion. According to legend, Hippasus, who discovered irrational numbers, was thrown overboard by his fellow Pythagoreans for denying that all phenomena in the universe can be expressed by whole numbers and their ratios. And if I was not 1/n
2.27.2021 3:40am - (2/2) If were not fearful of a similar fate, I would point you to present days machine learning research and its stubborn clinging to the belief that all wisdom comes from the data, hence expressible in probabilities. See "Radical Empiricism and ML" 2/2

2.27.2021 2:50am - I would be more charitable today and explain that what held statistics back was a stubborn clinging to the language of probability, not lack of insight. They did the best they could with the limited language in their disposal. Change of language is traumatic to most scientists.

2.26.2021 4:50pm - Acolytes tells us so much about the person. I was about to vote for @SenSanders in the primaries, but then he announced his foreign policy advisors, among them: James Zogby and J-street. Thank you Bernie for sparing me a painful regret.

2.26.2021 3:11pm - Glad Laura @shawfrank also found Purim's message as inspirational as I did I am somewhat more specific perhaps on where courage and fears lie these days of crisis. Thanks @shawfrank .

2.26.2021 2:43pm - Two corrections: (1) The edited version of "Radical Empiricism and Machine Learning Research" will not make it to Volume 8, but to Volume 9 of the Journal of Causal Inference. (2) The unbroken link is

2.26.2021 1:34pm - I think it carries lots of promise but, more importantly, we are about to understand how much promise it carries because some people on the team know causal inference so they can answer "what should the world be like for our method to work?" a question outside DL's vocabulary.

2.26.2021 11:49am - (Replying to @atafti) My question addressed the intuition behind this result. Or its counter-intuitive nature.

2.26.2021 11:47am - (Replying to @WaterFront8) True. But from over-arguing about Israel people tend to forget the context of 9 million human beings under unique circumstances of unending threats of annihilation. I am here to remind Bernie to this context which he prefers to forget - his voter base won't allow him.

2.26.2021 11:34am - (Replying to @timcash) Beg to differ. The memory of Nasser threats to annihilate me and my friends prior to the Six Day War is not a "dogma" but a vivid experience I share with my friends. It is used not "at the expense of others" but to monitor others with similar designs in mind. I heard it on radio

2.26.2021 11:17am - I conjecture that the "logic of hunches" will invoke metaphorical analogies to our toy world. E.g., "an electron acts like a tiny baseball, so it must have spin. Gee, how would I measure it if it does?". In other words, direct experience with baseballs act as a causal model.

2.26.2021 10:53am - (Replying to @timcash) Causal reasoning will tell us that as soon as you have a "state" you have a collective sharing memories of common experience, and such a collective is called a "Nation". So "Nation-State" cannot be blamed for all the horrible things that people can do.

2.26.2021 10:46am - (Replying to @onnoh) To the extent that "real scientists" use hunches, of course. But when we formulize the "logic of hunches," we should do it on top, not instead, of the logic of cause and effect.

2.26.2021 10:41am - I should add that Volume 8 of the Journal of Causal Inference will also contain an edited version of my blog post "Radical Empiricism and Machine Learning Research"

2.26.2021 10:32am - (Replying to @tvladeck) To the extent that facts support the allegations it is fair, but from what I read and know, the allegations were manufactured by less than humanitarian motives.

2.26.2021 10:03am - Bernie Sanders is a great humanitarian, with extremely wide followings. On his next humanitarian move we can expect to see him outraged at the Ayatollahs for planning to wipe out 9 million human beings from the face of the earth.

2.26.2021 9:19am - I've just turned this Tweet into an oped: My Purim Rabbi's Hat Happy Purim everybody.

2.26.2021 8:40am - The automatic generation of hypothetical, and continually refined and improving SCMs is possible. This is what scientists do, and they haven't done too bad. The aim of A.I. systems should be to construct an "automated scientist", and DL should be harnessed toward this goal.

2.26.2021 1:08am - (1/2) Disciplines that aspired to "find causal relationships" never made much progress. They should have first asked: "Suppose we have the causal relationships (ie, a model) what can we do with it?" And then: "What should the world be like to assure that we do the right thing?" 1/2
2.26.2021 1:08am - (2/2) (Replying to @yudapearl) Disciplines that aspired to answer these "easier" questions made lots of progress, because they asked scientifically answerable questions. Statisticians, demoralized by the impossibility of the impossible, never got to ask them, while economists,... for another discussion. 2/2

2.25.2021 4:51pm - (Replying to @causalinf @kneupane and 2 others) If you buy what statisticians claim about themselves, ie, that they are guided by and aim at causal understanding (see then Stigler barely touches on that aspect. Morgan is good if you believe that she understands structural equations - I have doubts.

2.25.2021 3:39pm - (1/ ) Glad to see that intellectual curiosity has not dried out on this educational channel. The answer to challenge is that there is a difference between "conditioning on" and "adjusting for". The latter invokes conditioning followed by averaging. The former will introduce bias 1/
2.25.2021 3:39pm - (Replying to @yudapearl) by suppressing variations of M), the latter perfectly washes this bias away. The fact that it works both for linear and non linear system is a miracle discussed in the Appendix here: Thank you all for a stimulating discussion. Viva la Miracles.

2.25.2021 12:24pm - (Replying to @peder_isager) I dont know what you mean by "unique". But trying to estimate p(y|do(x)) always makes sense, and is meaningful in practice.

2.25.2021 12:20pm - (Replying to @dylanarmbruste3) The question was whether adjusting for Z would introduce bias, given that Z suppresses variations of M.

2.25.2021 6:45am - (1/2) My My, this is a fairly strong indictment of economics. You are lucky you did not try to introduce new notation. As I said in an early tweet: [7.24.18 @3:46pm] - Economics has had all the potentials for becoming the queen of social science. It blew it by fostering an orthodox 1/2
2.25.2021 6:45am - (Replying to @yudapearl) and insular culture. Do you know that to post a comment on NBER Webserver you need to be an "approved NBER family member"? [exact words] I never figured out what they are trying to protect.

2.25.2021 5:31am - (Replying to @peder_isager) This is indeed one of the puzzles: Does it matter if the system is linear or non-linear?

2.24.2021 9:36pm - (Replying to @yudapearl @tdietterich and 2 others) Recall that adjusting for a son of M does introduce bias, so why not adjustment for a father of M, which can be as strong a suppressor (of M) as its son.

2.24.2021 9:28pm - (Replying to @tdietterich @pablogerbas and @JohnAnibalGomeV) This is a different question indeed. Our question is: We want P(y|do(x)) and someone says: "adjust for Z" and we are worried: what if Z suppresses M, would the adjustment introduce bias? Adjusting for M would, so why not adjusting for Z?

2.24.2021 8:54pm - This is as solid an argument as I have seen, and can easily verified by simulation. There remains only to explain why Z does not hurt us. After all, Z is a driver of M, so it constrains the variation of M, especially when their correlation is high, almost like conditioning on M.

2.24.2021 5:24pm - (Replying to @RaulMachadoG) The question is whether controlling for Z is bad for estimating X –> Y.

2.24.2021 4:08pm - (1/n) Friends. Tomorrow night, Feb. 25, we celebrate the Jewish holiday of Purim. Throughout my up bringing, Purim was a holiday of fun, joy, clowns and costumes, sort of Jewish Halloween, celebrating our deliverance from a genocidal plan in Persia, 580 BCE 1/n
2.24.2021 4:08pm - (2/n) (Replying to @yudapearl) However, in past years, I've come to see a profoundly personal meaning in Purim, especially in this powerful message that Mordecai sends to Queen Esther: “Do not imagine that you, of all the Jews, will escape with your life by being in the king’s palace,” says Mordecai. ".. 2/n
2.24.2021 4:08pm - (3/n) (Replying to @yudapearl) "For, if you keep silent in this crisis, relief and deliverance will come to the Jews from another quarter, while you and your father’s house will perish. And who knows, perhaps you have attained to royal position for just such a crisis.” I am using it often on high profile 3/n
2.24.2021 4:08pm - (4/n) (Replying to @yudapearl) colleagues who remain silent seeing their students intimidated by BDS cronies, or seeing their junior colleagues "cancelled" by EDI vice-squads. "You owe your stature to many who spoke out in such crises before" I say to them "it's now time to give back, or we'll all perish." 4/

2.24.2021 1:38pm - Very interesting! Because I have not met even a single convert among Bayesians. My theory: They fall in love with the power of probability calculus to capture prior scientific information and they refuse to see how this calculus excludes them from capturing most of science.

2.24.2021 1:22pm - Students of causality will find challenge in this variant: Z
Is it good or bad to control for Z to estimate P(y|do(x))? Z seems to be an "effect modifier" b/c it "clamps down" M, which is a mediator between X and Y. But is it?

2.24.2021 1:06pm - (Replying to @memosisland and @GaryMarcus) I would have liked to join your Colloquium, to refine (ie, falsify) my own theory of what it means to "understand" something: But I don't do facebook; any other entry?

2.24.2021 12:12pm - I was surprised to hear that Andrew Gelman agrees with me on this daring statement (we normally agree on generalities, rarely on substance). I checked the blog, and I don't see any agreement. Bayesians find it harder to understand causality than frequentists.

2.24.2021 7:27am - While so many are still busy digging for some malice behind Israel's victory over the Covid-19 vaccine, it is refreshing to read @ShMMor sober account of the factors that led to that victory.

2.24.2021 6:34am - The mentality of some writers ( @glcarlstrom) has reached the point where they can no longer see the moral deformity of their writings.

2.24.2021 5:58am - (Replying to @DavidHarrisAJC) And it will continue to rise, and rise, and rise... until we learn to spell the word Zionophobia. See why: See how:

2.24.2021 5:49am - Super-congratulations to @analisereal on his joining the University of Washington, and to causality researchers in the State of Washington on this terrific re-enforcement, which future breakthroughs will affirm.

2.24.2021 4:49am - (Replying to @avnerarik) I am not talking about "a pace workshop", but a clear commitment of the US State Department to eventual peace, entailing warnings to the ayatollahs an the PA that their textbooks -- the greatest obstacle to eventual peace -- ARE being monitored.

2.23.2021 8:37pm - I do not know myself, because Twitter refuses to "verify" me, and my university says they never dealt with Twitter. It seems I will forever remain "unverified" -- a wandering soul in search of identity. Help anyone?

2.23.2021 8:20pm - I am not a candidate for promotion but if asked to fill this form I think I would pass in flying colors citing my record of defending and promoting the inclusion the most excluded and intimidated minority group on campus: Zionist students. The first professor to try it: Hat off!

2.23.2021 2:54am - This conversation will take place Sunday, 2/28, 11 am PST. To watch, please click on http://standwithus.TV. I will be talking primarily on the anatomy of campus Zionophobia and how to neutralize it. For reference material, see my opeds:

2.22.2021 10:10pm - We have gotten used to hearing "commitment to Israel's security" but, today, it dawned on me: Isn't it really a submission to the idea Israel must live by its sword to the end of days? Why not "commitment to end the conflict" or "to end all threats" or "to peace education"?

2.22.2021 4:15pm - (Replying to @MatthewZGindin and @AllisonKSommer) I grew up in Bnei Brak and I know those Litvak Rabbis inside out. They turned super-Zionophobic "lights" after 6,000 Zionist kids sacrificed their lives to rescue them from the "Rivers of Blood" that Azzam Pasha promised us in 1947-48. Glad some of us still live to tell history.

2.21.2021 11:05pm - Michael Che an Idiot? Or is it the joke author? Or the Zionophobic producer of SNL? Neighborhood bully, standing on the hill Running out the clock, time standing still Neighborhood bully. (Bob Dylan, 1983)

2.21.2021 9:19pm - (1/ ) I'll tell you a story apropos causal calculus. Today I've installed Alexa app for my wife and tried to challenge her with the impossible: "Alexa! Read Book of Why." What do you know. She started with "The do-calculus, or mind over matter" on page 231, and her voice was so warm 1/
2.21.2021 9:20pm - (Replying to @yudapearl) and convincing, that I could not help but falling in love with do-calculus again, and decided that this is how I should present Causal Inference to the skeptics from now on. Try it, with Alexa, her voice is much more convincing than mine. & the historical context helps.#Bookofwhy

2.21.2021 1:43pm - (Replying to @MatthewZGindin and @AllisonKSommer) Zionophobic racists never get tired of teaching others what "true multicultural democracy" is, and how to escape dismantling or else; it's part of graduating a Zionophobia major, and trying out dismantling wars here and there.

2.21.2021 1:43pm - (Replying to @PHuenermund) "Arrow in network" is easy said when you have a network. Not so if you are sworn to avoid graphs. Try to say "Arrow" in Swahili.

2.21.2021 1:16pm - (Replying to @PHuenermund) "Arrow in network" is easy said when you have a network. Not so if you are sworn to avoid graphs. Try to say "Arrow" in Swahili.

2.21.2021 10:58am - Zionophbes are the only human species allowed to be openly racist; you can't deprive them of this privilege by taking the word "Zion" out of their rhetoric.

2.21.2021 9:55am - (Replying to @MathmoThe) Based on the overview I see a fair coverage of the basic material. However, I can sense some reluctance to adapting the First Law of Causal Inference, the do-calculus and the 3-level hierarchy of the Ladder of Causation. If I'm right, it can cause problems

2.20.2021 10:59pm - You should be grateful for not having to teach confounding using potential outcomes, where even top experts admit to be walking in total darkness. The "Crash Course in Good and Bad Control" may also be helpful

2.20.2021 10:32pm - What's the problem? Have you tried Primer chapter 3 Scott @smueller tells me he taught it in High School and the kids loved it.

2.20.2021 10:24pm - During my four years as a student at the Technion, the Bahai Temple in Haifa was one of the city's Jewels we treasured most. Colleagues and students tell me it has remained so -- a unique quest for the holy through beauty.

2.20.2021 8:14pm - @BristolJSoc , I will be joining you next Wednesday. I wish US students had your courage to tell their universities: Enough is enough! Compare their timid stand at UC-Irvine where the adm. deems the criminalization of Israel a "welcome" to Jewish students.

2.20.2021 5:57am - (Replying to @ianawren) You can mention, and you can soil but, as King Solomon said: For everything its season. Time to plant and time to uproot.... Time to tear and time to mend.... Time to admire the Acropolis and time to remember the slaves who built it.

2.20.2021 3:16am - (Replying to @ianawren) The mechanism of "thanksgiving" for the good we have ensures we will have more of it. The mechanism of soiling the good with the bad ensures little of the former.

2.20.2021 1:28am - (Replying to @ianawren) Doubly in awe of Western wisdom of filtering the noble from the ugly.

2.20.2021 1:19am - Science beautifully amplifies the courage that each of us possess when we ask our teachers: Are you sure?

2.20.2021 1:14am - A rare view of the Acropolis under snow. Always in awe: the cradle of democracy, of logic and of the scientific method.

2.20.2021 1:01am - A simple way of simulating Simpson's reversal is to use two structural equations, like eqs. [7-8] in, and making sure alpha<-beta*gamma. See section 3.1

2.20.2021 12:48am - (Replying to @emilykschrader) We had a neighbor called Trachtenberg, a son to a prominent Zionist family, with cultural roots that go back many generations, perhaps even to Abraham, Moses and King David. What would Trachtenberg the Zionist say to Trachtenberg the Zionophobe?

2.19.2021 7:26pm - Kenneth Roth has discovered a new logic for causation: "The conservation of lies." If a lie does not pass through the UN Security Council it must be pushed through other avenues until it does.

2.19.2021 4:13pm - The examples used in may be useful but his "resolution" of the Simpson's paradox is off. A "resolution" demands an explanation of why this reversal is deemed paradoxical for a century of statisticians and philosophers. See

2.19.2021 4:03pm - Perfect example. The grammar by which we wish to speak is always cleaner than the speech we actually produce. Correcting my grammar: "The principles by which we wish to live our lives are always nobler than the lives we actually lead."

2.19.2021 3:58pm - (Replying to @DanielNevo @itamarcaspi and 2 others) I can imagine people who would find challenge in further refining the distinction and add distinctions on distinctions. I cant join them because, in the nonparametric world every variable is presumed to be a moderator (or effect modifier).

2.19.2021 3:34pm - Sure. The principles by which we wish to live my lives are always nobler than the lives I actually lead. AI programs those principles.

2.19.2021 3:18pm - Hillarious!!! I think the US is catching up fast.

2.19.2021 7:55am - (Replying to @DanielNevo @itamarcaspi and 2 others) The term "moderator" is different from "mediator". The latter resides on the causal path, the former modulates the strength of that path from the outside.

2.19.2021 4:32am - (Replying to @DanielNevo @itamarcaspi and 2 others) The basic do-sentence reads: P(y|do(x), z) where x,y and z are all vectors. However, the interference problem (eg different contagious factors among different age groups) is still under-formalized.

2.19.2021 4:08am - I am glad they bring up causality as a major ingredient in creative writing, a deficiency of present day neural nets, but I am the last to say that computers will not grasp this in the future. They will.

2.18.2021 10:30pm - Very telling question: "Can you share the data?" My humble eyes see the data in Fig 1.1, with each circle representing one data point. But I suspect your question aims at something more profound; what is it?

2.18.2021 9:57am - (Replying to @petroniocandido and @ylecun) The dichotomy is not between individuals; it is a hard mathematical fact that you cannot jump from rung-1 to rung-2 unless you make certain assumptions. Individuals only vary in how seriously they take the mathematics and the assumptions.

2.18.2021 9:50am - You can get a solution manual for free, if you write to my assistant Enjoy.

2.17.2021 10:49pm - (1/n) I've noticed a sentence in your blog which might lead to some confusion. It says: "Causal Inference (a subset of Bayesian Networks)". The taxonomy I have in mind reads: (1) Bayesian Networks - Rung 1 objects. (2) Causal Bayesian Networks - Rung 2 objects, and (3) SCM - Rung 3 1/n
2.17.2021 10:49pm - (2/2) (Replying to @yudapearl) To explicate, (1) is a specification of a probability distribution, (2) specifies probabilities under (all) interventions and (3) a set of functional relations, each specifies an assignment, from which (all) counterfactuals can be computed. See 2/2

2.17.2021 5:20am - Thanks for the *emphasis* and spacing. I could not do a better job, and I stand behind those quotes.

2.17.2021 2:37am - (Replying to @Jabaluck @AdanZBecerra1 and 4 others) Now we are getting someplace. Once we disambiguate that nebulous "conceptual knowledge" we can do all kind of exercises and talk about "probability of ATE < p", "plausibility that confounding < q" and more. Each requires DAG-based logic. Sensitivity analysis is a good example.

2.16.2021 8:48pm - (Replying to @Jabaluck @AdanZBecerra1 and 4 others) Hence my answer stands: Without formalizing what you mean by "our contextual knowledge" the answer to your question is: It cannot be done. Because, if "context" is some mental DAG, the answer is trivial. If it is not, it could be astrology, premonition, fairy tale -- your pick.

2.16.2021 8:08pm - (Replying to @Jabaluck @AdanZBecerra1 and 4 others) If economists are trying to solve the problem of deciding which DAG is preferred, without telling us what knowledge they invoke in their "design", then I DO HAVE a solution to it. And the answer is "it cannot be done". This is not an opinion but a theorem. Logic is helpful.

2.16.2021 7:09pm - The purpose of SCM is not to "estimate a SEM model" but to answer causal questions, e.g., what is the efficacy of a given vaccination. Here, ( surprising to many SEM folks), you can get many quantitative questions answered just from the qualitative structure of the DAG plus data.

2.16.2021 6:06pm - (Replying to @jon_y_huang and @EpiEllie) THose SEM's that are justified by "fit stat" are not causal. DAGs/SCM incorporate causal assumptions, of course, but not consistency (its a theorem) and not faithfulness (which is needed for discovery, not for inference.

2.16.2021 6:00pm - (Replying to @NoahHaber @jon_y_huang and @EpiEllie) Same apply to SCMs. If all you need is prediction and retrodiction, use it as a statistical model. But if you want answer causal or counterfactual questions, the SEM literature is as helpless as Sewall Wright was. And I refer you to concrete questions

2.16.2021 5:52pm - (Replying to @Jabaluck @AdanZBecerra1 and 4 others) The only thing I keep insisting on is to have solved problems that have solutions and identified those that don't. If you are attempting to solve problems outside this range, I suspect your problem resides in the "don't" category, or it is not well-defined. "good/bad dag" is one.

2.16.2021 3:49pm - (Replying to @Jabaluck @dmckenzie001 and 3 others) So why is it so irritating to economists to accept that the reason they cross out umbrellas ---> rain is because they have a DAG in mind that says Umbrellas<--Clouds---> Rain and, importantly, they also have a logical engine there, telling them: The two are incompatible.

2.16.2021 4:43pm - (Replying to @Jabaluck @AdanZBecerra1 and 4 others) I find it unpleasant to communicate in a language laden with "bombastic claims" "you cannot build" "you cannot do" "you cannot this" and "you cannot that". Excuse me; this is an educational channel, not a horse market.

2.16.2021 4:37pm - (Replying to @VC31415 @Jabaluck and 4 others) It will take me a week to decipher that paper. Are the questions there so hard to summarize in a Tweet or two?

2.16.2021 4:31pm - (Replying to @Jabaluck @AdanZBecerra1 and 4 others) I have not seen such a well-defined problem yet, except those that causal logic tells me "you cant answer question Q given the information in model M". Do you really think that by repeating the phrase "you don't know" "You don't know" you are going to become convincing?

2.16.2021 4:17pm - (Replying to @Jabaluck @AdanZBecerra1 and 4 others) Whether something is "Potentially consistent with many underlying DAGs" requires a logic to confirm or deny this assertion. Welcome to the Causal Inference fold. You will not be disappointed.

2.16.2021 3:58pm - I come from Sewell Wright too, who is a hero of #Bookofwhy. Here are questions that Wright could answer: compared with questions he (as well as SEM folks today) could not:

2.16.2021 3:49pm - (Replying to @Jabaluck @dmckenzie001 and 3 others) So why is it so irritating to economists to accept that the reason they cross out umbrellas ---> rain is because they have a DAG in mind that says Umbrellas<--Clouds---> Rain and, importantly, they also have a logical engine there, telling them: The two are incompatible.

2.16.2021 3:35pm - (Replying to @Jabaluck @dmckenzie001 and 3 others) Beautifully put. The reason a human can do this could be that he/she already has a DAG in mind and can see incompatible with the one presented. AI can do it too. But you imagine a mind with a different model in mind, broader than DAG that, supposedly, economist have. Show one.

2.16.2021 3:25pm - (Replying to @PWGTennant and @EpiEllie) Non recognition of collider bias is only a symptom of neglecting the causal content of SEM, i.e., not asking the kind of questions that you find here:

2.16.2021 3:15pm - (Replying to @jasndoc) Of course! For many years. Peter is still in the opinion that SEM are just parsimonious and "meaningful" encodings of statistical information. He is reluctant to accept my assertion that what he means by "meaningful" is none other but "causal".

2.16.2021 3:10pm - (Replying to @ashtroid22 @pablogerbas and @EpiEllie) Hayes is awfully hazy on causation.

2.16.2021 3:08pm - (1/ ) The generalization from linear & parametric to non-parametric models is only ONE difference between SCM and traditional SEM. The main difference is disambiguating the causal content of the models and the tools developed to extract that content algorithmically. To witness, 1/
2.16.2021 3:08pm - (2/2) (Replying to @yudapearl) causal and counterfactual problems, pose insurmountable difficulties to SEM folks, even in linear models. See & 2/2

2.16.2021 2:55pm - The intent is the same, but the meaning of SEMs in #psychwitter is still debated by their users (see while Structural Causal Models (SCM) have been formalized and their causal content explicated w logical clarity (see

2.16.2021 2:35pm - (Replying to @Jabaluck @dmckenzie001 and 3 others) The disappointment is mutual. Instead of presenting ONE model that is BOADER than the DAGs variety, you hurl "self promotion" accusations at those who attempt to explain what the logic of causal inference can and cannot do for building convincing arguments for an estimator.

2.16.2021 12:31pm - (Replying to @FJnyc @sullydish and @jessesingal) I am sure Zionophobia plays a major role in the Barri-phobia pandemic. Zionophbes go nuts when their righteousness is examined on factual and moral grounds. The echo-chamber does not prepare them.

2.16.2021 7:48am - (Replying to @dmckenzie001 @jenniferdoleac and 2 others I like your survey of the confusion in the matching arena, but not the "Bottom line of all of these cases is the need to really understand the context well," which sounds like pre-scientific calls to "think hard." Today, "understanding the context" has been mathematized.

2.16.2021 4:34am - Of interest is the sizable gap between the top four and the rest of the world. Our global village is acting like it ain't a village.

2.15.2021 5:49pm - (Replying to @WaterFront8) Myself, and the great majority of Israelis (as well as American Jews) are atheists. To be a Jew means to see yourself as a carrier of a valuable collective tradition that you wish to pass on to future generations.

2.15.2021 5:38pm - (Replying to @rdisipio) Sorry, I am not familiar. But recall the beauty of the Ladder; it tells you that you do not need to become familiar with a proposed method to decide. If the method receives its input from the data only, then, regardless how clever it is, it will never be able to climb to Rung-1

2.15.2021 5:31pm - Prominent University College London Scholar Resigns After Academic Board Rejects Leading Definition of Antisemitism: ‘You Are All Going to Hell!’
2.15.2021 5:32pm - (Replying to @yudapearl) I am urging UCL's Provost, Dr. Michael Spence, to maintain the IHRA definition within UCL and ensure the safety of its Jewish students. I urge our colleagues at UCL to write to Dr. Spence on behalf of those students.

2.15.2021 6:50am - (Replying to @chriswolfvision and @tailcalled) My objection was practical: "While radical empiricism may be a valid model of the evolutionary process, it is a bad strategy for machine learning research." @tailcalled convinces me that it is more impractical than I thought. We cannot replicate past streams of cosmic radiation.

2.15.2021 3:31am - (Replying to @dela3499 @ks445599 and @LakeBrenden) It gives me the hope that, as much as these authors love the concept of "inductive bias", they do not consider causal models to be in the that category, due to the fundamental differences between the two.

2.15.2021 2:23am - This is the "radical empiricism" justification of ML culture. I have presented three reservations here:

2.14.2021 9:41pm - (Replying to @yudapearl @westurner and @UCBIDS) Why do I doubt it? Because "understanding a phenomenon" means understanding the causal ropes behind that phenomenon and that necessitates a language to describe those ropes. Such a language, best I know, is not taught in standard CS classes. It should.

2.14.2021 9:35pm - (Replying to @westurner and @UCBIDS) I looked into Data 8: The Foundations of Data Science and was intrigued by this sentence: "Given data arising from some real-world phenomenon, how does one analyze that data so as to understand that phenomenon?" This is indeed what SHOULD be taught in DS; but is it?

2.14.2021 9:21pm - (Replying to @MaxIlse) @MaxIIse , did you really get this comment from a reviewer? If so, the reviewer was brilliant in summing up the philosophy of CI in one line. With one twist added: replace "and argue" with "and shows those part of". Completeness results in CI allows you to identify those parts.

2.14.2021 7:16pm - (Replying to @westurner) I am not sure the author of "Inductive Logic" on Stanford Encyclopedia has read #Bookofwhy and why Bayes theorem is more than a trivial theorem in probability theory, implied by P(A,B)=P(B,A).

2.14.2021 2:47pm - (1/ ) Today we celebrate 125 years to the publication of Theodor Herzl's Der Judenstaat (Vienna, Feb 17, 1896,, the first call for the restoration of a sovereign Jewish homeland. To appreciate the explosive impact of this 86 page pamphlet we read that only 1/
2.14.2021 2:47pm - (Replying to @yudapearl) 18 months later, 208 delegates from 17 countries gathered in Basel for the pivotal First Zionist Congress, after which Herzl wrote in his diary "In Basel I founded the Jewish state." I have commented on that historical Congress here:

2.14.2021 2:15pm - As an Israeli-American, I am happy to see that Israel has ranked top 7th in Bloomberg Magazine "most innovative country", and the US ranked 11th. A bit disappointed though that Israel slipped from 6th place in 2020, and the US slipped two ranks, from 9th.

2.14.2021 11:49am - Please share if you get to understanding it.

2.14.2021 10:39am - (Replying to @charleskfisher and @NikosNi41644559) Let me put it this way. If your task is to tell cats from dogs you wouldn't use one "inductive bias" for cats and one for dogs -- that would be cheating. Similarly, if our task is to decide if the effect of X on Y is positive or negative, we must use the same inductive bias.

2.14.2021 9:56am - (Replying to @tdietterich @arxiv and 2 others) Concur!

2.14.2021 9:48am - (Replying to @NikosNi41644559) But it still remains the same over all problem instances. Occam's razor, for example, prefers simple theories over complex ones, a preference not imposed by the data, but does not change this preference when move to from one problem instance to another. Any example where it does?

2.14.2021 9:42am - (Replying to @charleskfisher and @NikosNi41644559) I am relieved. So, a CNN used in my kitchen would have the same "inductive bias" as the one used in my backyard. This is what I suspected, uniformity over problem instances. Not so "model-based" inference. Each problem demands its own model to start the inference engine going.

2.14.2021 9:35am - (Replying to @RaulMachadoG @PHuenermund and 2 others) Can you elaborate? What movement and kind of "leaving"? Am I going to be "canceled"?

2.14.2021 9:30am - (Replying to @charleskfisher and @NikosNi41644559) So, when you go from scene to scene, some person tweaks the "bias" afresh? Or you encode this prior knowledge once and for all, for all anticipated scenes?

2.14.2021 9:26am - Legitimate question! The list is quite long, so I'll provide a link to all our Lab's papers; all titles containing the terms "external validity", "generalizability", "selection bias" "data-fusion" etc. are part of the "algorithmization of EV." Enjoy

2.14.2021 9:14am - (Replying to @StillTr05207382) True, but some ML folks may argue that they can represent causal models in NN. After all, both are made of "functions". CI folks would argue that the functions learned by NN are among features in the data, not about reality. And so the arguments continue, decade after decade.

2.14.2021 4:16am - (Replying to @dataengines and @peabody124) The question is whether this method should qualify as "model discovery" or "inductive-bias". The distinction is not just a matter of nomenclature, because each term is charged with tradition and a set of tools and principles.

2.14.2021 4:02am - Thanks. This url seems to work.

2.14.2021 4:01am - (Replying to @NikosNi41644559) So how about if one stores Answer-1 for situation-1, Answer-2 for situation-2, etc. for all situations. We wouldnt need data then. Is it an "inductive bias"?

2.14.2021 3:34am - I am reading a recent paper on external validity (EV) file:///C:/Users/Judea/Downloads/ValidityProject_v4%20(3).pdf and find it hard to understand why one should continue to explain the necessity of EV when it is fully defined, formulated and algorithmitized.

2.14.2021 2:48am - (Replying to @NikosNi41644559) But it still remains the same over all problem instances. Occam's razor, for example, prefers simple theories over complex ones, a preference not imposed by the data, but does not change this preference when move to from one problem instance to another. Any example where it does?

2.14.2021 1:57am - (1/ ) The difference between "inductive bias" and "model-based" is that the former is a UNIVERSAL preference function; it remains the same from problem to problem. All the examples in Wikipedia are of that character, eg. minimum description length. In contrast, model-based inference 1/
2.14.2021 1:57am - (Replying to @yudapearl) the preference function varies over problem instances. The same data would produce different answers to the same question, depending on what problem-specific model is invoked. With this distinction in mind, ML + induction-bias do not cover causal inference. 2.14.2021 1:57am -

2.14.2021 1:40am - (Replying to @NikosNi41644559) So, according to this definition, NOTHING is beyond the capabilities of ML. Even if a task requires information that is NOT in the data, we will call that information "inductive bias" and, OOPS, it suddenly becomes part of what we can learn from data. Why not call it a "model"? (Replying to @StillTr05207382)

2.13.2021 11:39pm - (1/ ) Help! What precisely is "inductive bias"? Some ML researchers are in the opinion that the machine learning category of ‘inductive biases’ can allow us to build a causal understanding of the world. My Ladder of Causation says: "This is mathematically impossible". Who is right? 1/
2.13.2021 11:39pm - (Replying to @yudapearl) The definitions and examples given in,function%20is%20actually%20the%20best tell me "the ladder is right". But perhaps there is some new, broader definition that embraces causal models and other types of knowledge in its arms?? Is there? If so, what world knowledge is NOT "inductive bias"?

2.13.2021 11:16pm - (Replying to @shaayohn and @TanviSunku) Please share a blow or two.

2.13.2021 10:56pm - Hard to believe that the Christian community has not managed to prevent this moral perversion, WCC, from speaking in its name for almost a century:

2.13.2021 10:19pm - In memory of Ilan Halimi, I am re-tweeting a eulogy I wrote in 2006 Any connection to modern day madness is purely God's doing, not our responsibility.

2.13.2021 9:50pm - I am in receipt of a draft of new ML book "PATTERNS, PREDICTIONS, AND ACTIONS: A story about machine learning" by Moritz Hardt and Benjamin Recht. Beautifully organized, with clear connections to the origin of ideas, and even 2 chapters on causality.

2.13.2021 9:03pm - (Replying to @DavidHirsh and @LFischerBerlin) The situation described by @LFisherBerling is precisely the reason I have long stopped fighting the cesspool of anti-Semitism and have targeted instead a less evasive cesspool called: Zionophobia, no part of which can be deemed "legitimate".

2.13.2021 8:42pm - (Replying to @DavidHarrisAJC) The unique color that Judaism adds to this universal principle is to enrich the meaning of "hateful to you" through collective memories. See

2.13.2021 8:26am - (Replying to @samagra_sharma and @mdgiitr) Your brother must be extra bright. Try to explain it to a Potential Outcome disciple (say and you'll find a brick wall.

2.13.2021 7:56am - (Replying to @quantumciaran and @ciaran_lee) The pleasure was mutual. Can I (or you) share with readers the pdf version you sent me, or we would get arrested by the copyright vice squad?

2.12.2021 6:59pm - This is the first time I am reading the New Scientist article by @ciaran_lee : I recommend it for readers who are just about ready to dive into causal inference, and those who already know that causality is not important for AGI.

2.12.2021 5:40pm - Putting on my Rabbi hat:

2.12.2021 8:53am - (Replying to @yudapearl @bariweiss and 2 others) Apropos, here is how I found out that pride really works:

2.12.2021 1:09am - (Replying to @bariweiss @IzaTabaro and @NellieBowles) Pride certainly works and, although you need no extra courage from me, I would nevertheless tell you: I am so proud of you and proud of belonging to a people that has given a cause to leaders like you.

2.12.2021 12:53am - What UCI Chancellor Gillman did not state is: “For many Jewish students, Zionism is an integral part of their identity and their ethnic and ancestral heritage. These students have the right to openly express identification with Israel." A quote from a statement by the University 2.11.2021 11:30pm - (Replying to @dynamite_ai) I am arguing for model-based data-science, of which causal inference is one example, and a successful one, from other model-based domains can learn.

2.11.2021 10:13pm - What we saw at @ASUCI last night is partly a result of our fixation on anti-Semitism as a litmus test of moral deformity. Who cares if BDS is or isn't anti-Semitic. The question should be is BDS aiming at dismantling a homeland of a nation. Are they or aren't they Zionophobic?
2.11.2021 10:13pm - (Replying to @yudapearl) By fixating on anti-Semitism the AJC's survey has given the UCI administration an opening for avoidance and inaction. I'm waiting for @AJCGlobal to ask UCI's Chancellor to state explicitly whether Israeli and Zionist students and faculty are welcome on UCI campus. That's his job.

2.11.2021 7:55pm - (Replying to @AstralFleck) I am not meeting the new crop. That's why I am more concerned with the prominent "scholars" who divert public resources to "centers" and "institutes" in which data-fitting is the ruling paradigm.

2.11.2021 7:28pm - (1/ )Ideally, ML-scientists should use data-driven methods to refine science-based models! True. But the way it is practiced today, Data Science is dominated and guarded by model-blind researchers and educators. Just examine the "data-science center" on your favorite campus, and 1/
2.11.2021 7:28pm - (2/2) (Replying to @yudapearl) ask how many classes it offers in model-based inference compared with curve-fitting methods. Or how many of their instructors can solve a toy problem in Causal Inference. Data-fitting in addictive, and building another "data-science center" only worsens the addiction. 2/2

2.11.2021 2:13pm - UCI Student Senate Passes BDS Resolution via @jewishjournal
2.11.2021 2:13pm - (Replying to @yudapearl) The only words that come to mind: Well, the neighborhood bully, he’s just one man His enemies say he’s on their land They got him outnumbered about a million to one He got no place to escape to, no place to run He’s the neighborhood bully Bob Dylan 1983

2.11.2021 1:10pm - If they can do it, they have accomplished the first step into "deep understanding". The next is to imagine you dont have a full causal model of the world and ask: What must I assume so as to answer a given question. Reading #Bookofwhy would help, and so

2.11.2021 12:58pm - (Replying to @Simpl3_is_Hard and @DavidDeutschOxf) SCM stands for "Structural Causal Model" see or #Bookofwhy.

2.11.2021 11:21am - Passing or withdrawing a resolution matters not. What matter is having a stage, a microphone, and a discussion about the racist character of an accused. The "accused" in this case should be student organizations that malign identity symbols of other students. (eg Koran, Israel).

2.11.2021 11:07am - (Replying to @ruescasd) Yes, they also requires "unity", which is hard to understand in the propositional case, and which I am not sure would cover "undoing".

2.11.2021 10:21am - (Replying to @DavidDeutschOxf and @yaobviously) "Making sense" (in my sense) begins indeed with a problem: "I can't answer a bunch of questions about a world which happened to generate the sensory data I have." If the questions cover all 3-Rungs of the ladder (and theory is needed to answer them) I feel "things make sense"

2.11.2021 10:07am - (1/ ) They jump straight into full predicate calc. formalization, making it difficult for lay persons to test their notion of "making sense" on simple examples. I have proposed another criterion for "making sense" (called "deep understanding" which I am 1/n
2.11.2021 10:07am - (2/ ) (Replying to @yudapearl) not sure would be satisfied by the proposed triple "prediction, retrodiction and imputation", for it's missing "undoing" (Rung-3). If any reader understands the AIJ paper, please enlightened us how it would make sense of the firing squad example. Would prisoner P still be dead 2/
2.11.2021 10:07am - (Replying to @yudapearl) had soldier A decided not to shoot? Prediction-Retrodiction-Imputation can be accomplished in probabilistic Bayes Network (non causal); not so "doing" and "undoing". "Making sense" should handle "undoing".

2.11.2021 5:01am - (Replying to @SlobodanIsFree and @PhilSciComm) Testing ideas in a laboratory is speedier more informative than waiting for natural phenomenon to offer an adequate test. Data Science is our laboratory. ML programs can be taken apart and reassembled, modified and repaired, to tailor the test to the idea tested.

2.11.2021 4:52am - Fascinating talk which I could not stop listening to, till the very (bitter) end. I have only one reservation: Our ability to program a Popperian-spirited program is not as far of as @DavidDeutschOxf argues. I actually think SCM is a mini embodiment of such program.

2.11.2021 3:47am - An incredibly courageous article by @GilTroy , one of the most insightful historians of our generation.

2.11.2021 2:11am - (1/ ) On objectivity. Note that whereas causal assumptions may seem less "objective" than data, the question whether a set of assumptions is sufficient for a given task is no less "objective"; the answer to this question does NOT depend on subjective assumptions. It's purely logical.1/
2.11.2021 2:11am - (1/ ) (Replying to @yudapearl) Therefore, the entire CI framework can be seen as an exercise in logic, rather than a request to believe in one's assumptions. The logic, however, is not classical logic, but causal logic, based on SCM semantics. See Sure-Thing Principle

2.11.2021 1:45am - (Replying to @conjugateprior) I would strengthen it by replacing "some causal knowledge" with "an answer to a given causal question".

2.11.2021 12:39am - Data Science is the arena in which philosophical debates will be tested and decided, between the rationalists, empiricists and the in-betweenists.

2.10.2021 11:00pm - (Replying to @isaacdecastrog) The problem is not in *passing* such resolutions; they cause irreparable damage even when defeated, poisoning gullible students with fabricated anti-Israel accusations. The question is why we do not hear of counter-resolutions, exposing the racist character of the fabricators.

2.10.2021 8:57pm - (Replying to @AstralFleck) The pleasure is mutual but, to be honest, I don't have a strong enough word for the model-blind ML world and its race towards the unknown. While we were tweeting, a new building was donated for another Data-Science Center in some well-meaning university.

2.10.2021 8:13pm - I should note that such questions are beyond the standard language used by ML authors, since the notion of a "world" existing outside the data, approving and disapproving conclusions emanating from the data is alien to ML, where the criteria of performance too reside in the data.

2.10.2021 6:47pm - Nice perspective on modern Data Science. Thanks also for the refs. to Coveney's articles on Big Data.

2.10.2021 6:43pm - (Replying to @MichaelALewis10) I didn't mean to be tough. I meant to apologize for not spending time on searching for the golden nuggets, when modern philosophers cover up those nuggets with disputations on "how philosopher X differed from philosopher Y". My weakness, not theirs.

2.10.2021 12:12pm - Maybe. But, honestly, the (subjective) likelihood of finding a nugget of relevant idea per hour of sifting through mountains of irrelevant philosophical disputations is fairly low. Accordingly, the theory of rational exploration tells me to rely on experts for those nuggets.

2.10.2021 12:03pm - (Replying to @HomoModulans) Highly relevant, great quotes and accurate historical perspectives. I wish you would relate these arguments to modern Data Science and ML, as I tried to do here, but I could not penetrate your walls of sign-ins, passwords, etc....

2.10.2021 10:33am - A similar distinction divides ML and CI styles of research. ML papers rarely ask: What should the world be like for my algorithm to work, or, "Is my task doable?". These questions are the key for CI papers, eg. "Are the model assumptions sufficient for answering a given query?"

2.10.2021 9:27am - (Replying to @yskout) Yes, ecological equations constitute a model from which counterfactual queries of attribution can be answered, by simulation or analysis.

2.10.2021 1:28am - (Replying to @yskout) The key phrase is "death due to exposure", which is a counterfactual, not statistical notion.

2.10.2021 9:22am - The difference between classical and modern day contests among empiricists and rationalists is that the latter are raging under the microscope of mathematical scrutiny, so claims of what models are necessary for a given task are supported by mathematical proofs and demonstration.

2.10.2021 1:20am - Anyone who has an even mild interest in understanding the new realities in the Middle East should watch this clear, concise, no BS talk by @EinatWilf . An absolute must.

2.9.2021 10:08pm - (Replying to @DanSeligson) I proposed it as a definition of "deep understanding" here:, not necessarily scientific.

2.9.2021 10:05pm - Alarming findings. Anyone read the technical article to find out how they define "attributable fraction"? (I could not go past the paywall.) See Causality chapter 9 on "causal attribution."

2.9.2021 9:53pm - I didn't realize that. So now we have two solid links to the "microscope" papers. Fairly good papers! I just read them again. Reading these would put you ahead of the average stat PhD student at U. of Stockholm. Perhaps even Stanford (correct me if I am wrong, @StanfordHAI)

2.9.2021 9:34pm - (Replying to @Timothy_D_Sawe) This is a good example of a system that explains an output in terms of the data-fitting strategy it was programmed to follow. An extreme example: Q. Why was my loan denied? Ans. Because we used Dirichlet priors.

2.9.2021 8:11pm - Are you asking for links to the "linear microscope" papers? If so, I'll try again:, and They just worked for me.

2.9.2021 5:50pm - Sure enough, while tweeting on the impossibility of model-blind "Data Science" I get a proposal for "Undergraduate Data Science Minor at UCLA." Why? Mounting pressure from students, industry, Dean, Chancellor: "UCLA needs to place itself on the map". Join or perish!

2.9.2021 2:15pm - (Replying to @markcannon5) I personally have not looked into spatial relationships, though some people at Stat UCLA used causal models for interpreting visual scenarios.

2.9.2021 12:53pm - (Replying to @wster) Popper's refutation theory is all over the place, though in spirit only. I am not sure, however, that Popper was explicit about the source of our conjectures and about the profound mismatch between the language of conjectures and the language of data, with which we refute them.

2.9.2021 12:02pm - (Replying to @jgalgarra) Great question! We often have just a sketchy model (eg a DAG) and we need data, and ML, to fill in the missing parts of the model. Not for the sake of getting a complete model, but just enough to be able to answer important research questions. @Bookofwhy is all about this game.

2.9.2021 11:06am - Machine learning folks were way ahead of me begging for "explainable ML". But I am not sure even today they have internalized that there is no such a thing as "explainable" program w/o a model of reality. You can explain the program's data-fitting strategy but not why it matters.

2.9.2021 9:12am - (Replying to @borgesvit_r and @szollner1) Isn't Musk tweet the event whose effect we are trying to predict? If so, its predecessors and determiners should be considered.

2.9.2021 9:07am - Thanking Steven Pinker for bringing my "Radical Empiricism" post to the attention of readers who otherwise won't see it. I wish to invite more ML students to take part in the discussion, as Yoshua Benjo did -- it's their future that is at stake.

2.9.2021 12:01am - Speaking about "Good and Bad Control", I would like to recommend arXiv:1907.02435 (by Henckel etal) which provides graphical criteria for deciding when one adjustment set results in a lower variance than another (in linear systems).

2.8.2021 5:32am - (Replying to @aminsaadou) That was the intention. Unfortunately, many took econometrics to mean an invitation to introduce statistics into economics. So much so that causality almost disappeared, as Kevin Hoover lamented: "Where have all the causes gone?"

2.7.2021 9:16pm - (1/ ) I would start with structural equations, where everything is meaningful and in line with human intuition. Then I would raise the question: So how do we estimate those marvelous structural coefficients from either experimental or observational? This motivates conditional 1/
2.7.2021 9:16pm - (2/2) (Replying to @yudapearl) expectations, and partial correlation and so on. This is basically the road taken in the "microscope" papers, except that more time is spent on Wright's Rule, which allows you to ask at any point: suppose we only want to predict, not to act.

2.7.2021 8:07pm - Try this link:

2.7.2021 7:57pm - I have two gems for you: and I wish I could refer you to a decent textbook, by a younger author, doing a comparable job, but we live in a world where education lags behind the time.

2.7.2021 7:27pm - So who would be most qualified to teach regression? Theoretically, it should be economists, who are using both structural and regression equations, hence should be skilled in switching between causal and statistical interpretations. Some indeed are, but:

2.7.2021 7:09pm - (Replying to @jmhessel) I've always felt sorry for students taking linear regression class from a statistician. The tormenting tension between statistical and causal interpretations cannot be properly handled by even the best of instructors, unless he/she took a class in causal inference. Few do.

2.7.2021 3:15pm - WATCH: Four Israeli scientists unexpectedly win Oscar for contribution to film
2.7.2021 3:15pm - (Replying to @yudapearl) To my Engineering colleagues and students: Mt. Olympus is not beyond our reach. To my Israeli colleagues and students: You make me mighty proud.

2.7.2021 2:25pm - (Replying to @CrypTomer_pers) We have a lot in common, so, I do not understand why you regard our human story as "promoting a political view". It is an important story to tell to a world that has ceased to view Israelis as human beings. And I am saying it not as a rhetorical tool but as a painful fact.

2.7.2021 1:05pm - The greatest map produced in 2020 and the greatest victory of causality research.

2.7.2021 12:55pm - (Replying to @CrypTomer_pers) Of course I am not literally afraid of being arrested in the Hague and, of course, I used my arrest poetically, to call attention to the absurdity of the whole ICC "Israeli war-crime investigation". In this narrow sense I think I am acting quite rationally.

2.7.2021 12:31pm - (Replying to @CrypTomer_pers) What makes you so sure they won't arrest me? Have they acted rationally in the past? Are we so sure of the purity of their intentions throughout this circus?

2.7.2021 10:58am - As a former Israeli, I am guilty of a few war crimes, including: (1) Being born (2) Born free (3) Serving in the IDF to defend my first two crimes. Will they arrest me in the Hague? Glad Zoom protects me when speaking for Dutch conferences. But how long?

2.7.2021 7:46am - Proud to be a (honorary) alum of @UofT , not too proud of @nyuniversity who gave a "President Award" to SJP, a registered student organization that prides itself on interrupting meetings of other student organizations.

2.6.2021 9:32pm - (Replying to @eliasbareinboim and @VC31415) Thanks for reminding me that we were also guilty of using the term "research design" in the PNAS paper. But luckily we have defined it formally, leaving no room for ambiguity. In particular, we did not confuse it with "modeling assumptions" or "identification strategies".

2.6.2021 6:38pm - (Replying to @VC31415 and @jondr44) I believe you are asking for a simple "demonstration"? Also, it's important to specify "standard error" of what? Of the treatment effect? or of the covariate-specific effect?

2.6.2021 11:49am - (Replying to @hendersonhl22) Imbens paper only makes the mystery more puzzling (see, and saying: "isn't helping anything" isn't helping either, unless one explains how prominent economists find SCMs less than essential for deciding Good from Bad controls

2.6.2021 11:25am - (Replying to @artistexyz @VC31415 and 2 others) There must be some fundamental difference between "Research Design" and "Choice of model", else Rubin would not write "Design trumps Analysis" (critiqued here and seekers of "credibility" would not worship "design" over models.

2.6.2021 7:00am - Part of the problem is that we call them "antizionists", as if they follow some legitimate political ideology worthy of an 'ism' title, rather than an obsessive moral pathology. To get the record straight, I've been advocating "Zionophobia" Use it.

2.6.2021 6:29am - (Replying to @LoosyinTheSpace @hrw and 2 others) When you are under a 73 year of besiegement by your neighbors and told in no uncertain terms that whatever you do you will eventually be wiped out, you are likely to make mistakes, including mistakes that impede, or seem to impede your main goal: 2-states, equally legitimate.

2.6.2021 5:55am - The next fun thing is to explain why prominent economists, from Heckman to Angrist, still find this framework useless if not threatening. Theories anyone?

2.5.2021 4:27pm - (Replying to @KLdivergence) Thanks. But the novice (me) is still wondering if RAI is (1) A method (2) A collection of methods dealing with the same problem, or (3) A society of researchers challenged by the same type of problems, or (4) none of the above or (5) all of the above.

2.5.2021 12:10pm - I can't think of a clearer guidance than the one provided in our "Crash Course in Good and Bad Control" What perhaps is misunderstood in the PS literature is the equivalence of PS and adjustment. See

2.5.2021 9:10am - (Replying to @PhilHaile @VC31415 and 2 others) Having difficulty parsing your last sentence.

2.5.2021 9:02am - (Replying to @PhilHaile @VC31415 and 2 others) I think it makes people LESS careful. Once you label what you are doing "design," you endow it with an aura of authority, which makes it less likely to be scrutinized for validity. See Rubin's "Design trumps analysis..."

2.5.2021 8:52am - (Replying to @MichaelALewis10) Almost. Those variable should block all backdoor paths, true. You may want to include more, but blocking is both necessary and sufficient.

2.5.2021 7:03am - U.S. embassy in Jerusalem wins bipartisan Senate support in near-unanimous vote
2.5.2021 7:03am - (Replying to @yudapearl) This week has been begging for a piece of good news: Here it is! 97-3 vote in favor of making honesty a permanent fixture of US foreign policy. The exception, as expected, is Bernie Sanders, who must appease his voting base, for whom honesty is still secondary to populist slogans

2.4.2021 11:54pm - (Replying to @VC31415 @PHuenermund and @paulgp) I would be grateful even for a semi-precise definition or just clarifying which "choice of method" IS NOT "research design". Reading through the "credibility revolution" literature I've noticed that replacing every "design" with "my favorite method" makes things perfectly clear.

2.4.2021 11:02pm - Congratulations to Karthika Mohan @Carthica on her first academic appointment. Lucky are students and researchers in the State of Oregon where the face of AI will not be the same. Please share your next breakthroughs with us, on this educational channel.

2.4.2021 7:12pm - Thank you @BrumerDelilah and @GabrielleLashley (GABRIELLE LASHLEY) for bringing the latest developments to the awareness of students at the Daniel Pearl Magnet High School, Balboa Lake, California. Your excellent newsletter @ThePearlPost has been a source of comfort for years.

2.4.2021 12:25pm - Just posted, the video of Jake Tapper's Daniel Pearl Memorial Lecture:

2.4.2021 11:22am - This is very gloomy outlook, because I believe causality will never be incorporated into ML (educational blindness), ML must be brought to causality.

2.4.2021 8:52am - (Replying to @RaulMachadoG) By "individual level" I mean statements about a person named Joe. What data or assumptions we use to validate such statements is a separate issue. See for example what we can say about Mr. Joe here: page 95.

2.4.2021 8:43am - Confession: I re-read it in stages too, at least once per week. Not because I forgot what we said there, but because we said it nicer than my brain can say it today.

2.4.2021 12:12am - Have you looked into counterfactual analysis, exemplified here: Rung-3 of the Ladder of Causation is all about individual level inference, and covers attribution, causes of effects, probabilities of necessity and more. See also

2.3.2021 11:25pm - (Replying to @StacyPatton89) Am I in favor of a two-state solution? Yes, I am, but only if it is accompanied with "end of conflict".

2.3.2021 10:58pm - An organization that can't decide if Israel should live (2-states) or die (1-state) can't pose as a human right entity because, by my latest count, there are 8 million human beings on the Israeli side, the great majority of whom recall one's right to life & self determination.

2.3.2021 8:11pm - Hard to understand how organizations like @hrw @KenRoth and @BDS can continue to pose as "merely opposing policies." If they're not aiming for Israel destruction, they could easily say so explicitly: "2-states, equally legitimate and equally indigenous." Why haven't they? Why?

2.3.2021 6:32pm - Sharing Rep. Brad Sherman @BradSherman letter to the Ambassador of Pakistan @GovofPakistan in support of our appeal for Review of the Supreme Court ruling.

2.3.2021 11:00am - Why do we keep doing this? I tried to find an explanation and the only one that makes sense to me is: Because the authors of stat textbook never solved a causal problem from beginning to end.

2.3.2021 10:30am - Good point. This is another reason why Holland's "problem" is neither "fundamental" nor a "problem".

2.3.2021 9:14am - Thank you, Bernard Henri Levy @BHL for your insightful and informative article on behalf of justice and decency @JusticeForDanielPearl

2.3.2021 9:02am - (Replying to @urfriendlen @JohnMullahy and @lewbel) Why leave us hanging in uncertainty? What do the "Makarov bounds" tell us? Do they improve the bounds derived in Tian and Pearl (2000) ??

2.3.2021 8:02am - (Replying to @lewbel) We agree here, but I would articulate it in simpler terms: The fraction who benefit is a property of the Structural Equation Model, and so are all marginals and joint distributions of PO's. Random assignment reveals the marginals, not the joints. That's why I love SEM (not Roy).

2.3.2021 6:19am - (Replying to @ThatMarkElliott) Not many scientific endeavors see themselves as sole heirs to philosophy of science, requiring students of other endeavors to take stat-101.

2.3.2021 5:54am - Totally agree. The "fraction of beneficiaries" is a counterfactual notion captured by PNS (see As to "would buy treatment", it looks associational, unless you mean "would buy treatment after its efficacy is studied and becomes public"

2.3.2021 5:41am - (1/ ) Your question enticed me to ask: Why is it so hard for statisticians (and ML folks) to accept the fact that they can't estimate "effects" with Stat alone. My theory: Statisticians are brought up to believe that they are the executors of the scientific method, as outlined by
2.3.2021 5:41am - (2/ ) (Replying to @yudapearl) philosophers of science from Aristotle and Hume to Popper and Reichenbach. As top executors, it must be traumatic to accept that you need to go outside your textbooks to do simple things such as estimate effects. I know how hard it is because I spent two decades in that
2.3.2021 5:41am - (3/3) (Replying to @yudapearl) paradigmatic enclave -- it's addictive. That is why I recommend: Don't start causal analysis with stat -- you will never snap out of it; start with the logic of causal analysis. To be honest, I should really say: "Dont start stat with stat". We will get there in 5-10 years.

2.3.2021 5:06am - Glad University of Baltimore teaches Causal Analysis as an independent topic, not as a curious addendum to Stat or ML.

2.3.2021 4:58am - (Replying to @andrewcbancroft) Appreciating your input, and will share it with Dana.

2.3.2021 4:54am - (Replying to @tiagoandresvaz @EpiEllie and @rkuchen) Sure, we need statistics to lift us from finite samples to distributions, but we must lean on causal assumptions (ie DAGs) to lift us further, from association to intervention. The inference engine delineates what we can do w/o stat/data, and that's a lot.

2.2.2021 10:07pm - Thank you, Jake, for honoring our son Daniel with your lecture today, and for everything you and the journalistic community are doing to bring him justice and keep his legacy alive.

2.2.2021 9:59pm - (Replying to @tiagoandresvaz) You can't use the g-formula unless you have a causal graph to tell you which factors remain in the product. But once you have a causal graph, you are in causality land, outside the statistical enclave. [I heard that some folks teach g-formula w/o causal graphs - a grave mistake]

2.2.2021 6:17pm - The ladder says: NO. No statistical test, no matter how clever, no matter what statistical relationships it exploits, be it odd ratios, likelihoods, variances, entropy, mutual information etc etc. can predict the efficacy of treatment. Hard to swallow, for most statisticians.

2.2.2021 2:26pm - (Replying to @JCornebise and @tdietterich) Yes. People interested in programming free-will, rather than delving into the history of philosophical mystification of free-will will be curious to know if he adds anything to my humble 3 pages.

2.1.2021 7:13pm - One of the best articles I have ever read on the Middle East conflict, perhaps THE best. Reality vis a vis wishful thinking. Peace goals versus populist clichés . Highly recommended.

2.1.2021 5:18pm - (Replying to @econshishir and @_MiguelHernan) I havn't taken this course, but from what I recall, the Harvard group lumps together Rung-1 and Rung-2 of the Ladder of Causation. This would confound quantities computable from interventional studies with those that require counterfactual information -- causing confusion.

2.1.2021 11:52am - Yes, I addressed this question here:

1.31.2021 10:00pm - (Replying to @solvay_1927) If you can't validate the "toy" models, do you think you can validate the messy "real-life" problem, for which you have no model? Unless you give up on "validation" altogether and settle for something weaker. If so, why not apply same "weaker something" trick to the toy model?

1.31.2021 9:44pm - (Replying to @Jabaluck) It is answered by that logic, and it goes beyond "Appeal to substantive knowledge", but since you already know that it is "totally incorrect", there is no point in trying to communicate it. Communication only works among those who are prepared to learn something new.

1.31.2021 9:01pm - To Ruth and Judea Pearl, from Pakistan with love via @jdforward
Sharing a moving article by Amal Khan, a Daniel Pearl Journalism Fellow for 2016, which also sheds light on how Pakistanis feel about the surrealistic decision by the Supreme Court.

1.31.2021 8:02pm - (Replying to @Jabaluck) Whereas "most toy problems are not very useful" the ones I have presented have spawned general methods for: identification, mediation, testable implications, external validity, missing data, selection bias, causal discovery, attribution, personalized decision - "Not useful"?

1.31.2021 7:45pm - (Replying to @Jabaluck) What you call "the question at hand" is answered by the logic of causal inference which we discussed at length elsewhere. Here I call attention to analogies which are not "coarse" nor "dispensable", unless one is giving up on solving the 18 problems in

1.31.2021 7:19pm - (Replying to @Jabaluck) I would like to spend some time with readers who are more interested in what we can learn from the history of science.

1.31.2021 6:18pm - (Replying to @Jabaluck) I am happy that your students are not among the econ. leaders who shun and continue to avoid this important kind of problems. And if they also understand the First Law, they will be the leaders of next generation economics.

1.31.2021 6:08pm - (Replying to @Jabaluck) I explicitly said "the kind of understanding gained by solving OUR "toy problems" is essential". "Our" means the problems that have been presented in the CI literature in the past 2 decades and which are still shunned by most econ. leaders (unlike their creative students).

1.31.2021 5:55pm - (Replying to @Jabaluck) Econometricians always did things first, and it is important for them to maintain this professional pride, fine. But I am talking about SOLUTIONS to specific problem sets, of the kind presented in: If your students can solve them, economics has a future!

1.31.2021 5:40pm - (Replying to @Jabaluck) I said "essential," not "can be useful".

1.31.2021 5:32pm - (1/ ) Sharing an interesting observation from Frank Wiltzeck's book "Fundamentals." In the 17th Century, while the entire scientific world was pre-occupied with planetary motion and other grand questions of philosophy, Galileo made careful studies of simple forms of motion, e.g.,
1.31.2021 5:32pm - (2/ ) (Replying to @yudapearl) how balls roll down an inclined plane and how pendulum oscillate. To most of Galileo's contemporaries such measurements must have appeared trivial, if not irrelevant, to their speculations on how the world works. Yet Galileo aspired to a different kind of understanding.
1.31.2021 5:32pm - (3/ ) (Replying to @yudapearl) He wanted to understand something precisely, rather than everything vaguely. Ironically, it was Galileo's type of understanding that enabled Newton's theory of gravitation to explain "how the world works".
1.31.2021 5:32pm - (4/4) (Replying to @yudapearl) about the virtues of "toy problems," and the contempt with which economists (and others) view such problems (see I still maintain that the kind of understanding gained by solving our "toy problems" is essential for next generation causal inference.

1.31.2021 4:22pm - Confirming my second point. The "First Law" is absent in both courses and even the "Second Law" (d-separation) is nowhere to be seen. Coursera is hungry for a modern course in causal inference. Volunteers?

1.31.2021 4:12pm - (Replying to @ir_hafidz and @CarnegieMellon) How can one possibly mistakenly understand DAGs when DAGs is what we have in our mind when we say "I understand"? I bet your MSc thesis was ahead of its time.

1.31.2021 11:38am - Thanks for the reminder. Looking forward to seeing you all Tuesday at 3:30pm (PT). As usual, lux and bagels will be served in the (virtual) lobby.

1.31.2021 9:45am - Protests against Covid-19 vaccination remind me of those organized against smallpox vaccine #Bookofwhy p. 44 They may raise awareness though of the counterfactual concept of "personal risk", see, a poorly understood Rung-3 concept.

1.31.2021 9:18am - (Replying to @KordingLab @andpru and 5 others) It was a blind spot for Reichenbach too, who posited (1956) that "there is no correlation w/o causation". See #Bookofwhy p. 199.

1.31.2021 9:08am - (Replying to @andpru @KordingLab and 5 others) Who could imagine that the Bristol group, so fearful of the "tyranny of DAGs", would be the one to explain "collider bias" to Nature readers. Next we should expect them to explain backdoor, external validity, mediation, etc. etc. DAGs do Wag their Tails.

1.30.2021 9:59pm - We decided indeed to file for a review petition against the unjust Supreme Court decision, to ensure that those responsible for our son's murder will remain behind bars. "And the Lord said [to Cain]: Your brother's blood is crying out to me from the ground" (Genesis: 4,10).

1.30.2021 7:55pm - (Replying to @humlu) Machine Learning. Sorry for presuming it has become a universally recognized acronym.

1.30.2021 4:04pm - (Replying to @TheSyst00873084 @JimBlevins0 and @bariweiss) Beg to differ. BDS does aim to eliminate Israel, see what its leaders say:

1.30.2021 3:57pm - (1/2) Readers noted that to anticipate/discuss clashes with theory may be too tall an order for ML-trained authors to fill, because the very idea that there exists a theory of what can and cannot be accomplished using model-free ML methods is outside the vocabulary of ML-education. 1/2
1.30.2021 3:57pm - (2/2) (Replying to @yudapearl) I would like to believe my ML-friends, who assure me that ML-education is changing, and for the better. Speaking as an editor, the JCI would welcome ML-based papers that are guided by a theory of what's doable with ML methods 2/2

1.30.2021 3:11pm - Fight for Free Speech and Academic Freedom
1.30.2021 3:11pm - (Replying to @yudapearl) Please read carefully, you and I might be next. And I know first hand how those "diversity" commissars operate.

1.30.2021 2:44pm - If industrial research came to this conclusion I am tempted to become an "industrialist", because some of my "methodologist" colleagues have reached an opposite conclusion. See and

1.30.2021 2:27pm - (Replying to @TheSyst00873084 @JimBlevins0 and @bariweiss) Noam Chomsky described BDS as "hypocrisy rising to heaven" and you are sending readers to BDS website to learn "what-is-bds" ? To discover what BDS aims are, read what their leaders tell their supporters. Try this:

1.30.2021 11:17am - (Replying to @chardalarna) Jewish Americans solidarity with African Americans hinges not on shared victimhood but on the secret of resilience and auto-emancipation, of which we are icons of inspiration.

1.30.2021 10:14am - (Replying to @chardalarna) Excuse me, I was there. Once they arrived to Israel Jews faced a genocidal attack ("rivers of blood"), not "some discrimination", then 73 years of eliminational hostilities. American Jews remains allies of African Americans in the pursuit of civil rights and auto-emancipation.

1.30.2021 7:28am - The melody sounded familiar. Lo and Behold: "Yigdal Elohim Chai..."!!! The same prayer we kids used to chant every morning at school, about the same time, 1943-1945, but many miles away, in Israel-to-be, knowing and not really knowing what's going on in Europe.

1.30.2021 7:09am - (Replying to @phoenix1189) For the life of me, I don't know what your are talking about. What you may perceive as "straw man" for me is vivid reality that only the blind or wishfully blind can miss. To reiterate, Israel's creation was a the most successful struggle against oppression and colonialism.

1.30.2021 5:53am - (Replying to @phoenix1189) Truth and astonishments are part of Chutzpah. Israel is the only place where Jews battled successfully against oppression -- through "autoemancipation", ie., Zionism -- and against colonialism -- the British mandate.

1.30.2021 5:38am - (Replying to @CMastication) Agree! DAGs reside in your head and need to be drawn only if you wish to conclude something that cannot be computed in your head. Otherwise, another beer would suffice.

1.29.2021 4:15pm - This clips gives a fairly accurate account of the events and hidden ropes that led to the recent court decision. February 1st is when we mark the anniversary of Danny's death, and say Kaddish: Yitgadal Ve'Yitkadash Smei Rabba.

1.29.2021 9:32am - (Replying to @doinkboy @rlmcelreath and @analisereal) Of course. But one is rarely able to focus on path and test it. All we can do is test that testable implications of a model, which often entail several "zero paths" and several "zero confounders".

1.29.2021 7:13am - The Journal of Causal Inference just rejected another paper claiming to achieve "transfer learning" & "robustness" in cases where theory says robustness depends on the data-generating model. We urge ML authors to anticipate/discuss such clashes, and to cast claims accordingly.

1.29.2021 6:30am - (Replying to @doinkboy @rlmcelreath and @analisereal) Surely, one ought not to control for anything that analysis says is bad for you. The purpose of the analysis is to imagine nature choosing a structure which you do not know, you controlling for Z regardless. Would you be better off or worse off given that structure.

1.29.2021 5:05am - I am still convinced, but would change "beneath the criticism" to "deep beneath the criticism"

1.28.2021 1:35pm - (Replying to @UweSiebert9 @kareem_carr and @socmd) Is epidemiology "younger" then economics? James Lind was there before Adam Smith !!!

1.28.2021 12:21pm - The most cheerful image of the week! Reminds me how we sang in kindergarten: Tu Bishvat Higiaa," and how we lined up, tiny shovels on our shoulders, to plant trees in the neighborhood. Friends tell me the ficus tree I plated 80 years ago is still standing, next to the barber shop

1.28.2021 12:00pm - (1/3) When a killer is behind bars, responsibility is absorbed by one deranged individual. When a killer is freed, society as a whole assumes responsibility for the crime. Today, the Supreme Court of Pakistan has handed an indictment to an entire nation, institutionally, culturally
1.28.2021 12:00pm - (2/3) (Replying to @yudapearl) and morally, for one of the most horrific crime of the century, which forever will stain the moral standing of that nation. We urge the US Department of Justice to vigorously pursue a request for extraditing Omar Schiek to stand a trial in the US, and we hope Pakistan
1.28.2021 12:00pm - (3/3) (Replying to @yudapearl) responds positively to such request, to rectify the injustice brought about today by two of the three judges.

1.27.2021 11:11pm - (Replying to @zacharylipton) But is ad-hoc-ness a curable weakness of ML? I have a theory that it can be cured only by having a model of right and wrong, namely, a model of the world outside the data themselves. Such a model, unfortunately, is not in the vocabulary of applied ML.

1.27.2021 9:14pm - (Replying to @grisaitis) You got it right! no 23 k488 in A major. Divine!

1.27.2021 8:21pm - CA Progressive Zionists Decry “Anti-Israel” Test in Questionnaire via @jewishjournal
1.27.2021 8:21pm - (Replying to @yudapearl) I have just joined a new sub-party "Progressive Zionists of California" whose existence I just discovered. Glad there are people who devote their time and energy to protect the Democratic Party from deceit, distortion, Zionophobia and other BDS practices.

1.27.2021 7:36pm - (Replying to @databoydg) I wish I understood what you want me to concede. DAGs represent our state of understanding. They do not claim anything that deserves concession.

1.27.2021 7:25pm - (1/ ) Had a wonderful vaccination day!! The streets are were I left them, 9 months ago. The GPS remembers who is the boss. Mozart piano concerto sings from the car radio. Any readers from Salzburg? Thank you for giving us a Mozart, 265 years ago, Jan 27. Oh, vaccine? It took an hour 1/
1.27.2021 7:25pm - (Replying to @yudapearl) waiting in a well heated car. But who cares about waiting when Mozart plays the piano and lets you joke with the masked nurses at the same time. Wonderful vaccination day!

1.27.2021 11:04am - Same suggestion to the book-club that @careem_carr is navigating, with the obvious modulation vis a vis statisticians' pre-conceptions (if any?). Happy sailing.

1.27.2021 9:41am - Great honor for me (and Dana). Suggestion: let the book speak to you as if you knew nothing about Epi. and nothing about fancy named concepts (exchangeability, ignorability etc.), in short, imagine that you are a robot, trying to become an epidemiologist. Happy sailing.

1.27.2021 1:40am - (Replying to @smueller @ElonMuk and 2 others) Thanks for catching, no mockery intended.

1.26.2021 8:16pm - This just in: As you can see, I haven't given up on reclaiming the soul of data science. Don't ask me "Why Toyota?". Ask if @elonmuk was also approached. Not yet, but he is welcome to help with this tectonic shift. Just not another "data science" building.

1.26.2021 5:34pm - (Replying to @UweSiebert9 and @kareem_carr) It was an important moment for epidemiology, agree. The paper has more citations than the one on do-calculus (1985), all as a result of open-minded leadership by Sanders and Jamie. Economics was not that lucky, as I lament here:

1.26.2021 5:26pm - (Replying to @VC31415) " is vice versa true as well? " No! My debt to econometric thought shines through each one of my books, and each of my survey papers. More on that later. Must prepare for tomorrow's vaccination. Wish me painless needles.

1.26.2021 7:37am - (1/3) Editorial advice to authors submitting papers on "External Validity" (EV) to the Journal of Causal Inference (JCI). JCI has received a number of papers on EV, mostly from economists, seemingly unaware of decades of progress in EV using graphical models. We realize that the 1/3
1.26.2021 7:37am - (2/3) (Replying to @yudapearl) econometric literature has been less than reflective of progress in neighboring fields, and we wish we could accommodate for students and scholars educated in this tradition. However, we must adhere to our criteria of publishing original research, and we request therefore 2/3
1.26.2021 7:37am - (3/3) (Replying to @yudapearl) that reported results be related to those available in the current literature, and that the technical contributions be made explicit. The same applies to the topic of "generalizability" and "selection bias" which have been gaining popularity in data science circles.

1.26.2021 4:53am - If I speak at the UN again, I would urge them to balance the Holocaust Remembrance Program with the story of Jewish revival—Israel—and how it came into being in the years 1917 to 1948. Here I'm calling Holocaust Museums to highlight this untold story:

1.26.2021 3:19am - Tomorrow 1/27 marks the International Day of Holocaust Remembrance. Exactly 15 years ago I spoke at the UN Headquarters, as part of this Remembrance program, and I am here sharing my remarks: Evidently, this was the 1st time I used the term Zionophobia.

1.26.2021 2:27am - (Replying to @gottfriedmath) Causal discovery can be based on observational data and causal assumptions about the shape of the data generating functions. In applications where those assumptions hold, discovery can be effective. See Tool 7 - Causal Discovery, in

1.26.2021 1:13am - (Replying to @blakeflayton) There is a simple litmus test to gauge the centrality of Israel in Jewish education: How Independence Day (May 15) is celebrated. E.g., Does the Rabbi chant "Barech et Medinat Israel..." Does it feel (almost) like a "high holiday"? How many kids know about November 29 1947. etc..

1.25.2021 8:46pm - It seems that DC pundits have vested interest in prolonging the conflict. How else can we explain their fascination with both UNRWA and the "2-state solution" when it is so clear that the former negates the latter?

1.25.2021 8:22pm - @kareem_carr , would you say a paradox is OVER-HYPED if it provides evidence that human intuition is governed by the logic of causation, as opposed to some other logics?? This is what we learn from "paradoxes" such as MH, Simpson & Lord -- an UNDER-APPRECIATED conclusion!

1.25.2021 7:24pm - Harvard Stat is where I first (1993) presented a seminar on backdoor, invited by Don Rubin -- spies tell me they are still using crystal balls for identification. Jokes aside, I know things have changed in adjacent departments, and I'll be happy to visit and remove more barriers.

1.25.2021 11:36am - (Replying to @HenningStrandin and @kareem_carr) I had in mind @Bookofwhy, which has a chapter on paradoxes and what they are revealing. But any other book would have done as well. Primer starts with Simpson's, and Causality gave it a new slant.

1.25.2021 10:26am - (Replying to @kareem_carr) You haven't read my book? I can't believe it! And I thought the reason your Tweets made so much more sense (contrasting other statisticians) was that you did. I am devastated. But, as they say in the Mishna (Avot 4:3): "There is no man w/o a life-changing moment"

1.25.2021 8:22am - @kareem_carr , would you say a paradox is OVER-HYPED if it provides evidence that human intuition is governed by the logic of causation, as opposed to some other logics?? This is what we learn from "paradoxes" such as MH, Simpson & Lord -- an UNDER-APPRECIATED conclusion!

1.24.2021 12:44pm - The challenge in explaining "paradoxes" is not in showing that "there is no paradox" (surely life is consistent) but to explain why people perceive some phenomena as "paradoxical", and what assumptions create it. That's what #Bookofwhy does to Monty Hall, Simpson, Lord, and more.

1.24.2021 12:31pm - (Replying to @socioburak) Who says my contributions to Soc. Meth. are underrated? I've found Soc. methodologists much more open minded than say economists, who can't stomach results other than home-grown. And the paper you cite are really good, I enjoy reading them even knowing the final results. Enjoy!

1.24.2021 12:41am - @AJCGlobal , This shows how important it is to fight BDS for its Zionphobic, rather than antisemitic roots. One pathology they would never volunteer to teach is Zionophobia.

1.23.2021 5:50pm - (Replying to @yudapearl and @VC31415) A helpful refs. would be and As you can see, instead of keeping X in the graph one can add a node to the mutilated graph, a surrogate of X. ETT is identifiable in some CBN, not all, eg. RCT data on confounded X-->Y, w/ X ternary.

1.23.2021 5:17pm - (Replying to @VC31415) Good question, to which my immediate answer is: It is doable perhaps for small islands in rung 3, but not the entire rung 3. Why? b/c rung 3 contains questions about individual effects e.g., P(Y_x (u)) for a specific U = u. We can answer them from SCM (First Law), not from CBN.

1.23.2021 11:37am - Remembering What Larry King Wrote in “I Am Jewish” via @jewishjournal

1.23.2021 11:37am - (Replying to @yudapearl) On a personal note, our first TV appearance after loosing our son was Larry King show. My wife remembers distinctly how he made it comfortable for us to speak about the unspeakable. Later, his definition of Chutzpa helped me write "A personal journey" YZ"B

1.23.2021 7:19am - ‘Jews Don’t Count:’ Former New York Times Editor Bari Weiss Breaks Down Antisemitism on Left and Right in Megyn Kelly Interview
(1/2) (Replying to @yudapearl) The striking feature of @bariweiss interview is her treating Right and Left racism on equal footing, a stance considered heresy on most campuses. At UCLA, for example, after the Poway shooting, the VC of Equity Diversity and Inclusion (read their lips) set up a kangaroo panel 1/2
(2/2) (Replying to @yudapearl) on the evils of Right-generated antisemitism and, despite students protests, excluded any mention of Left-generated antisemitism. Please check the records of EDI office at your university; you might find it the most anti-diversity office on campus -- I've seen it at UCLA & USC.

1.22.2021 12:26pm - (Replying to @jfvs41 and @EinatWilf) I don't see where you read "perpetual conflict" when we beg for "end of conflict"? Are you reading us correctly?

1.22.2021 9:43am - For readers who wrote to me that DC pundits are aware of @EinatWilf argument, please watch them regurgitating the slogan "2-states is the only..." as if they think 2-state is possible w/o "end of conflict". Who are they kidding? Why not sing: "end-of-conflict is the only..."?

1.22.2021 7:12am - (1/2) I do worry, @MiriamElman , because we've both seen what happened at USC, following R. Ritch's resignation. When 43 prominent professors signed a letter saying (in essence): "We are Zionists, are we welcome on this campus?" what did USC EID officials do? 1/2
1.22.2021 7:12am - (Replying to @yudapearl) They organized (in collusion with Jewish academics) THREE kangaroo panels, on "hate", Antisemitism, Islamophobia, etc. etc., anything but Zionophobia. None of the 43 professors was invited to participate, and the word "Zionism" barely uttered -- a happy coverup for inaction.

1.22.2021 6:20am - Extremely encouraging to see causal graphs trickling into NLP, enforcing my speculation that " learning from the way causal reasoning was domesticated, would benefit researchers in other area of AI, including vision and NLP." (in

1.22.2021 5:46am - A timely and well-documented article. I differ on one item: "Antisemitism-awareness training" has become a COVERUP FOR INACTION. School administrators (in collusion w/ interested faculty) love the idea of initiating training programs on "Antisemitism-awareness". Why? 1/2
1.22.2021 5:46am - Replying to @yudapearl Because it absolves them of dealing with the real problem –Zionophobia - which they try to avoid at all cost, and which will continue to fuel the harassment and marginalization of Jewish students all over the country. Time to call the virus by its name:

1.21.2021 5:56pm - The hordes of bureaucrats and pundits now seeking jobs with the Biden's administration are refusing to hear what @EinatWilf is telling us. God, can't they see how stale, out-dated and content-less their words are, compared with the freshness and solidity of her ideas?

1.21.2021 3:48am - (Replying to @ajaydiv and @herdiants) I have used "inertia" on occasions but it connotes an innate, unshakable property, something there is no point fighting. The academic inertia we are facing IS shakable through informed leadership and enlightened policies.

1.20.2021 11:30pm - (Replying to @herdiants) Actually, I think the word "orthodoxy" is inappropriate here, for it connotes thoughtful ideological effort to resist change. What we are facing is more like short-sighted laziness, perpetuated echo-chamber reinforcement. "viscosity" would be a more appropriate word.

1.20.2021 11:07pm - (Replying to @soboleffspaces) I hope you start with the First Law.

1.20.2021 2:31pm - Readers ask: "And what about causal inference education in ML?". Sadly, I don't believe it's much better. Which is strange; statistics can decouple itself from causation -- ML can't. I hope Biden's adm. enacts new science policies to help AI lift itself from academic orthodoxy.

1.20.2021 3:26pm - (Replying to @SGruninger) You are right, and I am apologizing for using the words "absolutely absent". I should have used "almost absent" as I did in "Stigler’s [book] (2016) barely makes a passing remark to two (hardly known) publications in causal analysis."

1.20.2021 11:05am - A eulogy I wrote for Ilan Halimi 15 years ago,7340,L-3229504,00.html An edited version appeared in Le Monde, March 30, 2006. "Let there be no silence on your grave, Ilan, no rest, nor learned discussion... until another Zola rises with a lauder "J'accuse". Itgadal V'Yitkadash

1.20.2021 7:33am - (Replying to @MichaelALewis10) I have seen it, and I even wrote an endorsing blurb on the cover, predicting a "shock" among unsuspecting econometricians.

1.20.2021 7:20am - (Replying to @RaulMachadoG) Amazing story! What about Uppsala and its open-minded faculty? I wish someone could undertake the project of globally mapping the universities that offer CI courses. Like Covid-19 maps, it would display the struggle of science to lift itself to new heights. Any volunteers?

1.20.2021 6:59am - Tunisian president blames “theft of Jews” for instability via @tech2s The mentality of "stolen election" is not unique to US. I have heard the "stolen land" fantasy since 1948, all over the middle east. Tunisia, welcome to modernity!

1.20.2021 6:28am - New readers ask what "The First Law" is. Ans. It is discussed here: and debated in earlier tweets, searchable here:

1.20.2021 12:24am - (1/2) "What are the most important statistical ideas of the past 50 years?" A new paper by Gelman and Vehtari lists "counterfactual causal inference" as #1. This is in stark contrast to Stigler (2016) "The seven pillars of
1.20.2021 12:24am - (2/2) (Replying to @yudapearl) Statistical Wisdom" in which causality is totally absent. Can statistics students expect a renaissance? A reform? Unfortunately, the number of statistics departments offering classes in causal inference is miserably small and those teaching The First Law -- infinitesimal.

1.19.2021 11:32pm - Had @J_Insider asked me what to ask Tony Blinken at his confirmation hearing I wouldn't hesitate for a moment: "Have you read the Book "The War of Return" by Schwartz Wilf?" Blinken's answer would tell us immediately where the Middle East is heading.

1.19.2021 4:53pm - Are potential-outcomes "natural primitives" in economics? Countering Imbens, I claimed they are NOT. @Chris_Auld , in a comment:, sided with Imbens. I've just wrote a reply to Chris, explaining why the variables appearing in SEM are NOT potential outcomes.

1.19.2021 4:47pm - Rest in strength Petra. I will miss your logo: "Make no mistake: The Punditocracy that gets Israel wrong also gets a lot of other things wrong", which gave me strength each time I Tweeted you at @WarpedMirrorPMB .

1.19.2021 2:04pm - Sad. I once spoke at this synagogue. I has been a landmark of interfaith dialogue.

1.19.2021 1:26pm - (Replying to @Jackiew80333500 and @mishtal) And I call it "Trumpism," to discharge populist slogans on people who are dumb enough to parrot them.

1.19.2021 6:09am - (Replying to @JohnnyNewman03) (1) Use facts, not slogans, (2) Be constructive (toward co-existence) not destructive.

1.19.2021 5:40am - What impact will CI have on AI? I believe it will takes years before AI researchers realize that the limitations embodied in the Ladder of Causation, like those discovered in formal systems are universal in nature, cutting across tasks and disciplines.

1.18.2021 5:05pm - Perfectly articulated. That's why I don's feel comfortable with the equation: anti-Zionism = anti-Semitism. The latter affirms victimhood, the former negates emancipation; perhaps the only successful emancipation in modern history.

1.18.2021 4:46pm - (Replying to @WaterFront8) One "nice" thing Zionism has achieved: Jews are no longer waiting for approval before choosing life. The lexicon of "nice", "good", "bad" presupposes some "life" in the background; Zionism made this background a prerequisite

1.18.2021 3:29pm - (Replying to @Martin_Kramer) Sorting tirelessly through the unvarnished, documented history and its diverse interpretations, I still dare add my own:

1.18.2021 3:01pm - MLK Day encourages many to quote and interpret MLK speeches. I chose his famous: “When people criticize Zionists, they mean Jews. You’re talking antisemitism.” Today he would have said: “When people criticize Zionism, they mean Israel's existence. You’re talking Zionophobia.”

1.18.2021 2:29pm - (Replying to @aTomBeer @VC31415 and @PHuenermund) Thanks for posting!.

1.18.2021 10:08am - Informative thread for victims of missing data, as well as traditional analysts of missing-data. Nice example how Rubin's MAR differs from graph-based MAR, why the latter is more transparent, and how to recover from missingness when MAR is not satisfied and enjoy it.

1.18.2021 6:22am - An badly under-appreciate miracle of DAGs is their ability to quantify, not only matching subpopulations, but also individual-level effects, e.g., the extent to which your own child should worry when taking a vaccine. See Causality, Section 9.3.3, and

1.18.2021 1:52am - (Replying to @mishtal) David. Thanks for undertaking this important project. I have more material for you than you can handle. I call it "The Anatomy of Campus Zionophobia". Please contact me at

1.17.2021 7:44pm - Who could imagine, in 2018 (when we wrote #Bookofwhy) that vaccination would become a "hot item"? We thought Causality was, or should be, the hottest thing imaginable.

1.17.2021 2:10pm - (Replying to @geomblog and @2plus2make5) I get new ideas when my old ones are attacked. That's why I am on Twitter.

1.17.2021 7:58am - (Replying to @artistexyz and @causalinf) I like your mild statement: "DAGs make PO immensely clearer and more powerful." May it rings in all PO corners of the world. (Are they still any?).

1.17.2021 3:29am - Imben's paper should be read together with my response: I've described it as a desperate attempt by ideological orthodoxy to hold on to a stagnated & insular culture. It will result in a temporary slow down of progress among traditional economists, but
1.17.2021 3:29am - (Replying to @yudapearl) will stimulate curious researchers to actually compare the two frameworks, side by side, and move the field forward. See also ( especially Comment 17) showing how the choice of IV requires graphical analysis. T. Kuhn had nicer words to say about orthodoxy.

1.16.2021 6:46am - (Replying to @jgrey_nplus1) If one aims at co-existence, and if the Irish were to claim England as a stolen land and act accordingly, one should mention the latter when discussing the former.

1.16.2021 5:51am - Sorry, that link was wrong. The correct link to transcript is:

1.16.2021 5:33am - (Replying to @jgrey_nplus1) Valid question. A two-step answer: (1) Base your critics on independently verified facts, rather than BDS propaganda. (2) Add a word or two about the "treatment of Israelis by Palestinians" and how they can contribute toward co-existence. A Zionophobe will never pass step (2).

1.16.2021 3:53am - It seems that I am the only student of color adaptation by chameleons. Sharper microscopes would be helpful.

1.16.2021 3:22am - (Replying to @ShMMor and @halbfinger) Amazing! Even the NYT, as hard as it tries, can't convince Israelis that a band of frustrated, attention-seeking activists can turn overnight into "analysts" and, suddenly, see the light.

1.16.2021 2:42am - (Replying to @yudapearl @questionsin2014 and 7 others) Zionophobes, like chameleons, are masters of color adaptation. When seeking victimhood they are merely "critiquing the occupation", when seeking Israel's destruction they are merely studying the Western notion of a "State". Colorful creatures, Zionophobes.

1.15.2021 4:04pm - I am informed that Moshe Vardi's talk on Logic and Computation (part of World's Logic Day) is now available on Youtube: Please do draw parallels to the current struggle within causal inference to establish firm formal foundations.

1.15.2021 3:52pm - (Replying to @questionsin2014 @ShaiDeLuca and 6 others) My My!! "Stigmatized any critique of Israeli occupation"!! How inconsiderate!! That is why I am only using Zionophobia, no stigma. No true Zionophobe has ever complained for being "stigmatized" as such. They take it as a badge of honor.

1.15.2021 3:30pm - (Replying to @questionsin2014 @ShaiDeLuca and 6 others) What's the occasion? Has he repented? confessed to? or admitted Zionophobic bigotry?

1.15.2021 3:16pm - (Replying to @ZennaTavares @neuro_data and 4 others) I do not think Mackie is underappreciated (same with Rothman's "sufficient causes") The weaknesses and ambiguities in this approach are noted chapter 10 of Causality, and I do not see modern authors addressing them. See also

1.14.2021 8:05pm - (Replying to @VC31415 @yskout and @PHuenermund) The problem become sticky when we define a counterfactual Y_x with x an event, not an equation. If we have two equations: Y=f(X) and Y=g(X), there is no unique equation to shut-of in order to define Y_x. SCM requires that there will be a unique equation for every variable.

1.14.2021 7:55pm - (1/4) A letter I wrote to the California Board of Education: I strongly oppose the 2021 California Ethnic Studies Model Curriculum. I am particularly alarmed by its attempt to depict inter-ethnic relationships as a irreconcilable struggle between racially-defined “oppressed” 1/4
1.14.2021 7:55pm - (2/4) (Replying to @yudapearl) and "oppressors” and by the way it associates "whiteness" with "oppression" and "colonialism". I am a "white" Jewish American, and I believe that the history of my people is a model of emancipation from oppression and colonialism, culminating in the State of Israel which is 2/4
1.14.2021 7:55pm - (3/4) (Replying to @yudapearl) an inspirational model of an oppressed ethnic minority lifting itself from the margin of history to become a world center of art, science and entrepreneurship -- a multi-colored light-house of free speech and gender equality. I want my grandchildren to take pride in this 3/4
1.14.2021 7:55pm - (4/4) (Replying to @yudapearl) historical transformation and to share our experience with other minorities; yet sharing as equal partners in one colorful mosaic of ethnic diversity, not as guilt-stricken "Whites", burdened with undeserved privileges. 4/4

1.14.2021 6:12pm - (Replying to @VC31415 @yskout and @PHuenermund) So, when applied economists write a paper and use this "construct," they just use it, and do not feel the need to justify it, or reference another economist who justified it, or who first argued for its validity. Am I right?

1.14.2021 5:49pm - (Replying to @VC31415 @yskout and @PHuenermund) So can we call it: "common practice" in econometric, with unknown origin and potentially debatable foundations?

1.14.2021 5:37pm - (Replying to @VC31415 @yskout and @PHuenermund) Good. We are on the same page. But what is meant by "of course"? Is it self evident? Common knowledge? Common practice? A theorem? A definition? I recall Heckman objecting strongly to this equation-replacement operation. See; would he agree with you?

1.14.2021 5:10pm - (Replying to @VC31415 @yskout and @PHuenermund) I am not criticizing the writing, I am merely trying to understand the principle. Is Y(p) = Y_G_p same, or similar to Y_x = Y_(M_x) where M_x is the modified model, in which the eq. for X has been replace by X=x ?

1.14.2021 4:41pm - (Replying to @VC31415 @yskout and @PHuenermund) What I am missing is the principle by which counterfactual predictions are made, once we have a structural model M. Is the principle the same as the First Law (see or does it vary ("custom made") from model to model?

1.14.2021 3:10pm - I will be speaking at this important webinar on Sunday. I have two grand-children in California, and I want to see them grow up in a friendly environment, respectful of differences and free of finger-pointing.

1.14.2021 9:18am - (Replying to @_Srijit and @blakeflayton) As far as I know, Israel does not have a "constitution". Its "Declaration of Independence" is often used in courts as a substitute.

1.13.2021 10:58pm - (Replying to @neuro_data) No causal question is "bad" and, these days, no such question lacks the mathematical machinery to answer it. Specifically, your question calls for assessing the probability PS that each of the proposed events was SUFFICIENT for producing the light. See

1.13.2021 1:12pm - Please consider this a personal invitation to the 2021 Daniel Pearl Memorial Lecture with @jaketapper

1.13.2021 11:52am - HMM... Does it mean Biden will need to start all over from scratch?

1.13.2021 9:15am - I am delighted to see such heightened interest in teaching Causal Inference. The homework problems posted are not trivial, and should prepare students for further explorations. It's a fitted tribute to this educational channel which has swelled to 40K followers today.

1.12.2021 4:53pm - (Replying to @Ganduin) I think it was the midterm, later turned into HW4.

1.12.2021 4:23pm - To all teachers of Causal Inference: Homework & Solutions from the days I taught it:

1.12.2021 1:58pm - (Replying to @RaulMachadoG @PHuenermund and 2 others) Tell us more about that movement. No one has asked me to de-platform yet, so I intend to stay here till the First Law makes it to elementary textbooks.

1.12.2021 5:54am - (Replying to @neilturkewitz and @DorotheaBaur) I need help parsing the connection to counterfactuals and where the outrage is.

1.12.2021 5:45am - (1/2) This quote sounds mighty powerful when it comes on other people's posts. Reassessing its ramifications, I can't understand why economists continue to dismiss the importance of the First Law and insist they can do without it. I can see why potential-outcome folks would be
1.12.2021 5:45am - (2/2) (Replying to @yudapearl) threatened -- it strips potential-outcomes of their primitive status and turns them derivatives of structural models. But why should it threaten economists, also known as "people of the model"? I hope to see it soon in some elementary econ. textbook, way before they talk IV .

1.11.2021 9:37pm -;text=;size=l

1.11.2021 9:18pm - Check out "World-Logic-Day Moshe Vardi Lecture: From Aristotle to the iPhone" @Eventbrite

1.11.2021 3:19pm - (1/2) Just heard Jim Heckman talk here: Lo and behold, he IS using the "First Law of Causal Inference", except he does not view it as "the definition of counterfactuals" and does not give it a symbolic notation in terms of the
1.11.2021 3:19pm - (2/2) (Replying to @yudapearl) original and hypothetical models, M and M_x. Eco. students should rejoice, they now have Heckman's blessing for benefitting from The First Law. Not sure if they also have his blessing for the 3-step method of computing "probabilities of counterfactuals"; it should come soon.

1.11.2021 5:04am - Philosophers of science are taking a serious look at #Bookofwhy, as seen here: and other recent books and articles.

1.11.2021 12:27am - (Replying to @jf_raj and @sigfridlundberg) Such inclusion is essential, but it is not enough. Why? Because, if taught by DL people it would quickly give way to DL thinking to become a patch, rather than a starting point. CI must be taught ground up.

1.11.2021 12:12am - (Replying to @jiafengchen42 @guilhermejd1 and 2 others) The derivation of LATE (assuming monotonicity) is not in Causality text; it is in the homework to Causality. (Will Tweet a link). But given that Z qualifies as IV, the derivation should be the same.

1.10.2021 4:09pm - Not a dumb question at all. The answer is YES. See Fig. 7.8(b) in Causality, together with other variants and a simple criterion. Only @guido_imbens thinks economists can handle IV variants in their heads, see Comment-17 by @eliasbareinboim and Forney.

1.10.2021 3:31pm - (Replying to @DKedmey and @EVKontorovich) We do not expect universities to "thwart" Zionophobia, or to police any speech on campus. We expect them, however, to treat Zionophobia the same way they treat Islamophobia or white supremacy, namely, as morally disgusting, albeit protected, ideologies:

1.10.2021 3:48am - A beautifully summarized nugget of a post which my esteemed ML colleagues are not eager to discuss with students: Here is another such post: I hope Google does not cancel me before ML students cry: "Our Emperor has no cloths."

1.10.2021 2:49am - (Replying to @VC31415 and @alex_peys) The only consolation I can offer, short of another vodka, is an experience-based assurance that, like alcohol, all this excitement gets easily absorbed in one's blood stream to become part of one's thought process (if not personality) - bonusing a safe ticket to modernity.

1.9.2021 5:01pm - I'll be speaking tomorrow 2 pm on this panel and, as readers should expect, I will leave aside the cause-effect relationships between Zionophobia and Jew-hatred, and treat the former as a new moral deformity, demanding its own vaccine, in the spirit of:

1.9.2021 2:17pm - We used to forgive such rhetoric with: "Politicians do not mean what they say, its just sabre rattling". Today we know that "sabre rattling" need to be taken seriously - it kills.

1.9.2021 3:54am - (Replying to @VC31415) Not sure about super-short, but the first proof of the d-separation Thm was cooked up by Thomas Verma here; not only is it self-contained, it doesn't even mention "probabilities". Moreover, as told in #Bookofwhy, Tom thought it was a homework exercise.

1.9.2021 2:26am - (Replying to @itamarcaspi and @mktscompetition) Oneg Shabbat indeed to see the Ladder of Causation climbed in Hebrew.

1.9.2021 2:11am - (Replying to @rorykoehler) We can also be sufficiently sober to recognize how much of that "inhumane treatment" is clutching at sheer propaganda.

1.9.2021 12:59am - We are now told how Israel managed to lure Pfizer into making it the vaccination capital of the world. The inducement was to turn its highly digitized, centralized and socialized healthcare system into the planet's first open-access scientific laboratory for mass vaccination.

1.9.2021 12:06am - This unpublished paper : "Finding a minimal d-separator" contains lots of other goodies worth noting; for example, finding an adjustment set of minimum cardinality, or other measures of measurement costs and estimation precision.

1.8.2021 11:50pm - This teaching assistant at Johns Hopkins should actually be encouraged to continue and expose to the world the extent to which Zionophobic mentality has infected academic life in the US. She does it so willingly and openly -- a treasure.

1.8.2021 1:43am - The original statement by @VC31415 is true, and can be proven through the corollaries of The analysis of deals with choosing a definition for "confounder", as opposed to "deconfounders", whose definition is unambiguous.

1.7.2021 3:42pm - (Replying to @SidikiXavier and @jkriss) It's about the consequences of experimenting, from which we can deduce "how to experiment in order to learn".

1.7.2021 1:15pm - (Replying to @VC31415) Was/is there a link to the haiku?

1.7.2021 1:05pm - The most hearty expression to the way my colleagues in Tel Aviv are feeling.

1.6.2021 1:27pm - (Replying to @VC31415 and @PHuenermund) It comes with a bonus: The Instrumental Inequality

1.6.2021 7:34am - (Replying to @MairavZ) Calling out antisemitism and racism is not so amazing either. His test will be calling out Zionophobia in the left-wing of the Democratic Party.

1.6.2021 2:54am - (Replying to @AntonioCapretti) Confessing: Occasionally, I also peek at this paper; arguments tend to get rusty in time.

1.6.2021 2:21am - So refreshing to read a book that deals directly with the core and essence of the Israeli-Palestinian conflict; all others deal with secondary issues that authors enjoy writing about and audience enjoy listening to -- an indulgence that prolongs the conflict.

1.6.2021 1:48am - (Replying to @yskout @achimdomma and @CambridgeUP) I have a different theory; they simply are not used to deal with books that continue to generate new markets, rather than sell for 3-5 years and die when the market saturates.

1.5.2021 11:53pm - (Replying to @achimdomma and @CambridgeUP) Every tweet that wakes up @CambridgeUP may speed up the process. We are waiting to hear from them.

1.5.2021 11:18pm - (1/2) Another cheerful observation from 2020: the 2nd Ed. of CAUSALITY (2009) has reached an all-time record of Google Scholar citations (see continuing its exponential trend. I hope @cambridgeUP takes note and gets the revised corrected printing out soon.
1.5.2021 11:18pm - (2/2) (Replying to @yudapearl) Until then, please refer to the errata list on And may 2021 see further penetration of causal thinking into the sciences and pseudo-sciences. 2/2

1.5.2021 9:24pm - The British mutation named The Guardian (and the entire Corbyn herd) will never forgive Israel for proving their textbooks wrong; every Israeli success MUST be at the expense of some oppressed Palestinian - the @guardian Law of conservation of suffering. What an ugly herd.

1.5.2021 8:52pm - To all Hebrew speaking readers on this channel - celebrate #HebrewLanguageDay. And who can, on this day, refrain from singing along with Hava Albertstein: Eliezer Ben Yehuda What a saga of one man's vision! Hannibal over the Alps.

1.5.2021 9:15am - (Replying to @PHuenermund) And the AEA program committee did not invite you as a keynote speaker? Incredible! This would not have happened in computer science. Unthinkable!

1.5.2021 5:32am - (Replying to @michaeldickson) Ad Meah V'Esrim !!!

1.5.2021 1:19am - Armed with the domain-independent language of counterfactuals we should be able to articulate the problems of "optimization, equilibrium and learning" as new types of queries, demanding estimation and, perhaps, new types of assumptions. Worth exploring.

1.5.2021 1:04am - (Replying to @VC31415) It would be a thrill to understand the limitations that White and Chalak found in SCM and how they overcome them, something I was not able to do on my own, despite several efforts. We may even find that some of these limitations (differently labeled) have already been overcome.

1.4.2021 7:41am - (Replying to @BraneRunner) What if it's truly "Deep" and when it's far from banality? Would "profound understanding" make it less banal? Open for suggestions, but "Deep Understanding" deserves a name.

1.4.2021 6:28am - (Replying to @novicus) Considering the enormity of the task, a spice of immodesty is absolutely necessary.

1.4.2021 3:59am - (Replying to @VC31415 and @econostat_) Heckman and Pinto should always be read with a reflective companion: Moreover, the dispute should be settled by examining the 18 problems that Pearl presents here: and judge whether it's sensible to leave them out of econometrics)

1.4.2021 3:17am - (Replying to @craigwpickett @analisereal and 3 others) Solution manual for problems in Primer are available to anyone requesting them. (Write to They are not published at the request of instructors who want to use them in class, for grading.

1.3.2021 11:08pm - Having all but given up on economists, I've spent the Christmas break with AI folks, trying to define Deep Understanding. But now that I got your tweet, I am rushing to retweet it to econometrics students: Behold! Redemption is near!

1.3.2021 3:50pm - When AI ideas begin to influence scholars of English Literature you know that AI has become part of our modern culture. This paper in Modern Language Quarterly tells us how:

1.3.2021 7:41am - Gratified to see Deep Understanding recognized among the History of AI 2020 Awards:

1.3.2021 5:29am - Few remember that this culture of socialized healthcare system started in 1911 by Berl Katzenelson (1877 -1944). The words Kupat Cholim & Tipat Challav are still etched in our memory, and lionize the visionaries who understood what infra-structure is needed for a model State.

1.2.2021 8:56pm - Each of these "computational theory of xyz" deals with a chunked aspect of intelligence; it's hard to believe that we can come up today with a computational theory of ALL of human intelligence. I therefore advocate a more modest goal: "one mini-Turing-test at a time".

1.2.2021 12:25pm - (Replying to @HomoModulans and @sigfridlundberg) When we tried to operationalize these notions, as in the days of Bayesian Networks, we ended up with "Most Probable Explanation", which is far from Best Explanation. The latter is a causal concept which cannot be captured by Bayesianism.

1.2.2021 6:37am - I considered myself a rational person, a "thinking organism" to be accurate, but watching these Ethiopian women kiss the ground of their ancient homeland, made me feel swept by an ocean of emotions of unknown origin. Genes? History? Culture? Upbringing? Go figure.

1.2.2021 2:44am - (Replying to @DavidRouquie and @BrianWandell) The neglect was surprising until we proved that Cartwright's dictum "no causes in, no causes out" is actually a theorem, not merely a recurring phenomenon. At that point, the data-centric culture ceased to be an exploration of possibilities & became a confession of impossibility.

1.1.2021 2:07pm - (Replying to @sd_marlow) Good point. We indeed do not have a formal definition of what "reaching the moon" means, but we have a formal definition of an intermediate-level achievement, necessary for reaching the moon: climb the Ladder of Causation.

1.1.2021 12:25pm - If I were a philanthropist, I would naturally donate another building for "Tree-climbing-science". But not when we have a mathematical proof that "trees are too short". Sadly, donors do not read proofs, and tree-climbers enjoy the climb, the low-hanging fruits & the nth building.

1.1.2021 8:21am - An Israeli friend wrote me how he got a phone call that someone did not show up for their vaccination appointment, if he wants to get one. In half an hour he and his wife were vaccinated, no wait, no uncertainty; polite, effective and done. Amazing - over a million people!!!

1.1.2021 6:45am - Countries are still "able" to invest, but try to talk to NSF or DARPA about it, and they will advise you to submit a proposal under some initiative authored and reviewed by same data-centric scientists. Strange, but only eccentrics are liberated from this self-perpetuating circle

1.1.2021 6:30am - Whatever name we give the interpretation process, be it abduction or IBE or Popperian hypothetic-deduction, we must keep in mind the the hypotheses are causal, and cannot therefore be expressed in the language of algebra or statistics.

1.1.2021 6:22am - Sorry if I did not make it clear. We need an enlightened billionaire to fund a Causal-Science Center, in one or several of our esteemed universities, to start the tectonic shift from data-centric to science-centric education.

1.1.2021 6:14am - (Replying to @sigfridlundberg) One tweak: The hypotheses we need to construct cannot be tested by statistical inference alone. Counterfactual and causal logic need be employed to tell us what data would falsify a given theory. Not sure Popper was aware of it.

1.1.2021 5:52am - (1/2) I keep asking myself, over and over again: "How could they allow it to happen?" I do not blame the scientists themselves - they're paid to pursue the paradigm that made them successful. Nor do I blame funding agencies, industry and university -they are run by same scientists 1/2
1.1.2021 5:52am - (2/2) (Replying to @yudapearl) What surprises me is the absence of a visionary leader who dares ask: Who is going to train our PhD's when the pendulum is swung to the "data-interpretation" side? That's why I keep saying that the future is in the hands of one enlightened billionaire who can see it's coming. 2/2

1.1.2021 2:52am - (1/3) Sharing 2020 quotes that will guide me into 2021: "The data-fitting school is driven by the faith that the secret to rational decisions lies in the data itself, if only we are sufficiently clever at data mining. In contrast, the data-interpreting school views data, not as a 1/3
1.1.2021 2:52am - (2/3) (Replying to @yudapearl) sole object of inquiry but as an auxiliary means for interpreting reality, and “reality” stands for the processes that generate the data". "Ten years from now, historians will be asking: How could scientific leaders of the time allow society to invest almost all its 2/3
1.1.2021 2:52am - (3/3) (Replying to @yudapearl) educational and financial resources in data-fitting technologies and so little on data-interpretation science?" (Quoted from:

12.31.2020 1:44pm - (1/2) My practical dilemma for the Next Decade in AI: How do we ensure that this high-level convergence does not result in the erection of a dozen more "Data Science Centers" in which data are modeled void of science. My theoretical dilemma: Can we converge on a definition 1/2
12.31.2020 1:44pm - (2/2) (Replying to @yudapearl) of "Deep Understanding"? I have proposed one here:, and I can see a bright decade ahead if we can agree.

12.31.2020 6:18am - Mark your calendar (2021, not 2020), and join us with @jaketapper at the Daniel Pearl Memorial Lecture, February 2, 2021.

12.31.2020 5:59am - Another thought on an this article. Everyone understands that "forecast" is far from "understanding" (See Toulmin's 1961) and the difference has been a point of fierce contention in AI. Yet I am not aware of a formal criterion for this demarcation - strange for computer science.

12.31.2020 4:02am - (1/ ) Year 2020 in review: The past year has seen an increase from 25K to almost 40K Twitter followers, musing over 3,000 tweets (searchable here, from discussions of misguided theories to the understanding of Deep Understanding. I am grateful to all readers 1/
12.31.2020 4:02am - (2/n) (Replying to @yudapearl) for stimulating these discussions. Our impact on the research community can be gauged by the number of citations to causal-modeling papers. 2020 has seen a record of 8000 citations to my papers and, as this chart reveals, the exponential growth 2/n
12.31.2020 4:02am - (3/3) (Replying to @yudapearl) promises an accelerated penetration into the walls of skepticism and orthodoxy. Thank you all for supporting this incredible progress, and may 2021 see the establishment of a wisely-directed, well-funded Causal-Science Center in one of our esteemed universities. 3/3

12.30.2020 11:02pm - (1/2) You are not the only one. Teachers, researchers, even authors of classical books confess that the theory of missing-data, especially the MAR conditions, are incomprehensible, and that, when no one is watching, they think about it differently. But the capacity of students 1/2
12.30.2020 11:02pm - (Replying to @yudapearl) to endure torture under tyranny of orthodoxy exceeds all bounds, second only to the capacity of orthodoxy to maintain its grip on education and mainstream literature. I hope 2021 will see more cracks in that grip, including the grips of PO and model-blind machine learning.

12.30.2020 2:18pm - That's a great New Year gift, thanks. I hope it brings new readers into the Deep Understanding fold. I am especially intrigued to see how you introduce do-calculus before d-separation.

12.30.2020 3:16am - (1/n) Good news for all victims of missing-data. We are informed that our paper "Graphical Models for Processing Missing Data" (with @Carthica ) has pacified all reviewers' objections and will be published in JASA, see So, if you are confused, as most people 1/n
12.30.2020 3:16am - (2/n) (Replying to @yudapearl) are, about what MCAR and MAR mean, or whether your imputation algorithm would ever converge, or whether your assumptions can be tested etc. --- those days are over. Expressing "the reasons for missingness" in a causal graph turns the "missing-data problem" into a science, 2/n
12.30.2020 3:16am - (3/3) (Replying to @yudapearl) enjoying transparent assumptions, methods of recovering causal and statistical relationships, and methods of testing one's assumptions. Recall: "behind every statistical model there stands a causal model begging to be expressed", speak to it, and there will be light! 3/3

12.29.2020 5:30pm - A new statement by the US Justice Department, in support of our appeal and introducing the idea that the US stands ready to request the extradition of our son's murderer.

12.29.2020 5:15pm - (Replying to @lihua_lei_stat @BetsyOgburn and 5 others) I would advise against the "super-population" metaphor, because you lose sight of the fact that all results are still sensitive to the structure of the causal model, and the query is still a causal query, not just statistical functional over the observables.

12.29.2020 1:01pm - (Replying to @BetsyOgburn @jiafengchen42 and 5 others) We can't replace "estimand" witn "estimator" for the same reason that you can't replace the "mean" with the (possibly many) "estimator of the mean". "Causal estimand" is ok for "query", but the latter is so much shorter and refreshing, even a nonstatistician understands it.

12.29.2020 7:13am - (Replying to @NeatWitTweet) In retrospect, every success (even the miracle of Chanukah) can be "simply attributed" to a combination factors A, B, C..., none very surprising in itself. Yet patterns of success do form patterns and can stomach some sugarcoating, to help confrontation of the next danger.

12.29.2020 6:29am - (Replying to @eccastro) To "enforce" a logic means to educate people, including those in power, that the relationships between data and the questions we ask are not whimsical, but obey certain principles. Those principles need to be taught in Data Science curricula, now attracting millions of students.

12.29.2020 5:41am - I see a familiar pattern in the way Israeli society recognizes the seriousness of a common danger and the determination with which it mobilizes itself to confront it.

12.29.2020 4:48am - I was one of those fortunate high school kids who studied geography from the 1st edition of Brower's Atlas (1950). Until then, the only Hebrew atlas we had was Jabotinsky & Perlman (1923) where the world was 1/2 British and 1/2 French. Great to meet the man behind my geography.

12.29.2020 3:59am - An edited transcript of my talk at Montreal AI Debate 2 is now posted here

12.29.2020 2:31am - (Replying to @nenetto) Try this one it worked for me. And thanks for luring me into reading this chapter again. I now appreciate how important it is to start with the logic of modeling, something economists are so reluctant to do.

12.28.2020 4:40pm - (Replying to @lihua_lei_stat @artistexyz and 3 others) There is another twist here. Suppose there are two distinct estimands for the some query (say two distinct sufficient sets) one contains zero-probability conditioning sets, the other does not. Is the query identifiable? If so, do we need to examine all estimands before deciding?

12.28.2020 4:18pm - (1/ ) Many readers are asking this question because, sadly, most teachers of CI are PO-limited. Luckily, there is very little that you need for joining the CI generation: (1) The First Law, (2) The second Law. Both are explained nicely in Primer, so I would 1/
12.28.2020 4:18pm - (2/ ) (Replying to @yudapearl) recommend to start with Primer. The First Law tells us what PO expressions mean, the Second tells us HOW to accomplish things w/o sweat and tears. As an economist, you might wish to see them formulated and used here: This post 2/
12.28.2020 4:18pm - (3/3) (Replying to @yudapearl) further elaborates on why the First Law is so important, and makes you think why the mainstream econometric leadership is reluctant to use it as a bridge between structural and experimental economics. 3/3

12.28.2020 9:00am - (Replying to @DavidDeutschOxf) Maajid Nawaz is the product of these institutions and, since he never heard of counterfactual logic or causal-inference education, he thinks it's all in the data.

12.28.2020 7:19am - Correct! The answer is NOT in the data, nor do we have the theory to answer it. But we have a logic of counterfactuals that tells us what kind of theory is needed to answer our questions, and what kind of data is needed to support that theory. My question: Who enforces the logic?

12.28.2020 4:42am - I welcome @samplingdude talk: for it clearly distinguishes individualized from population level effects. However, the focus on principal stratification is stifling, for it misses the general analysis of probabilities of causation.

12.28.2020 4:27am - (Replying to @lihua_lei_stat) Vaccine efficacy is usually measured by ATE, which is a population quantity. The quantity P(Y(1)=0|Y(0)=1), like probabilities of causation, PNS, PN, PS, measures individual level effects, tightly bounded here: and visualized here:

12.27.2020 11:30pm - (Replying to @lihua_lei_stat) That's great to know. Add to it the "First Law" and Stanford is on its way to modernity.

12.27.2020 3:01pm - (Replying to @lihua_lei_stat) I must have been a victim of selection bias -- the CG knowledgeables have either avoided me, or pretended to be talking Swahili. Is there any course at Stanford in Causal Graphical models, or equivalent?

12.27.2020 1:09pm - (Replying to @lihua_lei_stat @ShalitUri and 2 others) Apologize for asking, but I could not contain my surprise. Usually, when I speak to Stanford folks, I have to assume zero common understanding, and a curtain over the communication. Why don't you offer @drfeifei to teach a course in CI? Or renovate Koller's class in Graphical M?

12.27.2020 12:23pm - Thanks, the definition of Zionophobia has not changed. Some of their tactics did.

12.27.2020 12:19pm - (Replying to @lihua_lei_stat @ShalitUri and 2 others) Interesting question to explore, there may be some neat patterns of identifiability. Wait a minute! You are at Stanford! So, where did you get your causal inference understanding?

12.27.2020 6:48am - (Replying to @djinnome) Of course, of course, who could deny the axiomatic theory of criticism. But have you noticed what the NYT has been attacking and negating in the past year or two? Why @Bariwiesse had to resign, and why @PeterBeinart (among others) was invited to preach the end of Israel?

12.27.2020 5:16am - (Replying to @WaterFront8) Is this a confession of guilt, an application for a medal, or a plea for an exemption?

12.27.2020 4:04am - Hilariously retweeted by a shame-ridden white Jew.

12.27.2020 3:38am - Jokes aside, has someone conducted a thorough scientific evaluation into the psychosis the NYT editors have with Israel? It comes with another puzzle: the enduring ability of NYT readers to smile at what/how NYT writes about Israel and skip, forgivingly, to the next story.

12.27.2020 3:04am - (Replying to @lihua_lei_stat @ShalitUri and 2 others) I am not familiar with a systematic analysis of identification under non-positive densities. I suspect it should be possible to prove non-identification in most such cases by showing P(M1)=P(M2) and Q(M1)=/=Q(M2) for some contrived M1 and M2.

12.26.2020 9:37pm - (Replying to @lihua_lei_stat @ShalitUri and 2 others) I would put it a bit differently. Positivity allows us to estimate the effect of non-manipulable variables, because Nature has been doing the interventions for us.

12.26.2020 2:15pm - (Replying to @19Naranjito82 and @VC31415) Same term in SCM: "positivity"

12.26.2020 9:29am - These wave-breakers are man-made, and the man who made them was my mom's cousin, Joe, sent to Rotterdam to study "port-engineering". From our conversations, I learned there're fancy tricks to the trade, all of which I forgot. What remained are these infinite waves. Tribute to Joe

12.26.2020 8:17am - Keeping our hopes high, as high as facts on the ground permit.

12.26.2020 1:50am - (Replying to @oacarah) The future belongs to the brave (and to the vaccinated)

12.24.2020 9:16pm - Sending warm wishes to all those beginning their Christmas celebrations around the world. I lived in Haifa during my 4 years of college, and I remember well how ecumenical and joyful the city becomes at Christmas time. Merry Christmas to all.

12.24.2020 7:57pm - (1/2) We refuse to believe that Pakistan government and Pakistani people will let such a travesty of justice tarnish its name. We are heartened to hear that the Government of Pakistan is filing an appeal against the release order so that our son's murderers will remain in jail 1/2
12.24.2020 7:57pm - (Replying to @yudapearl) and justice will prevail. We also have full confidence in the Supreme Court of Pakistan to provide justice for our beloved son and reinforce the paramount importance of the freedom of the press.

12.24.2020 1:53pm - (Replying to @luislamb @GaryMarcus and 13 others) Thank you @liuslanb and Kuddos to @GaryMarcus and the Montreal Team for putting together this AI Debate #2. Here is a tweet I just posted from my corner of the wood. Happy holidays.

12.24.2020 12:23pm - (1/ ) There were abstract points of agreement in the panel but, as a foot soldier, I would have liked to see agreements on concrete bridge-building steps as well. For example, if Stanford Institute for Human AI would see to it that Stanford offers at least one class in Causal AI 1/
12.24.2020 12:23pm - (Replying to @yudapearl) (Say, along Primer or other models of Deep Understanding, as defined by the Ladder). How about it @drfeifei ? For my side, I would have liked to see class material on the logic of ethics and values. Any pointers? @GaryMarcus

12.24.2020 12:17am - (Replying to @WhalingPhoning) Nothing written yet. But the structure of the template is the same as that of the Inference Engine in

12.24.2020 12:12am - (Replying to @zeemo_n @KevinZollman and 2 others) Not sure I serve the title, but Descartes' new dictum: "I emulate therefore I understand" implies that most philosophers will soon become computer scientists.

12.23.2020 12:19pm - (Replying to @DeMelkbroer) You won't miss much, just a few erratta. Merry Christmass !!!!!!!!!

12.22.2020 3:11pm - (Replying to @JaapAbbring @Chris_Auld and 5 others) Thanks for the introduction; I've learned MDP from Ron Howard's book (1960's). What would you like to see about generic SCM methods that you don't see?

12.22.2020 1:31pm - (Replying to @JaapAbbring @Chris_Auld and 5 others) I am not familiar with John Rust's work. How is it related to SCM or to structural econonomics?

12.22.2020 1:15pm - (Replying to @Chris_Auld @VC31415 and 6 others) LATE, as defined by "the causal effect of treatment on outcome, for the subset of individuals who are responsive to the instrument" is structural, because it is a unique property of the SCM, once you choose treatment, outcome and instrument, regardless of identifiability cond.

12.22.2020 10:59am - (Replying to @VC31415 @PhilHaile and 6 others) An estimand itself is just a functional of the joint distribution, hence it is purely statistical. However, what LATE attempts to capture, ie, the effect on the subset of responding individuals is a causal quantity, since it is a property of SCM that may or may not be identified

12.22.2020 10:52am - (Replying to @sbuhai @lewbel and 6 others) Not bad. I would only add that some counterfactual questions cannot be answered by RCT, no matter what. The last sentence may give readers a different impression.

12.22.2020 3:20am - (1/ ) @eliasbareinboim called my attention to an important omission in our discussions over the role of DAGs in estimating counterfactuals from economic models. When skeptics complain: "But we do not have the fully specified structural model" they should be shown how certain classes 1/
12.22.2020 3:20am - (2/ ) (Replying to @yudapearl) of counterfactuals can nevertheless be estimated using the DAG structure to make up for the missing specification. Direct and Indirect effects stand out prominent in this class. Consider the 3 mediation problems, Section 3.2.5 in my Haavelmo's tribute: 2/
12.22.2020 3:20am - (3/ ) (Replying to @yudapearl) These are very hard problems to solve in non-linear models, and I doubt anyone has attempted them prior to yr 2000, even with a fully specified structural model. I don't expect therefore that seasoned economists will be impressed by the ease with which DAGs-students can solve 3/
12.22.2020 3:20am - (4/4) (Replying to @yudapearl) them today (see However, considering that mediation analysis aims at unraveling how the economy works and that it holds the key to questions of discrimination and fairness, modern economists will appreciate what DAGs can do for counterfactuals. 4/4

12.21.2020 8:57pm - (Replying to @data4sci and @matplotlib) I believe we need to emphasize the computational benefit of the method as discussed in the 2nd paragraph of page 75, Primer.

12.21.2020 9:50am - (Replying to @tdietterich @quantadan and 2 others) This is a good compromise. I would even describe it as "carefully engineered highly complex terrain fitting."

12.21.2020 9:40am - (Replying to @PhilHaile @VC31415 and 5 others) The reason I am eager to hear about "something else" is that I would like to know if SCM should be enriched with new capabilities. I would appreciate the idea behind the "something else", namely how the needed assumptions are of totally difference nature, beyond SCM.

12.21.2020 8:49am - (Replying to @PhilHaile @VC31415 and 5 others) I agree too (rare event) but I would add that, since both DAG and PO are derivable from SCM, and "the something else" has still not revealed itself, we might as well agree that SCM is the natural breeding grounds for counterfactual analysis.

12.21.2020 8:03am - (Replying to @VC31415 @JaapAbbring and 5 others) We went through this distinction with Jason long time ago. What you call "RFsts would try to design experiments or look for quasi-experiements " is none other but scanning their mind's DAG, which they prefer to keep implicit for a variety of reasons.

12.21.2020 7:19am - (Replying to @VC31415 @JaapAbbring and 5 others) I am not clear what "Agnostics are skeptical about", when it is such an easy matter to express skepticism (in assumptions) using the economic models themselves. You do not "impose a heavier structure" by hiding the structure needed for your method to work.

12.21.2020 7:10am - (Replying to @JaapAbbring @VC31415 and 5 others) Yes. This is what the model implies.

12.21.2020 6:20am - (Replying to @VC31415 @JaapAbbring and 5 others) It is a combination of rejecting the First Law and agnosticism in structural economic models. The latter begs the question: what alternative to structural models are there, that can yield counterfactual conclusions formally and systematically from meaningful assumptions?

12.21.2020 6:10am - (Replying to @JaapAbbring @VC31415 and 5 others) Why would fixing Q at q_0 be different from fixing P at p_0? The same method should apply for all interventions. Unless there is a catch to Q that I dont see.

12.21.2020 4:25am - (Replying to @VC31415 @JaapAbbring and 5 others) And yet, confused and shaken by this great schism, economists resist the one bridge that can bring the two cultures together into a one symbiotic whole -- The First Law. I have bemoaned this schism in section 4, here:,

12.21.2020 4:15am - (Replying to @VC31415 @JaapAbbring and 5 others) The equilibrium process proposed by economist Arthur Goldberger is analyzed here:, Section 7.2.1. Note the ease with which a counterfactual question is answered in accordance with the "First Law." (One law for all queries, no hand-waving, no hiding)

12.21.2020 3:49am - (Replying to @VC31415 @JaapAbbring and 5 others) You're lucky the editor did not insist on "can you state your assumptions in the form of conditional ignorability statements, to pacify harmless reviewer #3"

12.21.2020 3:41am - (Replying to @VC31415 @JaapAbbring and 5 others) (1) Notation is secondary, substance is what counts, and begs a definition in the form of a procedure applied to one's model that delivers what one later labels as CF. (2) Can you give an example of do(X=X') which is not breakable into elementary do(X=x) interventions?

12.21.2020 3:35am - (Replying to @quantadan @tdietterich and 2 others) My My! They even have three chapters on "curve fitting." It's so reassuring to see honest people take pride of honesty.

12.21.2020 3:24am - (Replying to @JaapAbbring @VC31415 and 5 others) What theoretical framework do they rely on? IOW, do they have a definition of counterfactual Y_x (what Y would be had X been x, contrary to facts) with X endogenous? If they reject the "First Law" as "unnecessary", what do they consider to be "necessary"?

12.21.2020 3:08am - Thank you @RitchieTorres

12.21.2020 1:37am - Yahya Sari is the only one who understands that the Abraham Accord is moving towards an honest peace, not like the cold peace agreements with Egypt and Jordan. His worries are the world's most noble hopes.

12.21.2020 12:40am - (Replying to @quantadan @tdietterich and 2 others) Where is the text from. It seems that someone here agrees with me that "curve-fitting" is more accurate and meaningful description of the aims of DL than "differentiable spongy foam" which describes the end product, not the aim.

12.20.2020 6:23pm - (Replying to @VC31415 @Chris_Auld and 4 others) For the life of me, it's 100 yrs after SEM were introduced to economics and economists are still arguing about their interpretation. Doesn't this mean that instead of asking what another economist wrote or thought we should start asking what he SHOULD have written or thought?

12.20.2020 3:22pm - (Replying to @JaapAbbring @VC31415 and 4 others) Sure. But remembering that a cycle is just a shortcut for a process that brings about equilibrium helps ensures that we have the right semantics for handling cycles.

12.20.2020 7:12am - (Replying to @VC31415 @DonskerClass and 3 others) I meant D and S are actual demand and supply measured (or anticipated) not "curves". The excess S-D leads to inventory and the cost of inventory causes the agent to raise the price P. etc.

12.20.2020 6:05am - (Replying to @VC31415 @DonskerClass and 3 others) Two comments. (1) Having unobserved variables (= functions) in the diagram does not make it non-standard DAGs. (2) It is possible to interpret D and S as actual demand and supply, and the price-determining agent as responding to S & D and minimizing cost of inventory S-D.

12.20.2020 4:09am - (Replying to @jchalupa_) I would say that the physicist is someone with "deep understnanding" of gravitation, yes. Not the equations, because they are written in algebraic form, hence do not allow for answering "undoing" questions. The physicist can make up this deficiency in her mind.

12.20.2020 4:02am - (Replying to @VC31415 @DonskerClass and 3 others) I see only 4 variables in Figure 10. D,S,P,O . Where are the latent intermediate nodes you were pointing to?

12.20.2020 3:39am - (1/2) One more DEEP. We heard about "deep nets","deep learning", "deep mind" and how they struggle to define themselves. Here is my own deep: "DEEP UNDERSTANDING" with very clear definition: A mathematical object that supports reasoning across all 3 levels of the causal hierarchy. 1/2
12.20.2020 3:39am - ( /2) (Replying to @yudapearl) The definition is both clear and unique. I know of only one object offering this capability -- SCM. Moreover, the capacity to reason across all 3 levels is necessary for achieving a "state of understanding." So, enjoy the new "DEEP" &, please, don't waste it on a new start up. /2

12.19.2020 8:50pm - (Replying to @DonskerClass @lewbel and 3 others) I am fascinated by the graph you present. What computation does this "computational graph" perform? Can you compute counterfactuals from it? How about testable implications? What is your research question that this graph helps you answer?

12.19.2020 2:15pm - (Replying to @FelixHill84 and @peabody124) The logic of counterfactuals should be able to tell us what it is about the world that makes "undoing" possible despite lacking a model. One possibility is monotonicity, another is (almost) determinism. Why not run it on this example and find out.

12.19.2020 8:12am - (Replying to @tdietterich @VC31415 and @economeager) Really? Never again! But how do you explain it to people who do not know what "nonlinear regression" is? And, btw, what is it?

12.19.2020 7:56am - (Replying to @vkatikireddi @mendel_random and 2 others) The only way to annoy me is to misrepresent DAGs and SCM. Plowing through your paper, my annoyance level is way below the one described in And I am glad that, with the exception of analogical reasoning, you find DAGs capable of capturing Hill's intuition.

12.19.2020 6:59am - (1/2) I'll be honored to participate in this awesome panel.
My Title: The Domestication of Causal Reasoning: Cultural and Methodological Implications Links:
1. "The Seven Tools of Causal Inference" 2016 Article: Summary:
12.19.2020 6:59am -

12.19.2020 2:46am - CI-minded readers asked what "Target Trial" is. I was wondering myself, and here is what I've learned: Go to Figure 2 of, make Q (Query) the result of some trial you wish to conduct, label all data-driven routines "big data". Now it is a "Target Trial".

12.19.2020 1:08am - I did not think I would live to see "The Tale Wagged by the Dag" again after showing that "what the authors aspire to achieve by abandoning the structural framework has already been accomplished with the help and bliss of that very framework." Take a look

12.18.2020 5:21pm - (Replying to @DKedmey) you are pressing too hard. No Zionophobe can stomach the question you posed w/o betraying his/her bigotry, wait till they finish decorating themselves with the crowns of "universal justice, equity, diversity, inclusion, etc. etc."

12.18.2020 3:54pm - (Replying to @CodecMendoza and @DemMaj4Israel) Nice acrobatics. Now replace Catalan with Palestinian, see how all hell breaks loose.

12.18.2020 7:41am - (Replying to @ChurchillMic) To "solve", yes, but only if it means "do the best we can". The goal of squaring the circle may be noble, but not if a proof exists that it is impossible. Math saves us time and sweats.

12.18.2020 6:34am - (Replying to @neilpearce53x @Yqzhong7 and @PWGTennant) We also managed to build Gothic cathedrals before the science of forces & beams came along. I cannot agree that "drawing a DAG is a formality". When you have several sets of covariates and you are trying to pick the best (in terms of power and cost) DAGs are indispensable.

12.18.2020 2:20am - (Replying to @MarcioMinicz) Agree. That's why it was necessary to augment mathematics with a new connective: <--- or ":=" an assignment, and all the calculus that it entails.

12.18.2020 2:17am - (Replying to @robertwplatt @jon_y_huang and 4 others) The middle ground is "method = best we can do". And this is precisely what is claimed about DAG+SCM. Or, more accurately, "method = best we can do today, counting all available tools, awaiting eagerly for alternatives."

12.18.2020 2:04am - (1/n) Your arguments against "YOUR algorithms of the truth" can be equally choreographed against any use of mathematics in the sciences. If you were to demand the expansion of math tools (say from algebra to calculus to group theory etc.) I would join you with open arms. But you 1/n
12.18.2020 2:04am - (Replying to @yudapearl) argue against ALL algorithms, as if they come at the expense of, rather than to benefit and enrich the "social production of science". To this I must say: Sorry, I am for the mathematization of the sciences, & I believe the history of science attests the benefit of my philosophy.

12.18.2020 1:35am - Hanukkah 2020 at the Israel Philharmonic Menorah with nine branches - “Dreidel Song” via @YouTube
12.18.2020 1:35am - (Replying to @yudapearl) My final tweet for Chanukah of 2020. It has been a pleasure sharing with you the joy and aspirations of this most meaningful of all Jewish holidays.

12.17.2020 11:57pm - (Replying to @robertwplatt @analisereal and 4 others) Good point. Popular culture has it that the appropriate outcome of an epi study is "better treatments". But the logic of CI tells us that an observational study can only aim at "best treatment under the assumptions". Therefore, "scrutinizable assumptions" => "better treatment".

12.17.2020 8:17pm - Just finished lighting the 8th candle of Chanukah with my family and, as we came to the last hymn, we spontaneously burst into the one song that represents the essence of Chanukah: Hatikvah (The Hope 1886): "To be a free nation in our historical homeland"

12.17.2020 5:21pm - I have a perfect picture of our neighborhood in my head. Yet, its awfully helpful to be able to glance at the GPS map once in a while, just in case I skipped the right turn. Not to mention days when they have road constructions, and I need to make a slight detour.

12.17.2020 3:10pm - (Replying to @neilpearce53x) It would be devastating if DAGs would show us things we do not already know. Forget about colliders; if we need to choose a minimal sufficient set for adjustment, I dont know how to ensure minimality and sufficiency w/o a DAG. (And I doubt any of the DAG-less experts know).

12.17.2020 2:24pm - (Replying to @VC31415 @PhilHaile and @steventberry) I wish Twitter would allow me to click the "like" button 3 times. And I know Marschak, Koopman and Hurwitz would have joined me with no 2nd thought -- they wrestled with this question but, lacking graphs, could not make much progress.

12.17.2020 1:53pm - (Replying to @VC31415 @PhilHaile and @steventberry) But the question of whether it is identified IS decidable in polynomial time and, so is the question HOW, if the answer is positive. This is the first two questions a student of economics would want to ask, before addressing more esoteric questions.

12.17.2020 1:41pm - (Replying to @PhilHaile @steventberry and @VC31415) Good. So let's just note that leading econometricians in 2020 still considered the problem of deciding, for an arbitrary structural model, whether an arbitrary policy can be nonparametrically estimated to be just "ONE valuable example of structural modeling", TBD by experiments.

12.17.2020 8:06am - Too bad I did not have this paper when asked (in an interview): "Any real-life problems solved by DAGs?" All I said was: "In my corner of the wood, one generic toy-problem is worth 100 messy "real life" problems". Now I can cite >200 papers to pacify the Gods of messiness.

12.17.2020 7:29am - Theoretical Aspects of Rationality and Knowledge (TARK) is having its next meeting in Beijing: Worth considering. I had my first paper there in 1994: (with Darwiche) and it made quite a splash.,

12.17.2020 7:18am - (Replying to @TimothyLash and @ken_rothman) Wait, Wait, now that I have you on twitter. Why are you depriving your readers of the "First Law of Causal Inference?", and all its flowers ? Has no one complained about this omission?

12.17.2020 6:56am - (Replying to @TimothyLash and @ken_rothman) My faithful Google Scholar whispered in my ears the title of your chapter 3 and provided me a link to it. I hope I did not offend the publisher by clicking on it.

12.17.2020 6:31am - (Replying to @yudapearl @steventberry and 2 others) Likewise, finding some problems that are not easily solved in DAGs does not justify the glaring absence of DAGs from econ. textbooks, education and journals, and the misrepresentation of DAGs capabilities as expressed by top economists such as here:

12.17.2020 6:12am - (Replying to @steventberry @PhilHaile and @VC31415) Taking an arbitrary structural model (triangular) and asking whether the effect of an arbitrary policy can be estimated non-parametrically and, if so, how, is light years away from the caricature you are depicting of cherry picking a toy problem that is "easily solved" with DAGs.

12.17.2020 1:26am - (1/n) I am in receipt of the 4th Edition of the now classic: Modern Epidemiology, co-authored by a partially new team: Kenneth J. Rothman, Tyler J. VanderWeele, Timothy L. Lash, Sebastien Haneuse. The 3rd chapter "Formal Models" is an improvement over previous editions, and should 1/n
12.17.2020 1:26am - (2/n) (Replying to @yudapearl) and should definitely be read by economists and machine learning folks: Some critical comments on first reading: The chapter does not highlights the commonalities/identities among the models discussed. In particular, the logical identity of SCM and PO 2/n
12.17.2020 1:26am - (3/3) (Replying to @yudapearl) is not there, nor the fact that Y_x is derivable from SCM (First Law of CI), or that the Ladder of Causation is made up of 3 rungs, not 2. The usual hang-ups over "manipulable variables" still mar the discussion, eg, but are forgivable, given the rest. 3/3

12.17.2020 12:35am - I was hoping for an inspirational piece to mark the 7th night of Chanukah. The key words here are: "Against All Odds", which characterize the spirit of every scientific endeavor and finds its cultural roots in the story of the Maccabees and their historical heirs.

12.17.2020 12:14am - (Replying to @economeager) I think you should add Simpson's Paradox to your examples because you can get students and gurus to freak out without even using numbers. IOW, it gets deep into the logic that governs out thoughts, and demonstrates the limits of statistics itself. eg.

12.16.2020 1:52pm - (Replying to @lewbel @VC31415 and @PhilHaile) Happy days are here, we are now searching for a problem for which PO and DAG are NOT useful. Haven't we forgotten SCM and what it can do? For example, the bounds defined by the Instrumental Inequality were derived using SCM; see

12.16.2020 1:34pm - (Replying to @steventberry @PhilHaile and @VC31415) My point is that, in order to decide whether a problem is more or less suited to DAGs we need to know what DAGs can do. Moreover, if its less suited to DAGs, we have SCM to work with, and the First Law, and more. My question is whether you know an econ problem that is beyond SCM.

12.16.2020 8:56am - (1/2) I see no substantive change here from the failing paradigm of the past 2 decades [We discussed it in Doha, 2005, remember?], according to which pursuing a 2-state solution will lead to one, while we ignore the giant elephants in the room -- refugees and historical legitimacy. 1/2
12.16.2020 8:56am - (2/2) (Replying to @yudapearl) It does not seem like you had an Israeli on your "task force". And I am talking about an Israeli who is tuned to the overwhelming consensus of Israelis: No 2-state solution without a final end of conflict and end of claims. 2/2

12.16.2020 7:22am - (Replying to @VC31415) Victor, I thought & think so too, but you havn't been reading what the woke-commissars now say about Zionism, and why it is the source of all evil. That's why I am certain they will find a reason to rename Picasso School and Cervantes University, given the slightest excuse.

12.16.2020 5:54am - (Replying to @VC31415) Scientists? God forbid! Fisher and Shockley where Eugenicists and Einstein and Martin Luther King were Zionists, which is much worse. Eagles are safe and owls are even better.

12.16.2020 4:49am - Schools should be named after animals, not people. No one would accuse them of not demonstrating that what matters to us did not matter to them. "San Francisco eagle elementary school" "Berkeley whale High School". Permanent role models.

12.16.2020 4:19am - (Replying to @TaliGoldsheft @HenMazzig and 2 others) The authentic panel running tonight is infinitely more authentic than the @PeterBeinart - @RashidaTlaib charade. But those charades will continue to confuse the uninformed until someone organizes a panel on: "Dismantling Zionophobia" where hawks could not pretend to speak for us.

12.16.2020 3:44am - A pleasure to add international migration to "Simpson's Paradox: The Riddle that would not Die"

12.16.2020 3:26am - (Replying to @BeEngelhardt) These scatter diagrams remind me of those we used in Simpson's paradox, eg., but I did not understand why we would want to find the directions of maximal variation.

12.16.2020 3:18am - (Replying to @blakeflayton) Aesope new fable. The fox the wolf and the hawk ran a panel on how to save doves from extinction. The hawk spoke first: "As a bird, I can speak for my fellow doves. We birds do not really deserve your kindness...The idea that doves need to fend for themselves is absurd..."

12.15.2020 5:53am - My computer will be undergoing a major surgery in the next day or two. I hope to be back to this educational channel with renewed vigor, once the surgery is over.

12.15.2020 3:26am - I'll be sending this #Hanukah video to my family, so that, tomorrow night, when we zoom-light the 5th candle, we will listen again by the message of Chanukah that we sang since kindergarten: A miracle has not happened to us, We carved the rock till we bled And there was light.

12.14.2020 8:58pm - (Replying to @VC31415) Not standard? What could be more standard than identifying causal effects in some structure (not necessarily DAGs)? I do not insist on non-parametric, 14 of the problems are in linear systems. Marschak thought this to be the core of econometrics, w/ all other problems secondary.

12.14.2020 8:30pm - (Replying to @VC31415 I would love to try, but you have to tell me what is special about the problems that the Berry-Levinshon-Pakes (BLP) approach attempts to handle, and whether we can demonstrate a concrete example of those problems in a structure. What's the input and what's the research question

12.14.2020 7:40pm - (Replying to @PhilHaile and @VC31415) I do not doubt its all "standard". But which of the 28 toy problems in, is solvable by the "standard" method and, concretely, how? Again, I am not challenging the many decades, just curious to learn how.

12.14.2020 7:20pm - (Replying to @VC31415) Because whenever I try to convince economists that the methods mentioned are useful, I get the answer: Oh, we can solve it using the "approach" of [name, but no solution], or "it's textbook econometrics" [again, no solution], or "economists are not dumb", & no solution] or "how..
12.14.2020 7:31pm - (Replying to @yudapearl and @VC31415) "how about non-toy problems (Imbens &Co)" & no solution. I am genuinely trying to learn how/if these problems are solved by classical econometric methods, not to challenge economists ability to solve them. I have difficulty learning it in Banach spaces - can we start with toys?

12.14.2020 6:27pm - (Replying to @VC31415) Yes, let's discuss it off twitter, but I fail to understand why the reluctance to discuss concrete problems that everyone understand and go to "approaches," away from the concrete.

12.14.2020 5:40pm - (Replying to @VC31415) I am re-examining the 28 toy problems I posed to economists in, Section 3.2. Can you point to those that can be solved by the method you describe?

12.14.2020 4:10pm - (Replying to @VC31415) I thought "partial identification" (eg Manski type) refers to bounding the estimand when it is not identified. Have I misinterpreted it? If so, where can I find a systematic method of deciding whether g(theta) is identified, when theta isn't. ?? Better yet, what's the priniciple?

12.14.2020 4:00pm - (Replying to @BanTorture and @jeremycorbyn) Thanks for the grammar tip. Glad you feel uncomfortable with the "priest of Zionophobia" title. I am sure @jermycorbyn views it as a badge of honor, so I won't deprive him of this priesthood.

12.14.2020 3:49pm - (Replying to @attilacsordas) I would like to see a specific example of a problematic "pulled together" variables, or other cases requiring conceptual moves.

12.14.2020 3:44pm - (Replying to @VC31415) Yes, you missed Marschak's Maxim. i.e., cases where no theta is identifiable and still the effect of the policy is. See, and many examples in Causality chapter 5.

12.14.2020 7:41am - (Replying to @causalinf) Now you owe me a blurb. But make it mild, don't offend the thin-skinned.

12.14.2020 7:19am - (Replying to @DavidDeutschOxf) I am talking about a robot WITH free will.

12.14.2020 5:08am - A no-nonsense indictment of EU's handling of anti-Semitism.

12.14.2020 1:35am - Jeff Witmer, the current Editor of J. of Stat. Ed. called my attention to this paper which encourages and outlines the teaching of CI in stat classes. Worth following. Jeff has also expressed these ideas here: There's hope for stat
12.14.2020 3:30am - (Replying to @yudapearl) And if there is hope for statistical education, there is also hope for many of its satellite disciplines. Imagine a student who took stat-101 with CI flavor, following it with a standard eco-101 class. Count the conceptual hurdles that that student would be spared.

12.13.2020 10:43pm - (Replying to @BanTorture and @jeremycorbyn) It also appears to be a consensus that, invariably, the priests of Zionophobia perceive themselves as "fact-based people". Therefore, a rational measure of fairness need to be invoked, based on 1.5 million refugees in European DP camps, whom that Palestinians refused to count.

12.13.2020 10:32pm - (Replying to @Stebbing_Heuer and @jeremycorbyn) Palestinian arguments were heard over and over again. However, voting rights are given to those who are willing to accept the final decision, pro or con. Palestinians announced they will not accept, and will respond with "rivers of blood" (Azam Pasha, Cairo, 1947)

12.13.2020 4:01pm - That's marvelous text, thanks for posting. I have not read it past 32 years. My, I was a good writer then, but badly misguided conceptually.

12.13.2020 3:32pm - (Replying to @ThatMarkElliott) Reminds me of the red rag of Vapnik Chervonenko Theorem, which I ventured to call "the Bernoulli Theorem of the hind-sighted scientist"

12.13.2020 11:41am - (Replying to @DanielNevo) [quoted] "I base this observation on three faithful indicators: statistics textbooks, curricula at major statistics departments, and published texts of Presidential Addresses in the past two decades. None of these sources can convince us that causality is central to statistics."

12.13.2020 11:24am - I wouldn't go as far as "CS envy", but I would highly recommend attention to CS's respect for language and symbols. That's why #Bookofwhy starts: "Every science that has thriven has thriven upon its own symbols" [A. de Morgan, 1864]. Corollary: "The First Law of CI".

12.13.2020 11:14am - (1/ ) There is no "lashing" on my side. There was a surprise that a survey of 10-year curriculum of statistical teaching did not include CI. Then a response to class notes that ended with the question: "Is it a good idea to teach what we were not taught". As to the attitude of 1/
12.13.2020 11:14am - (Replying to @yudapearl) "Tell me more" you must be lucky to have such colleagues up there at Tel Aviv University. In my interview with David Hand I list 3 types of evidence for concluding that interest in CI among mainstream statisticians is vanishingly small. No lash, just fact.

12.13.2020 10:58am - (Replying to @yudapearl and @VC31415) You can see this "statistics-envy" attitude in the way Haavelmo's work was received by his contemporaries, in the works of Hendry and his disciples, and in Heckman and Pinto's pride in their "fix" operator being statistically Kosher (it isn't, see

12.13.2020 10:40am - Interesting. But I am tempting to believe that the Rasch's theory did not involve as drastic a paradigm shift as causal inference. The latter demands assumptions that could not be articulated in the language of statistics, eg. treatment does not change sex. Can you write it down?

12.13.2020 10:34am - (Replying to @VC31415) Perhaps it was such rejections that caused economists to compromise their causal intuition, in order to show they are good statisticians. I discuss it here 3.4. What Kept the Cowles Commission at Bay? [Note, I said "teach it systematically and friendlily"]

12.13.2020 10:17am - As a funny anecdote, the editor who initiated the interview, a very famous statistician, refused to publish it. That's how strongly statisticians want to know about causal inference.

12.13.2020 10:05am - "Is it a good idea to teach something (CI) we have not been taught ourselves"??? It's today's key question for statisticians, economists, political scientists, and more. It's the key reason for the reluctance of those fields to teach causal inference systematically and friendlily

12.13.2020 7:36am - (Replying to @BorkRiet) Thanks

12.13.2020 6:37am - Glad to see causal inference appearing among Gelman's "most important ideas". However, he is reluctant to call it a "revolution", perhaps because he hasn't seen the Ladder of Causation or the First Law of Causal Inference. So, I'll link to it:

12.13.2020 5:21am - (Replying to @zeemo_n and @michaeldickson) I am a bit weak on Bhutani history and ideology. What would make a tiny kingdom in the Himalaya take side in David and Goliath?

12.13.2020 4:11am - (Replying to @AimeeSMcCoy) I'll bet the next 10-years will continue along the same well-paved paths, and people will be citing your article to justify their curricula. Survey articles shape standards and standards shape values, this is how scientific cultures slow down science.

12.13.2020 3:45am - (Replying to @AimeeSMcCoy) I read "Challenges for the Next 10 Years" in the title, and assumed the article summarized what statistics educators perceive to be a 10-year agenda. My disbelief targets those educators, not the authors. BTW, weren't you a bit surprised by the findings? This is 2020!

12.13.2020 3:05am - Can you believe it? This is #Dubai! and this is Chanukah! The holiday that even @nytimes couldn't stomach, and for a good reason. Chanukah is the most visible refutation of the picture the @nytimes editors are trying to paint of Israel and of Jewish history. Thank you, Chanukah!

12.13.2020 1:37am - "Not aware" is not what their leaders say. See my Interview with David Hand: On the contrary, "weren't you a bit harsh" they ask "in stating that stat is all about data summarization?" Let's look at this article, sub-titled " a 10-year projection."

12.13.2020 1:10am - Incredible. The Journal of Statistics Education publishes an article on "Data Science in 2020" with no mention of causal inference. Next we will be hearing from stat leaders how important causal inference is for statistical practice, and how they have been doing it all along....

12.12.2020 8:02am - One cannot but wonder what role John Kerry's "No, No, No and No" theory played in this delay. He pronounced it as if no price could prove the "4-No's" theory false.

12.12.2020 7:42am - Wright's failure to convince the Cowles Commission to accept path-diagrams as a new language of causation has resulted in a long econometric winter, from which the fields is still recovering, albeit slowly and painfully, see

12.12.2020 7:07am - (Replying to @RaulMachadoG) Measurement errors are nasty. I did some work on it here:, but much more needs to be done on this frontier. See also

12.12.2020 7:02am - (Replying to @causalinf) You showcased a DAG at NBER? Historic moment! The last guy who tried was Sewall Wright (It was the Cowles Commission, 1947, an even higher authority). I got two un-melted brain cells left -- please share the experience.

12.12.2020 6:33am - (Replying to @michael_at_work) Thanks

12.12.2020 6:21am - This link resists my efforts. See if anyone can unfreeze it: [PDF] A causal theory of error scores. R van Bork, M Rhemtulla, K Sijtsma, D Borsboom - 2020 Abstract In Modern Test Theory, response variables are a function of a common latent variable ...

12.12.2020 6:06am - Breiman's (2002) paper on "Statistical modeling -two cultures" came to serve as an arena for various tensions in the ML vs. Statistical communities. Not sure if this paper reaches same conclusions as mine, but it deserves attention.

12.12.2020 5:43am - I once tweeted: "behind every statistical model there stands a causal model begging to be expressed." Psychometric theories of "error scores" were once thought to be statistical, but this paper thinks differently file:///C:/Users/Judea/Downloads/preprint%20(1).pdf

12.12.2020 5:28am - Correction: "but for" = PN (not PS). It just occurred to me, what does "Thank you" mean? A common word, said hundreds times a day. Conjecture: "Thank you" is ascribed to actions or situations that score high on the PN scale. Eg."If it wasn't for you, I would have been miserable"

12.12.2020 4:34am - On a second reading, I am surprised that the author did not evaluate the role of PN, PS and PNS in the perception of actual causation, as done in "Oxygen, matches and fires" Cheng's "causal power," for example, turned out to be pure PS, and "but for" = PS

12.12.2020 1:35am - To our Spanish & Ladino-speaking readers, here is an unforgettable gift for the 2nd night of Chanukah, Speechless & inspired, this 200+ children choir assures me of the future of humanity; perhaps more than Beethoven's "Ode to Joy" or Handel's Halleluiah.

12.12.2020 1:01am - (Replying to @CaiXueYong) Comment: The estimand needs no data, the estimate does. Comment: Pearson would reject the arrow into "causal model". He thought such models are bad science.

12.12.2020 12:51am - A new mathematical model of causal judgement published in "Cognition": deals with important dilemma: If all information is derivable from structural equations, why bother with "actual causation" as in "the actual cause of death".

12.11.2020 2:46pm - (Replying to @RashidaTlaib) Equally inspired:

12.11.2020 5:53am - This is the Menorah that my grandfather brought with him when he came to Israel in 1924 and around which we, his grandchildren sang our Chanukah songs as far back as I can remember: The original is with my cousin in Iowa, but I have a replica.

12.11.2020 3:31am - What's the difference between Jared Kushner and John Kerry? Kushner read the Bible: "Seek peace and chase after it" (Psalm, 34:14) Kerry thought peace just happens if you launch a shuttle diplomacy to prove your theory right.

12.11.2020 12:42am - Had I known that my blurb would appear below Imben's, I would have added another sentence, perhaps a quote from our friendly discussion:

12.10.2020 5:42pm - Best Chanukah news I've ever received: Peace between Israel and Morocco. As they say in that beautiful Chanukah song: “A miracle did not happen to us; We have not found a vessel of oil; We carved the rock till we bled; And there was light!”

12.10.2020 4:13pm - (Replying to @VC31415 and @economeager) The point of "backdoor" is to take whatever knowledge your have about the word (be it 3, 4 or 100 relevant variables), find out if that knowledge is sufficient to answer your question by controlling for some measured variables and, if yes, who are they, then choose the best set.

12.10.2020 4:03pm - (Replying to @dante_gerardo) I think Scottt @smueller can help you here.

12.10.2020 3:46pm - (Replying to @yudapearl @HannesMalmberg1 and 4 others) How timely to retweet this one.

12.10.2020 7:16am - (Replying to @Dodamalka) Love it! It sounds like the super-intelligent robots we are building using neural networks. Can I used it in my next paper?

12.10.2020 2:31am - In my opinion, Imben's comparison should always be retweeted with my rebuttal: Because readers may not notice the inconsistencies, especially the absence of SCM.

12.10.2020 12:29am - Therefore, all the "advantages" that Imben's wishes to ascribe to to PO (eg. shape restriction) already exist in SCM, leaving PO with two fundamental disadvantages: (1) opaque assumptions and (2) opaque testability. I mentioned this blunder in January

12.10.2020 12:18am - (Replying to @yudapearl @autoregress and 3 others) The main conceptual blunder is the total absence of SCM (Structural Causal Models), of which DAGs are but a graphical abstraction and in terms of which potential outcomes are defined. Thus, whatever capabilities one attributes to PO, those same capabilities are realizable in SCM.

12.9.2020 8:47pm - (Replying to @autoregress @eliasbareinboim and 2 others) I see it as a desperate attempt by ideological orthodoxy to hold on to a stagnated & insular culture. It will result in a temporary slow down of progress, but will also stimulate young researchers to actually compare the two frameworks, side by side, and move the field forward..

12.9.2020 7:29pm - Preparing for tomorrow - the first night of Chanukah. Please rehearse w/ me a few Chanukah songs that you probably recognize: Or, join our choir (about 11 min into this video:) in Sevivon Sov Sov Sov (The Dreidel) Happy Chanukah.

12.9.2020 4:14pm - McGill University Principal and Vice Chancellor Fortier: McGill Must Defend Prof. Salzman and Academic Freedom - Sign the Petition! via @Change

12.9.2020 7:16am - Flying to Jerusalem, through Zoom Something I would not miss, even it meant not speaking on SEM as epistemic state.

12.9.2020 4:22am - While econometrics scholars are laboring to avoid graphical models, their colleagues in microeconomics have found them useful for modeling competing narratives: Not surprised; SEM is the most natural model of an agent's epistemic state.

12.9.2020 3:41am - (Replying to @artistexyz) I wish I would have kept up with my quantum education since I left superconductivity, in 1965,but I havn't. Trying to catch up. Perhaps through d-separation?

12.9.2020 3:26am - The last project Guy describes is precisely what we haven been waiting for when we asked "What oracle judges the correctness of DL programs?". Guy takes P(x) as an oracle to evaluate performance of classifiers. I would suggest SCM to evaluate interventions and counterfactuals.

12.9.2020 3:06am - Another paper worth watching tomorrow at #NeurlPS2020 If it weren't for imitation learning I would have no role models.

12.9.2020 2:51am - Immensely instructive lecture, especially to out discussions of model-blind vs. model-based causal inference. @Bookofwhy

12.9.2020 12:57am - "What Hanukkah means to me?" This year, with a wink to @nytimes , it would be doubly appropriate to re-post this op-ed: "Chanuka - Our Trust Deed to History" Join me in celebrating the most meaningful holiday I've known. @EinatWilf @bariweiss @StandWithUs

12.9.2020 12:11am - (Replying to @Claire_Voltaire) And, don't forget, you have to get on their pre-approved "minority list" before they would squeak "who did you say you are?"

12.9.2020 12:05am - Hail to Britania, God save the Queen These times are good times Else we would not sing..... Singing Oh what a merry land is England.

12.8.2020 11:23pm - Have you heard about the National Academy of Inventors? I was proud to know that, this year, UCLA (where I used to hang out before Covid) topped the nation with 4 of its faculty inducted to NAI: Congratulations go to my CS colleagues J.Cong and G.Varghese.

12.8.2020 7:21am - (Replying to @joshualeifer) It's not scandalous but pitiful to see a person torn away from sweet childhood memories, and an editor of a Jewish journal equating such memories with "rituals and religion" instead of family, peoplehood, freedom, songs, and joy. Happy Chanukah.

12.8.2020 3:51am - (Replying to @PhilHaile) Business is full of successful accountants who can't use algebra. This does not disprove the necessity of algebra in high school education. Evaluating low-hanging counterfactuals does not tell us how to evaluate the high ones, where the future of economics may rest, eg mediation

12.8.2020 12:24am - (Replying to @MarkusSchacher and @CarboJavier) They can ask statistical questions, such as: Will the next object be black or white? but not causal questions, such as: which experiment to perform next or even "what if I act differently.?"

12.8.2020 12:06am - (Replying to @Abel_TorresM) I dont read the mainstream stereotypes, but by AGI I mean all human capabilities including emotions, empathy, free will, consciousness etc.

12.7.2020 11:58pm - (Replying to @RandomlyWalking) This type of inference is called "statistical inference" (ie. going from samples to distribution and back to future samples). Let's honor it with the title "statistical inference" and protect it from the controversies surrounding "causal inference".

12.7.2020 10:04pm - (Replying to @quantadan) All Choirs are based on inductive generalizations. E.g., whatever was'nt done in the past won't be done in the future, or, progress seen in one corner will be matched by others. Agnosticism is inevitable.

12.7.2020 9:51pm - While I agree that deep learning and neural nets lack human-level understanding, I cannot join the anti-AGI choir. Having seen such incredible progress toward human-level understanding in my little corner of the wood, I became a true believer in the eventual realization of AGI.

12.7.2020 4:50pm - AI - Feynman: Understanding or curve-fitting? Just got this article:

12.7.2020 4:17pm - (Replying to @RandomlyWalking) "inductive reasoning" has become too broad a term, can we focus on an example? Do you mean extrapolating from some instances to others?

12.7.2020 9:41am - (Replying to @PhilHaile) So we agree on "helpfulness" and disagree on "necessity". Great progress. Now it is my turn to learn something new: can you describe a question (in the context of SEM) that could not be expressed as a question about some counterfactuals? Genuinely curious.

12.7.2020 9:07am - (Replying to @PhilHaile) I think the disagreement is only on how we describe the disagreement. I am proposing that defining counterfactuals would be helpful in structural economics because it would allow solution to one problem to be transported to another. If you agree than we have no disagreement.

12.7.2020 8:45am - (Replying to @DiogoFerrari) The fact that some statisticians use potential outcome notation to analyze RCT does not make it model-sensitive. Fisher analyzed RCT's w/o potential outcomes. It is therefore desirable to classify RCT as a causal estimation method, but not "causal inference."

12.7.2020 8:29am - (Replying to @PHuenermund and @Andrew___Baker) This is something I also experienced in conversation with economists. It is very hard for them, almost impossible, to grasp the idea that an IV earns its instrumental qualities by virtue of its position in the DAG.

12.7.2020 4:44am - (Replying to @PWGTennant) CI does not insist on the use of DAGs, as long as you start with a model of how the world operates. Potential outcomes could be a contender, if only scientists knew what "conditional ingnorability" is, or how to communicate with it.

12.7.2020 3:14am - The question comes: How can we tell if an algorithm "uses" some causal model of the world? Ans. Run it in a different universe (or domain) and see if it acts differently. So what about data-fitting? Change the universe keeping same sample distribution, will it act differently?

12.7.2020 2:51am - True, but the word "inference" connotes a certain intellectual activity, be it by human or machine, of going from some premises to some conclusions via some logic. Take away the premises and what you have is table lookup, or data-fitting, hardly "inference".

12.7.2020 2:34am - I need to second this tweet, b/c it is becoming increasingly important. Call it "causal inference" (CI) if it uses some causal model of the world. Else, call it by the name of whatever it models. For example, RCT is not CI because it is model-blind, though it estimates effects.

12.6.2020 5:57pm - (Replying to @sbuhai and @PhilHaile) It didn't occur to me that economists would be "insulted" by suggestion of scientifically proven ideas. I happened to befriend Jacob Marschak and Jim Heckman, who were always proud of helping economics overturn the stigma of being a pseudo-science. What happened? Truly sorry.

12.6.2020 4:19pm - (Replying to @ThatMarkElliott @PhilHaile and @sbuhai) Quibble accepted. Replace "science invented algebra" with "scientists (Galileo) recognized the benefits of Algebra"

12.6.2020 4:01pm - The latest round in the discussion of whether economists can benefit from the First Law of Causal Inference -- defining counterfactuals in terms of a structural model.

12.6.2020 3:53pm - (Replying to @PhilHaile and @sbuhai) It's a "very minor" difference that makes big difference between science and non-science. Science invented algebra, so that we won't need to labor the solution to every equation afresh, but use their commonalities. Every quadratic eqn looks different, yet the formula is the same.

12.6.2020 5:47am - Inspired by our recent discussions with economists who do not see a reason for a general definition of counterfactuals, I am posting a piece on Marschak Maxim - - the idea that policy questions are not tied to model parameters but have life of their own.

12.6.2020 4:25am - (Replying to @sbuhai and @PhilHaile) It was not clear to me before we started, because I heard Berry say: "It's all textbook stuff", and Angrist has a whole video on "the road not taken". No computer scientist would believe that a field that swears by counterfactuals would not define it. Now I understand. Relieved.

12.6.2020 3:53pm - (Replying to @PhilHaile and @sbuhai) It's a "very minor" difference that makes big difference between science and non-science. Science invented algebra, so that we won't need to labor the solution to every equation afresh, but use their commonalities. Every quadratic eqn looks different, yet the formula is the same.

12.6.2020 2:50am - (Replying to @sbuhai and @PhilHaile) I think you are reading too much into this discussion. Luckily, it has narrowed down to one easy question: whether it is helpful to define Y_x in terms of a general SEM or let each researcher define his/her own research question from scratch on the specific SEM at hand. Easy.

12.5.2020 4:03pm - (1/ )(Replying to @PhilHaile and @sbuhai) This is an interesting take, and it explains why economists had hard time understanding what I mean by "define a counterfactual Y_x" or even "identify a query", instead of "identify a parameter". It also explains why economists where less than super-excited with the "First Law 1/
12.5.2020 4:08pm - (2/ ) (Replying to @yudapearl @PhilHaile and @sbuhai) of Causal Inference" I once went with Heckman on this issue, and we agreed (so I thought) on the need to DEFINE a counterfactual as a means for making ALL queries "well posed". Evidently, our agreement did not penetrate the rest of the eco-sphere, 2/
12.5.2020 4:15pm - (3/4) (Replying to @yudapearl @PhilHaile and @sbuhai) leaving me with the burden of convincing economists of the wisdom, benefit and beauty of defining Y_x as a property of (every) structural equation model and thus making so many other research questions "well posed", including mediation, explanation, attribution, and more. 3/4
12.5.2020 4:23pm - (4/4) (Replying to @yudapearl @PhilHaile and @sbuhai) Thanks for alerting me to this missing link in communication. I will attend to it in the near future and provide a full discussion on all those research questions that become "well posed" once we define Y_x. In the meantime, enjoy "well posedness" here 4/4

12.5.2020 2:23pm - (1/n) (Replying to @PhilHaile and @sbuhai) Your basic characterization is my Oracle. We need however to expand a bit on the process of taking "quetions" and "properly posing them within a given model." That process requires a DEFINITION of the terms appearing in the question. For example, if I ask for "the effect of ..1/n
12.5.2020 2:29pm - (2/n) (Replying to @yudapearl @PhilHaile and @sbuhai) education on earning" I need to find a variable named "education" in the model, another variable named "earning", and something that would capture the term "effect on". In linear systems "effects" are conveniently captured by parameters like structural coefficients, so ..2/n
12.5.2020 2:36pm - (3/n) (Replying to @yudapearl @PhilHaile and @sbuhai) so analysts can skip the step of DEFINITION and go straight to identifying the parameter of interest and conclude: My query is identified. Done. But for more complicated queries a DEFITION is needed, else confusion would emerge. Consider: "What is the DIRECT EFFECT of ... 3/n
12.5.2020 2:46pm - (Replying to @yudapearl @PhilHaile and @sbuhai) education on earning, unmediated by 'skill'" in non-linear (or nonparametric) systems. We must now DEFINE what we mean by "unmediated", and it may not map into one parameter, but many. In nonparametric systems it might map into some combination of the underlying structural ...
12.5.2020 2:54pm - (Replying to @yudapearl @PhilHaile and @sbuhai) functions which is identifiable, even though each component function is not identifiable. That is the role of a DEFINITION. ie. "find that combination of functions which properly captures your question". If you can do if for the counterfactual question: "what would Joe's salary
12.5.2020 2:58pm - ( /n) (Replying to @yudapearl @PhilHaile and @sbuhai) be, had he had 2 more years of college, given that his current salary is S" then I would say that we have DEFINED a counterfactual, and that that the question is "properly posed" within a model. As you and I know, this mapping is not trivial; mediation analysts have had great /n
12.5.2020 3:04pm - (n/n) (Replying to @yudapearl @PhilHaile and @sbuhai) difficulty defining what they mean by "unmediated by M". But, once we define what we mean by "Y_x", namely "the value of Y had X been x" ALL questions of interest would become "well posed" for EVERY structural model. Isn't it great? That's why I asked: how do you define Y_x. n/n

12.5.2020 7:21am - (Replying to @peder_isager) In this case, the illusion was the appearance of correlation (or statistical dependence) between the independent causes, giving rise to the theory that some locations receive more bullets than others (perhaps by being more exposed) when in fact the enemy sprayed them uniformly.

12.5.2020 6:27am - I think UK Epidemiologists would find the "First Law of Causal Reasoning" to be an eye opener, see I know that even DAG-full epi. texts deprive readers of this enlightenment, and leave them helpless when it comes to probabilities of counterfactuals.

12.5.2020 6:05am - (Replying to @LeiferJoshua) Jordan Ellenberg's "How Not to Be Wrong" has a whole section on Abraham Wald and on "explaining away", though no DAGs (too bad). For DAGs, try #Bookofwhy and then Primer

12.5.2020 5:08am - (Replying to @ernestosakurai) A "variable" need not be quantitative. "Location name" is as good a variable as "temperature", its values are: {engine, tail, etc.}

12.5.2020 4:47am - (Replying to @PWGTennant) Here is a classic case of selection bias: I've met only ONE DAG-less UK epidemiologist (Davie Smith, who wrote "The Tale Wagged by the DAG" and I rushed to assume they disappeared. Hasty me! I don't see the "typical epidemiologist". Survival bias.

12.5.2020 4:24am - (Replying to @ben_golub) ???

12.5.2020 4:16am - (Replying to @PWGTennant) Where are those "traditional epidemiologists"? I thought they all disappeared in 1999, after this paper

12.5.2020 4:01am - (Replying to @socialscienceg3) My intuition came from the Garden of Eden, where Eve said: "The serpent deceived me, and I ate." Hoping this would "explain away" her action and reduce her responsibility. See #Bookofwhy

12.5.2020 3:29am - The corresponding DAG for Wald's airplane was a collider
# of holes ---->survival<----location The army observed more holes (per sq. inch) in the fuselage than the engine, & recommended more armor on the former. This dependence however was created by selection bias.

12.4.2020 6:24pm - (Replying to @JulienSLauret and @sejnowski) I am waiting for someone to explain it to me "from a causal perspective". So far, the closest connection I've found is, and I can't see the relation to Sejnowski's paper. Open to learning.

12.4.2020 5:33am - "The unreasonable effectiveness of deep learning in artificial intelligence": Though I do not agree with the overall philosophy of DL, I find this article by @sejnowski to be both entertaining, informative, and challenging.

12.4.2020 5:09am - Some authors say that graphs are not as useful as PO methods, because "it requires inputs from the user about the true underlying causal structure that are hard to come by." See: What they mean: "I've never solved a problem by the 2 methods, side by side."

12.4.2020 4:45am - (Replying to @sbuhai and @PhilHaile) Dozens of well-intending economists are advising me on how to interact with economists MORE CONSTRUCTIVELY. I am truly grateful. But, humble me, I come to learn, not to teach. All I am asking for is a definition of counterfactuals. Help anyone? @causalinf ? @EconBookClub ? @sbuhai ?

12.4.2020 3:26am - Rashida Tlaib @RashidaTlaib never misses an opportunity to undermine the Palestinian cause. She can't resist warning the world that a "state" for Palestinians means death to their neighbor.

12.4.2020 3:11am - (Replying to @sbuhai and @PhilHaile) Talking with Heckman, like citing links to 100 papers,, didn't help me define a counterfactual. I am hungry for ideas, methods, algorithms, not links. Do you buy the "one line long" definition? It contains an idea, a method and leads to algorithms of computing counterfactuals.

12.4.2020 2:55am - (Replying to @sbuhai and @PhilHaile) I keep on begging for an alternative definition, out of genuine curiosity. Because I am genuinely interested in understanding how economists think, and how they managed to survive w/o the First Law. Links do not help, I am thirsty for ideas and concepts.

12.4.2020 2:45am - (1/2) Given the interest I see among economists in the relationships between structures and counterfactuals, I venture to retweet a blog entry titled "The First Law of Causal Inference" which explains why it deserves this title:, and what you may be missing 1/2
12.4.2020 2:45am - (2/2) (Replying to @yudapearl) by ignoring it, as econ. textbooks invariably do. It's followed by "The flowers of the First Law": (1), (2) (3), which invite all economists to take advantage of the First Law, for fun, progress & more. 2/2

12.4.2020 2:08am - (Replying to @sbuhai and @PhilHaile) I would love to believe its true, and it is probably true "in essence". Yet whenever I ask an economist: define the counterfactual Y_x(u), the answer I get: "It's defined since Haavelmo" or "see Hurwicz, Koopmans and others ". Never a definition which just ONE LINE long. Strange!

12.4.2020 12:18am - (Replying to @PhilHaile) Am I right to assume that "The First Law of Causal Inference" is of no interest to economists, because they solve each problem from scratch? Hard to believe! Grouping problems into categories has much wisdom, when they share solution methods.

12.3.2020 3:24pm - (Replying to @PhilHaile) I take it then that the standard econ. approach is to take a structural model M and try to answer a query Q directly from M, without bothering to categorize Q as "counterfactual" or "policy-type" or "predictive". Am I right here? We still find some taxonomy into ATE, LATE, etc.?

12.3.2020 3:08pm - (Replying to @sbuhai and @PhilHaile) I am not begging for inclusion of DAGs. I am merely curious how you manage things w/o DAGs and w/o the fundamental definition of counterfactuals. Truly curious.

12.3.2020 7:03am - I have a soft spot for Czechoslovakia, perhaps because it saved Israel in 1948, by shipping it 4,500 riffles to defend itself against the Arab invasion, perhaps b/c the Golem of Prague was my inspirational Robot, and perhaps b/c HaTikva's music is taken from Smetana's LaMoldava.

12.3.2020 6:26am - (Replying to @stuz5000 and @hildeweerts) I dont think the distinction between "model-based explainers" and "real-world-based-explainer" is constructive when constructing explainers, since the former is the best science can give us.

12.3.2020 4:42am - (Replying to @bariweiss and @blakeflayton) Trouble is, "leaders of the organized Jewish community" take to silence when students talk. They take it as a threat when students plead that the anti-hate machinery they have been using all along is utterly outdated and in fact harmful. @DavidHarrisAJC @ADL and @HillelIntl !

12.3.2020 3:37am - We are in receipt of . For readers (like us) who thought that it bounds individual treatment effects (as in the answer is no. It deals with the distribution of outcomes under one regime, either under Treatment or under Control.

12.3.2020 2:08am - (Replying to @DKedmey and @DavidDeutschOxf) Natural language interface would surely help the invitation, but there is no need really to wait for it in order to test philosophical theories of induction or explanation. We can judge whether the theories produce plausible answers even in their raw formal dressings.

12.2.2020 10:52pm - (Replying to @Chris_Auld @PhilHaile and 7 others) While it's true that structural equations can model a decision maker "Behavior", including actions and explorations, the word "Behavioral" has an unfortunate association with Skinnerian "Behaviorism", so it may be misleading. As to "causal" why not define it formally?

12.2.2020 10:44pm - (Replying to @artistexyz) Thanks for sharing. It's a lot to read and the x/underbar notation does not make it easy. Will plow.

12.2.2020 8:58pm - (Replying to @PhilHaile @steventberry and 7 others) I agree with the distinctions you make in your slides, but some of them could be made formal, rather than conversational. The one needed most is defining counterfactuals in terms of structural equations, as given here: Any reservation to this definition?

12.2.2020 3:30pm - Remember the USC faculty response to their anti-Zionist scandal? (see Well, they are running a panel discussion tomorrow, 5 pm: with Bari Weiss. For background material, here is panelist Blake's article

12.2.2020 3:06pm - Glad convergence between CI and mainstream econ. is not so far as some thought. I have only one comment to yours: Why "your stuff" and not "ours"? Is there an alternative definition of counterfactuals beside: Y_x = Y (at M_x) ?? (Eq. (4.5) in Is there?

12.2.2020 12:37pm - (Replying to @NiekTax @hildeweerts and @dfahland) Glad to hear and ready to de-bias my sample. Where do I get more samples? The papers I read from the "explainable AI " community (eg. have no mention of "causality" or causal explanations.

12.2.2020 12:16pm - (Replying to @hildeweerts and @dfahland) My only addition to your comment is that I doubt any of the people developing or writing about "explainable AI" is aware of the causal ladder.

12.2.2020 7:09am - (Replying to @causalinf) You seem to still be living in an era when one needs to apologize for using a language of thought to represent one's thought. I was hoping that era is long gone.

12.2.2020 7:06am - (Replying to @RyanDEdwards and @causalinf) How else would you represent what you know about the world?

12.2.2020 7:02am - I can only imagine would it would save for grad students.

12.2.2020 6:42am - (Replying to @hildeweerts) The point of these example is to illustrate the difference between explaining the data-fitting strategy of a model-blind fitter vs. explaining real-life events such as death or survival. Users who expect the latter will get disappointed when they get the former.

12.2.2020 4:42am - When I see an article on explainability, ask yourself: "What does it explain? The data-fitting strategy of a model-blind fitter? or real-life events such as death or survival?" This draft belongs to the former kind. No reader of #Bookofwhy would have authored it.

12.2.2020 4:03am - (Replying to @PhilHaile @steventberry and 7 others) Agree on substance. But should a field like economics permit confusion to persist, when the relationship between counterfactuals and structure is so lucidly given in Eq. (4.5) of, and so clearly computed in Sec. 4.2.4, for both cyclic and acyclic systems?

12.2.2020 3:18am - (Replying to @AlexKale17) It's not only a matter of branding. When someone decides to explore the "philosophy of field X" it means that field X has life of its own, the aims and assumptions of which differ from those of other fields. Is this the case for "data science"?

12.2.2020 2:37am - .... And the angels are watching. The angels of justice, who slept on their watch on January 31 2002 will not let it happen again, in 2020. They will be watching vigilantly over the Supreme Court tomorrow. They are watching already.

12.1.2020 7:07pm - We are informed of a new branch of philosophy: Data-science philosophy, sponsored by ASA. It leads me to wonder what makes "data-science philosophy" different from "philosophy of science" and "philosophy of statistics"?
12.1.2020 7:07pm - (Replying to @yudapearl) \

12.1.2020 6:45am - (1/2) I was 13 when the first wave of Yemenite Jews arrived at Israel, and I soon got engaged in a campaign to collect toys from the well-fed kids of Tel-Aviv -- as welcome gifts for the Yemenite children who haven't seen a toy before. Trucks full of toys arrived at the tent sites. 1/2
12.1.2020 6:45am - (2/2) Replying to @yudapearl My memories are fairly muddy: Tents, mud, toys, mud, prayers, mud, food-line, mud, another rain?
"Our lost brethren," mutters one teacher. "On eagle wings," echoes another.
Where are they today, my brethren in mud? The tents are gone -- my brethren are my colleagues now.

11.30.2020 9:44pm - (Replying to @storyroo and @GaryMarcus) Saying "machines cannot travel backward in time" is saying: "I have not examined SCM in detail".

11.30.2020 3:05pm - Some straight lines are mathematical boredom and some are poetry in motion.

11.30.2020 6:47am - As a final note on the UN Vote of Nov. 29 1947, I am sharing @Martin_Kramer insightful essay. I'll add one more reason for marking this anniversary. It isn’t just a reminder of Arab responsibility for past misfortunes but of their continued incapacity to accept their neighbors.

11.30.2020 5:56am - (Replying to @borgesvit_r) Begging to disagree. Jews were oppressed in Muslim countries precisely because they were both (1) different and (2) stateless. Respect blooms among the equals.

11.30.2020 4:56am - (Replying to @storyroo and @GaryMarcus) I love the expression "causal inference brain". It truly captures the amazing capacity of CI to navigate all three levels of the causal hierarchy. "Trapped in mathematical probability"? Who cares if the engine under the hood speaks Swahili?

11.30.2020 4:49am - My wife had a similar experience, when forced to leave Iraq with one suitcase (1951). Yet she would not beg the UN to restore her rights to property & return. Instead, she appeals to the Iraqi Govt to recognize Israel, allowing her family to live in peace & dignity, as equals.

11.30.2020 1:38am - I normally congratulate Israeli athletes when their medals make me proud, but this lady, Linoy Ashram, is beyond normal, and breathtaking to watch. It's like seeing a beautiful Theorem unfold that I could not possibly begin to prove with my own skills. Kol Hakavod Linoy!

11.30.2020 12:49am - (Replying to @ShMMor and @Elliot_Keck) And I had a gig or two with AIPAC itself, the pinnacle of all shady powers. Gee, I wish I knew how to use their weaponized espresso.

11.30.2020 12:16am - (Replying to @quantadan @akelleh and 2 others) Hard to believe. I can't imagine how the conditions needed for correct post stratification could be articulated in Gelman's vocabulary. I would be curious, if you have a link.

11.30.2020 12:06am - (Replying to @akelleh @eliasbareinboim and @deaneckles) One more thing to notice; we did not call it "post stratification". Who invented this terminology? Note we never ask: "Is post-stratification helpful?". Instead, we ask: Is there a way of recovering from selection bias, be it post-stratification, crazy-stratification or whatever?

11.29.2020 11:55pm - (Replying to @DKedmey and @DavidDeutschOxf) Popper's philosophy, like others', factor into my ideas with a joy of invitation: Rejoyce! We can now model a "scientist's mind" on digital computer! Bring in your best ideas on induction, conjectures, explanation, refutation etc, load them on, send them to the field, and test!

11.29.2020 11:42pm - (Replying to @heem2335 @arepasdecarne and 4 others) More urgently, we cannot let a few Ex-Jews, who denounce their history and heritage take a whole nation as hostage and redefine what Jewishness is.

11.29.2020 3:54pm - (Replying to @thenateway and @jeremycorbyn) Repeating this slogan might be misinterpreted by your adversaries to mean: "ONLY Palestinian lives matter, not others," then used to distort your intentions. So, how about joining Israelis in celebrating coexistence, and commemorating Nov 29, 1947.

11.29.2020 3:26pm - (Replying to @eliasbareinboim and @akelleh) You are right! This was the first. Wow, we were young then.

11.29.2020 3:14pm - Speaking for myself, my family, friends and colleagues in Israel, we thank you, 33 countries, for giving us what other nations take for granted --a homeland, and a chance for a new beginning, in sovereignty and dignity.

11.29.2020 2:57pm - (Replying to @alphasixtyone and @jeremycorbyn) It was last year, on November 29. The 1947 UN Vote was a vote for independent Palestine and we celebrate it each year to remind ourselves of that historical commitment to coexistence. Join us.

11.29.2020 2:23pm - (Replying to @swittbit @AdaptiveAgents and 7 others) SCM respects cycles. See Causality p 27-28, 215. Why use "mental models" if it is define formally and so friendlily?

11.29.2020 7:33am - (Replying to @jeremycorbyn) Wait a moment, I thought November 29 marks the 1947 UN Vote on the Partition of Palestine and the world's commitment to coexistence. Shouldn't we celebrate both?

11.29.2020 6:58am - An enlightening podcast with @DavidDeutschOxf , partly on Popper's philosophy, partly on philosophy of science and in particular on empiricism. The relations of empiricism to AI and machine learning was lightly touched upon in my blog:

11.29.2020 4:02am - (Replying to @yudapearl @steventberry and 8 others) I am asking because I find myself unable to answer many of the questions you are asking each other. Sometimes I hear that structure is defined in terms of counterfactuals, and sometimes the other way around. What comes first? I ask myself, and how should an econ student think?

11.29.2020 3:40am - (Replying to @steventberry @lewbel and 7 others) E. Leamer's wrote (1995): “economists know very well what they mean when they use the words ‘exogenous,’ ‘structural,’ and ‘causal,’ yet no textbook author has written adequate definitions.” I have tried to explain it in: (p.135) Have things settled?

11.29.2020 2:54am - November 29, 1947, 73 yrs ago, Flashing Meadow, NY, 4 pm, the UN was holding a Vote that lasted only 3 minutes, yet changed the course of history. Gil Troy describes the mess before the Vote: This picture describes the minute after:

11.29.2020 2:06am - (Replying to @bratton) If it weren't for the Turing Test I would not take a detour from philosophy and tell myself: OK wise guy, if you think you know something about causation, write an algorithm that takes a story and answers causal questions about it as plausibly as a child does. Grateful to Turing.

11.28.2020 3:04pm - (Replying to @steventberry @lewbel and 7 others) Interesting! I have two questions. (1) Is there an operation that preserves counterfactuals which differs from the identity operation? (2) Are these fundamental questions discussed someplace in the (accessible) econ literature? Has consensus been reached on some of them?

11.28.2020 2:30pm - (Replying to @osazuwa @AdaptiveAgents and 8 others) At this point, all I am trying to understand are the inputs and outputs of PP. Not its goals, nor its methods. Given this level of description, am I right to assume: Input ="an arbitrary set S of probabilistic expressions". Output = Tightest bound on any desired expression E ??

11.28.2020 6:09am - @GaryMarcus has a much harder task than I have, trying to assess limitations of existing AI systems in the broad arena of AGI. Lucky me, focusing on the tiny arena of causal reasoning, I can use a friendly oracle that proves certain tasks undoable by current AI approaches.

11.28.2020 5:48am - (1/ ) Wishing all readers a joyous and meaningful Thanksgiving weekend, and adding a personal note for Friday, November 29. I have a secret contract with Lady History: She forgives my moments of weakness and I remember her moments of grace. One such moment was November 29, 1947
11.28.2020 5:48am - (2/ ) (Replying to @yudapearl) 2/ when I, my family, our neighbors and the whole town heard the voice from the radio: "Yugoslavia abstained! The resolution [UN 181] has passed!" And suddenly, a terrifying shout, 80 generations of age, came out of the throats of everyone in town, including pets, fish, and
11.28.2020 5:48am - (3/3) (Replying to @yudapearl) babies in their cribs: WE HAVE A STATE !!!!!!!! To fulfill my obligation to Lady History, I am sharing a 2018 oped that further describes the significance of that moment: Happy Thanksgiving!

11.28.2020 5:38am - (Replying to @BRSLWP) My point is that the combinatorial explosion prevents cheating like Searle's Chinese monkey trick, and only those who have a tractable computational model of the world (or chunks of it) can pass the test. Having such a model is intelligence. see #Bookofwhy

11.28.2020 3:10am - (Replying to @tom_santis) As a "Zionist at least", please join us for Sunday's celebration of Jewish Thanksgiving Day, commemorating Nov. 27, 1947, the day when 33 nations declared themselves "Zionist at least", by voting for coexistence. See

11.28.2020 2:45am - (Replying to @SOdaibo) Optimization is only as intelligent as the objective function that one aspires to optimize. When the latter is misguided, so is the former.

11.28.2020 2:38am - (Replying to @lewbel @causalinf and 7 others) Are your students happy with the terminology "defines behavior"? LeRoy's book expresses a 30-year frustration of econ students in understanding structural equations, i.e., what kind of 'equations' are they if they do not obey the rules of algebra? Aren't your students asking it?

11.28.2020 1:57am - Beg to differ; the Turing test has been, and still is, the engine behind AI ideas. The reason it hasn't been passed by any machine is that intelligence is a funny commodity, 'faking it is having it'.

11.28.2020 1:35am - With all the doubts I have about 'big data,' it would not have elected Libya, Sudan and Venezuela to its Human Rights Commission -- even 'small data' would have known better.

11.27.2020 6:26pm - (Replying to @Chris_Auld @steventberry and 7 others) Yes.
Q_D <----P--->Q_S Q_D---->[diff=0]<---Q_S
where diff is a selection node, observed to take on the value 0.

11.27.2020 6:17pm - (Replying to @steventberry @causalinf and 7 others) My statement is about the definition of a counterfactual. If you agree with my definition, and the 3-step procedure, than we agree on all the rest. However, this definition invokes (step-2) shutting off equations to which Heckman objects vehemently. So where are we? In agreement?

11.27.2020 4:47pm - (Replying to @lewbel @causalinf and 7 others) Is it acceptable in respectable economics circles to say that structural equation are "not algebraic" ??, namely that they are not invariant under algebraic transformatins (eg. substitution) ??

11.27.2020 4:42pm - (Replying to @Chris_Auld @causalinf and 7 others) Whatever the causal structure is, if it is expressible in equations, it sure can be expressed in SCM and enjoy the 3-step procedure for computing counterfactuals. A very helpful "gem".

11.27.2020 4:34pm - (Replying to @steventberry @causalinf and 7 others) Who claimed "don't understand", God forbid!!!. Pearl humbly suggested that it would be helpful, and any helpful trick is a gem. BTW, what IS the definition of a counterfactual, say Y_x , in structural models?

11.27.2020 4:02pm - (Replying to @causalinf @Chris_Auld and 7 others) Heckman and Pinto paper should not be read w/o tongue-in-cheek interpretation, see and DAG are acyclic, but SCM can represent cycles as shown in Causality p.215-7. Note the DEFINITION of counterfactuals -- a gem for economists.

11.27.2020 2:53pm - (Replying to @Chris_Auld @lewbel and 7 others) I see no practical harm in labeling some relationships "cause" and others "super-causes" to emphasize some perceived minor difference. Troubles begins if most causes of interest are in the "super" category, and journals reviewers start insisting on calling everything "super".

11.27.2020 8:50am - I don't think you read me (and UAE) correctly. We are saying "Zionism is coexistence" and, axiomatically, coexistence means "free Palestine," side by side w/ Israel. So, where did you get the "you do not let" idea? We are celebrating Nov 29, Sunday, a commitment to COEXISTENCE.

11.27.2020 8:23am - Where/how did Emiratis learn how to win Jewish hearts? They do not say "we are against antisemitism" (who isn't?) or "we are both Semites; descendants of Abraham" (big deal). They get straight to the core of Jewish identity: "We are for Zionism, hence coexistence"

11.27.2020 7:56am - Puzzled over the "Jewish Thanksgiving Day", November 29, readers asked to see the official resolution that was voted on by the UN General Assembly. Here it is: As you can see, it calls for Arab and Jewish states, with Jerusalem under International status.

11.27.2020 3:01am - (Replying to @AdaptiveAgents @djinnome and 6 others) I believe the discussion will be illuminated if we gain a common understanding of what "probabilistic program" (PP) is. Am I right to assume that PP takes an arbitrary set S of probabilistic expressions and outputs the tightest bound on any desired expression E implied by S. ???

11.26.2020 6:38pm - I am curious to see the reaction of economists to this book. eg @causalinf , @PHuenermund , @EconBookClub , @steventberry , especially those who took part in our discussion on "reduced form" @Chris_Auld , @analisereal , @jdramirezc , @lewbel , @PhilHaile , Is it still an issue?

11.26.2020 11:02am - (Replying to @CloudsWithCarl) Its a new algebra, axiomatized by Halpern. See Causality p.228.

11.26.2020 10:42am - (Replying to @PhilHaile @PHuenermund and @IvanWerning) I didn't get what "tyranny" is being critiqued and what resides in the twilight zone between "causal inference" and "structural empirics".

11.26.2020 10:32am - Congratulations @IrwinCotler

11.26.2020 9:56am - (Replying to @AJCGlobal and @DavidHarrisAJC) Let's not forget the Jewish Thanksgiving Day

11.26.2020 8:18am - Wishing all readers a joyous and meaningful Thanksgiving day, and adding a personal note for the Jewish Thanksgiving, Sunday, November 29: Here is an account of the events leading to the 1947 UN Vote on the partition of Palestine:

11.26.2020 6:40am - We've discussed the dire need of econometrics for a Causal Inference text; here is one: Unfortunately, it enslaves econometric equations to the rules of algebra (See Preface Eqs. (0.3)(0.4)), thus taking structure out of economics.

11.26.2020 1:11am - Just heard on Israeli TV of the death of Diego Maradona, and watched with the rest of the world some of his most brilliant moments on the soccer field and the millions of grieving Argentinians who declared a 3-days mourning to their hero and their legend. We mourn with you.

11.25.2020 12:31pm - (Replying to @jacobgmathew @tlbtlbtlb and 3 others) I agree with him. Who doesn't? Even some statisticians do, but are afraid to admit.

11.25.2020 10:11pm - Another arrival: Video of my interview at the Causal Inference Seminar: Reading material:
1. History of Bayes nets:
2. Data vs. Science:
3. On Machine Learning Research:

11.25.2020 4:34am - (Replying to @jaketapper) Tlaib is dying to be "suppressed and silenced", how else can she make her stale "Evil Israel" rants sound relevant?

11.25.2020 4:24am - (Replying to @Claire_Voltaire and @PhotonAlchemist) Not exactly. The Arabs claimed that a Jewish state would turn some Arabs, locally, from majority to minority and this, so they claimed, is unacceptable, and Jews should know why. See a brief account of claims

11.25.2020 12:42am - (Replying to @RuochengGuoASU) You are right. The invariance you show only holds If Z is a cause of X tht is not also a cause of Y. If Z is a child of Y, then P(X|Y) is invariant. We also need to assume faithfulness.

11.24.2020 10:28pm - (Replying to @learnfromerror) Now all we need to do is moving from "not giving up on" to "actually doing it." BTW, I listened to the interview - nice and illuminating. But, again, most of the issues are waiting for a formal language to be aired and, you know my bias: we have a language, and it ain't stat.

11.24.2020 2:10pm - Just arrived: Video talk by @eliasbareinboim on "Causal Data Science: Recorded as part of the Causal Data Science Meeting on April 1st, 2020.

11.24.2020 6:08am - What @RashidaTlaib means to say is: "I know everyone is bored with my "Evil Israel" rants, perhaps @ABlinken can wake up my audience. Please, silence & suppress me, I am dying to have something to say. "

11.24.2020 2:01am - It's the old puzzle of cause and effect: Did John Kerry pretend to be so sure of what he was saying to justify the failure of his "shuttle diplomacy" or did it fail because he was so blindly sure? I met him once, slippery as an eel, I can't decide between the two theories.

11.24.2020 1:38am - No comment, just inspiration.

11.23.2020 4:46pm - (Replying to @Blattberg @anniefofani and 2 others) The political community in the US names streets and schools after G. Washington and MLK, not Napoleon or Chez Guevara and, in this way, it gathers more and more common memories, to slowly and organically become a national community. Not to exclude, but to unite.

11.23.2020 3:48pm - (Replying to @Blattberg @anniefofani and 2 others) The words "favor the interests" make it sound immoral. I would say: Nationalism is an ideology that leverages the commonalities of the national community to create a cohesive and functioning society. Take US, we leverage our memories of Lincoln, MLK etc. to function as community

11.23.2020 3:02pm - Hey @RumanaRashid6 , you really blew my mind with your comment. You can't believe how honored I am to know that my book can impact bright minds like yours. Your BTW comment has changed my gloomy perspective of the future of science education. Carry on, to impact others' minds.

11.23.2020 2:41pm - (Replying to @questionsin2014 @EBluemountain1 and 17 others) Thanks @questionsin2014 for suggesting me, but I made it a rule not to participate in any panel on antisemitism unless it explicitly addresses the moral deformity of Zionophobia.

11.23.2020 8:38am - (Replying to @RuochengGuoASU) One assumption, for example, is linearity. In linear systems every counterfactual statement is identifiable whenever the causal effect is identifiable. Causality chapter 5. Also Primer chapter 4

11.23.2020 8:23am - (Replying to @thefeministcop @zeemo_n and 2 others) I am talking about words like "bullshit" which I do not understand because, as I said, I was brought up in a civilized country. Have a good day.

11.23.2020 8:18am - (Replying to @thefeministcop @zeemo_n and 2 others) "the Palestinians could never accede to Israel's demand that they recognize it as the nation-state of the Jewish people. ... I cannot change my narrative." "I am the proud son of the Netufians and the Canaanites. I've been there for 5,500 years before Joshua. " Saed Erakat, 2014.

11.23.2020 8:12am - (Replying to @thefeministcop @zeemo_n and 2 others) I dont understand this kind of language, sorry. I was brought up in Israel.

11.23.2020 8:07am - (Replying to @thefeministcop @zeemo_n and 2 others) Palestinians need a "narrative" very very badly, as attested by two observations: (1) Their relentless effort to invent one on a (questionable) genetic linkage to the Canaanites. (2) Their relentless effort to deny Jewish connection to the land. They know: Narrative = Trust Deed

11.23.2020 7:50am - (Replying to @thefeministcop @zeemo_n and 2 others) Great Try! But you won't make it in Yemen. You know why? Because you do not have the historical connection that it takes to make you "indigenous". Morocco tried it with Spain, but no Moroccan youth could remember Spanish landscape good enough to campaign for a "return to Spain."

11.23.2020 7:38am - (Replying to @zeemo_n @thefeministcop and 2 others) It's not who was there first that counts. What matters is who's maintained historical, cultural and linguistic connection to it. The Palestinians call it "narrative", and they are trying very very hard to create one. They understand that a "narrative" is a trust deed to a land.

11.23.2020 7:27am - (Replying to @thefeministcop @zeemo_n and 2 others) I thought we agreed on the "equally indigenous" principle. If we did, then it is not "migrate to a land" but "return to their land" and all these populist slogans "expelling" "occupy" "steal" etc. turn empty. If we did not agree, we should go to the definition of "indigenous".

11.23.2020 7:17am - (Replying to @readingafatbook @kyleivan31004 and @EinatWilf) I thought the Palestinians aspire to have one. No? And I also heard they were offered one (1947) but decided to wait till they dismantle their neighbors' home. Correct me if my history book does not match yours.

11.23.2020 6:57am - (Replying to @thefeministcop @zeemo_n and 2 others) When a group of people has right to self determination and their neighbors prevent them from realizing that right, attack them, & promise to attack them again if strong enough, the group naturally defends itself and tries to prevent the attacker from repeating. No ethnicity here.

11.23.2020 6:43am - (Replying to @thefeministcop @zeemo_n and 2 others) Take me; I am an atheist, yet Israel is my ancestral land. Not because the "Holy One" said so, but because my family celebrates holidays connected with that land, my childhood heroes (Abraham, David...) acted there, and my father (+his) prayed 3 times/day to return to that land.

11.23.2020 6:21am - Wrong link. The translation between ignorability and admissibility is given here:

11.23.2020 6:01am - (1/n) Here is a new and comprehensive survey of methods for combining experimental and observational studies: I would supplement it by noting that: (1) " The SCM and PO frameworks are" NOT "complementary, with different strengths discussed." Rather, 1/n
11.23.2020 6:01am - (2/n) (Replying to @yudapearl) the two are equivalent: Every result obtained in one is also obtainable in the other. (2) Imbens (2019) is the last place I would look for comparing the two frameworks (see why: (3) A translation between the ignorability condition used in PO and the 2/n
11.23.2020 6:01am - (Replying to @yudapearl) and the admissibility condition used in SCM is given and discussed here: Finally (4) I am not aware of any assumption-free method of combination that would improve over RCT, except for counterfactual queries, as in Any ideas?

11.23.2020 5:11am - (Replying to @zeemo_n @thefeministcop and 2 others) Hating Jews is not more evil than being a Zionophobe and hating only those Jews who want to live in a sovereign Jewish state. Islamophobes, likewise, do not hate ALL Muslims, only those who believe in the teachings of Prophet Mohammad. They claim that even some Muslims don't.

11.23.2020 1:58am - A expression I'm getting from many Zionophobes, on first confrontation with the facts: Come on! Me? a racist? Me? A super-righteous hailed by all BDS trumpeters? You must be kidding! Me? Champion of freedom for all oppressed people? What? Another people? Jewish? A people? 1/2
11.23.2020 1:58am - (Replying to @yudapearl) Equally indigenous? Seeking dignity? Coexistence? The villains? You must be joking! Denied normalcy? That can't be! I'm not a bigot! I'm super pure! ..... Tell me more about Zionism.... Tell her more @EinatWilf They don't teach Zionism-101 in schools any more. Too old-fashion.

11.23.2020 12:57am - (Replying to @RuochengGuoASU) I use different notation for counterfactuals in order to distinguish sentences estimable from experimental studies from those that aren't. Do-calculus can be used for counterfactuals (same as using prob. for experiments) but it ain't sufficient, w/o additional assumptions.

11.23.2020 12:15am - (Replying to @thefeministcop and @EinatWilf) Glad you are exploring Zionism by asking a Zionist, not a BDS propagandist. A "Jewish state" means a national homeland for the Jewish people, same as a national homeland for the Palestinian people (recall, majority of Israeli's are atheist -- it's peoplehood that makes us Jewish)

11.22.2020 11:37pm - (Replying to @thefeministcop and @EinatWilf) My book is full of learned studies and practical advice on how to fight classical antisemitism. It's missing one anti-body, against a new mutation, Zionophobia, an ideology that denies one group of people the right to live in dignity, as equally indigenous, on their land.

11.22.2020 3:57pm - (Replying to @ThomSeaton and @RabbiWolpe) My My, You won't believe it, but instacart just delivered 3 bars of Hagen-Daz to my door - I am not making it up, and will rush to check my computational skills.

11.22.2020 3:11pm - (Replying to @IsaacRothberg and @cholent_lover) Shocking! isn't it? How else can you explain Saeb Erakat attempt to appropriate the history of the Netufians and the Canaanites, of which Palestinians know absolutely nothing? As to peaceful aspirations, have you met a Palestinian leader willing to accept their neighbors?

11.22.2020 1:50pm - (Replying to @blagojevism and @EinatWilf) Luckily we are enjoying another form of nationalism, 2000 year old, fresh as a baby, thriving, creative, friendly and compelling. The kind that Palestinians are dying to create, with such great difficulty, lacking history and lacking peaceful dreams.

11.22.2020 7:40am - We have brought this kind of Kangaroo panels on ourselves by calling anti-Zionism "antisemitism", and agreeing to let it be handled by scholars of antisemitism. What used to be a convenient license for inaction now turned into a weapon for the worst action.

11.22.2020 12:53am - I am surprised @doctorow takes "underspecification" to be the "wrecking ball" of ML. To me, the paper shows how difficult it is for ML folks to diagnose faults in a language that suggests their remedies, perhaps suspecting that any such remedy entails a ground-up reframing of ML.

11.22.2020 12:02am - (Replying to @Blattberg @AndrewPessin and @EinatWilf) You asked for example of dangerous anti-Zionists. Here they are: They hide under "dismantling antisemitism" to dismantle Israel, the soul of Jewish identity. Charging them with bigotry they can't pretend to dismantle, Zionophobia, would stop this charade.

11.21.2020 10:41pm - (Replying to @artistexyz @joscani and 2 others) No hablo español, solo canto español.

11.21.2020 10:36pm - (Replying to @Blattberg @davidmanheim and 2 others) Surely they were not atheists, no one was atheist then, everyone had a God of some sort. The question is what was that God a symbol of. In the Jewish tradition, God's role in shaping a people's history is central, more so than, say, His role in managing weather or the afterlife.

11.21.2020 10:13pm - As a co-author of a book on probabilistic graphical models, Daphne Koller knows what "interventions" are and how to model them for drug discovery. But how do her ML employees take it?

11.21.2020 10:00pm - (Replying to @kyleivan31004 and @EinatWilf) The logic is right, but labels help strategize. The difference between a Zionophobe and an antisemite is that the latter never admits what he/she is. That makes it extremely hard to press charges and win a court case. The former can be roasted on its own admission.

11.21.2020 9:36pm - (Replying to @Blattberg @AndrewPessin and @EinatWilf) The most dangerous anti-Zionists today are the congressional Squad. Take Israel away, and they would love Jewish food and culture. They are not 'crazy,' but rather hateful, calculated and methodical, working compulsively towards the elimination of the world's greatest evil.

11.21.2020 2:04pm - (Replying to @Blattberg @davidmanheim and 2 others) Judaism was a peoplehood before becoming a religion: Exodus came before Mt. Sinai, Amech Ami V'Elohaich Elohai. Anochi Adonai.. Asher Hotseticha... (what qualifies me as Adonai is our common history)

11.21.2020 5:28am - This is a commercial, you can skip. How do I know that I've written a damn good book? Simple, I enjoy reading it even 2 years after publication. In fact I was thrilled to read it today, after @Joscani mentioned S. Wright and his lonely stand against the statistical establishment.

11.21.2020 5:13am - Who said DNN folks know what it takes to go beyond curve-fitting? Today I read a couple of articles on @Deep_AI . Lo and Behold: authors truly believe you can get explainability and interpretability by local linear operations on the NN. Join me in reading

11.21.2020 3:14am - (Replying to @joscani @eoteromuras and @kamromero) Wright became my hero after I understood what it means to tell the high priests of statistics: "Sorry, you have a terrible blind spot in your thinking". Now I wish the high priests of ML/DL would read Wright's rebuttal of Sam Karlin's "model-free approach" (#Bookofwhy p.88).

11.21.2020 12:19am - (Replying to @eliasbareinboim @guilhermejd1 and 3 others) I would go as far as saying "least committal", unless someone demonstrate that certain parametric knowledge can be articulated more reliably than structural knowledge.

11.21.2020 12:02am - (Replying to @awhillas and @EyalShay2) On the contrary, the media I consume paints Arabs as angelic victims, incapable of wrong doing. And if you come to my campus you can see where it comes from - they actually teach it in the History Dept. That's how Zionophobes learn what media is watched by people they don't know.

11.20.2020 2:46pm - (Replying to @deaneckles @eliasbareinboim and 2 others) No reason for frustration. Anyone with new ideas and promising directions can formalize them and launch an effort to "solve" them. This does not absolve the new formalizer from knowing the extent to which the earlier formulation of the problem was "solved".

11.20.2020 6:42am - (Replying to @HolgerSteinmetz @psforscher and 3 others) See if this paper answers you question:

11.20.2020 6:24am - (Replying to @PWGTennant @AndersHuitfeldt and @PHuenermund) For those who have seen the same methodology capable of taking us from any context to any other, the distinction between generalizability and transportability seems over-stated. See This includes recovery from selection bias.

11.20.2020 6:36am - (Replying to @AndersHuitfeldt and @PHuenermund) To "solve" means to find out if the assumptions I am willing to defend are sufficient for estimating the research question I asked. If not, I want to know it. If yes, I want to know HOW.

11.20.2020 6:07am - (Replying to @AndersHuitfeldt and @PHuenermund) By all means; there is much to be done in the future. I still have good reasons to wonder how much they appreciate the completeness results we have today; such appreciation would help them find "less restrictive" solutions.

11.20.2020 5:53am - (Replying to @saribashi) Rigor teaches us that a stain of racism soils those who single out Israel when: (1) Majority of states are based on "ethnic identity" (2) Palestinians are laboring to fabricate "ethnic identity" as a basis for their statehood. (3)Israel's basis is much more than "ethnic identity"

11.20.2020 5:39am - (Replying to @nimish15shah and @johnurbanik) I did use harsh and perhaps inappropriate words when I misunderstood the intended audience. At the same time, someone must take the risk and scream "wrong way" when leaders are not doing it. And they can't, they don't have a toy problem to guide them. I don't blame them.

11.20.2020 12:31am - (Replying to @edzaizv) This precisely what "external validity" is about: "retain deep commonalities and discard deep disparities". I am emphasizing "deep," because you can't do it by surface commonalities and disparities. Causal models give you the depth.

11.19.2020 10:31pm - (1/2) I have always maintained that Zionophobia needs to be fought as a unique moral pathology, separate from anti-Semitism: Today's decision by the State Dpt to classify BDS as anti-Semitic may help legal efforts to curb BDS's abuses of US campuses, but 1/2
11.19.2020 10:31pm - (Replying to @yudapearl) but much more needs to be done in the moral arena. BDS leaders continue to confess their ultimate goal of eliminating Israel.

11.19.2020 5:37pm - (Replying to @DaveBrady72 @SLMsociology and 2 others) Model is a heavy responsibility, I hope I can live up to expectations.

11.19.2020 5:30pm - Exciting News: First "causal inference" initiative, proudly named, not hiding under "data-science" or "machine learning" umbrellas: Kuddos to Maya Peterson and Mark van der Laan. I hope it leads to 12 "Causal Science Centers" by visionary donors.

11.19.2020 9:39pm - Happy to see political scientists realizing that the so called "external validity" problem is essentially solved using modern CI methods: I am not sure though if practitioners and disciples of Campbell's school truly appreciate how complete the solution is

11.19.2020 3:41pm - (Replying to @eigenhector) We need a glossary of ML jargon: I do not know what "ML unit testing" is. Should I? It does not sound like a toy example, else I would hear about it.

11.19.2020 2:55pm - (1/n) In fact toy examples *MUST* be ADDED to explain why more data ain't the solution. It's been over 20 years now since DARPA has launched program after program in ML robustness, "domain adaptation", "lifelong learning" "transfer learning" etc etc. Hundreds of millions of dollars 1/n
11.19.2020 2:55pm - (2/2) (Replying to @yudapearl) have poured into those programs and I havn't seen ONE paper trying any of their products demonstrated on a toy example. It would have revealed early on that, w/o causal inputs, robustness is unachievable. In short, the dismissal of toy examples costed US taxpayer a fortune. 2/2

11.19.2020 2:30pm - (Replying to @johnurbanik and @eliasbareinboim) Well taken, and thanks for explaining the positive side of "underspecification" papers. I like your (quote): "Toy examples do not enlighten someone who thinks the solution is just more data." But I think toy examples should be ADDED to explain why more data ain't the solution.

11.19.2020 6:16am - Elaborating: toy examples come with theoretical boundaries of what's doable, what's not doable and what must be measured or assumed to make things doable. By showing how your algorithms run on such an example, we learn if the'r aimed at expanding those boundaries, circumvent them
11.19.2020 6:16am - (Replying to @yudapearl) or just scale them up. Absent such demonstrations, bystanders can't begin to understand what is attempted, especially when the algorithms are hyped as "new framework", "new approach" or "new paradigm" and decorated with home-grown titles like..[witheld to maintain friendships]...

11.19.2020 5:52am - (Replying to @alexdamour) I am happy indeed to know that you and like-minded ML teams are heading toward coherent inference. But I must confess: It does not show. You lose us, bystanders, by trying to leap straight towards big success stories, instead of demonstrating your ideas on toy examples first.

11.19.2020 4:12am - Authors of modern causal analysis: If you ever get your paper rejected with the argument below: (1) Remind the Chief Editor that this is the 21st Century, and (2) Educate the reviewer that clearly stated causal assumptions are no less "causal" than any so called "causal design".

11.19.2020 3:51am - (Replying to @omaclaren and @alexdamour) While we should not insist on DAGs, we should insist on a coherent logic of inference. See, for example, the general structure of the inference engine in Figure 2 of

11.19.2020 3:37am - Nature Comms' insistence on deleting causal language takes us a century backward, to the days of Karl Pearson. Still, the use of "matching" should have been justified on causal grounds to rebut the Reviewer's objections.

11.19.2020 2:59am - (Replying to @NandoDF) I need to look into the "game setting" and make sure I understand what is meant by "spurious correlation". Any link to where it is described using 2-3 variables. The paper you cite says: " we propose an alternative data-driven solution". It worries me, knowing it's impossible.

11.19.2020 2:01am - (Replying to @awhillas and @EyalShay2) The "racist stench" stains all who preach "Me Me Me" and can't utter the words "equally indigenous". Thus far, only Zionists made this their credo, and not a single Zionophobe that I know.

11.19.2020 12:33am - (Replying to @alexdamour @XiaohuaZhai and 15 others) @alexdamour , I am retweeting this: to make sure it reaches some of your co-authors in the ML bubble, where the ideas that something is "provably doable" or "provably undoable" need still to take roots.

11.19.2020 12:05am - (1/3) I am also concerned with "communication strategies." Here is what I tweeted earlier about "underspecification": I am putting myself in the shoes of a ML researcher and wondering if I would be convinced that "identification matters" without ever 1/3
11.19.2020 12:05am - (2/3) (Replying to @yudapearl) seeing identification achievabled, and provably so, in some simple examples, and without ever studying some basic limitations on identification as, for example, in the Ladder of Causation. More specifically, since our main concerns were "robustness," I would not be convinced 2/3
11.19.2020 12:05am - (Replying to @yudapearl) unless shown an example where the same probability shift is caused by two different "domain shifts", indistinguishable in my data, yet each calling for different adaptation strategy, again provably so. (As shown eg here )

11.18.2020 8:36pm - (1/2) Hilarious! And bordering on the grotesque. 40 Google scholars searching for a bug and all they can come up with is, hold your breath: "undespecification." Same diagnosis that statisticians discovered when they could not understand Simpson's paradox 1/2
11.18.2020 8:36pm - (2/2) (Replying to @yudapearl) It's the same diagnosis, "underspecification", that a pre-algebra child would discover trying to solve 2 equations with 3 unknowns. The same diagnosis that one would pronounce when too lazy or too tired to think about what it takes to correctly specify the "underspecified". 2/2

11.18.2020 7:53pm - To supplement your nice exposition of the *Front-door Criterion* here is a simple (1993) derivation of the formula, not using do-calculus, just ordinary probabilities:

11.18.2020 6:34pm - (Replying to @LentoBio) I have not met this forgetfulness.

11.18.2020 4:26am - Covid-19 has almost made us forget another crisis, "Climate change". Here is a 20 year review of what we knew, know and wish to know, by a top geoscientist, Michael Ghil: Seeing counterfactual logic illuminate this field adds a grain of future hope.

11.18.2020 2:57am - For readers requesting more information on the early days of probabilistic reasoning and Bayesian Networks, here are some personal recollections:

11.18.2020 1:28am - (Replying to @awhillas and @EyalShay2) Zionophobes can't get it through their skulls that the act of "kicking someone off the land" applies symmetrically to two equally legitimate and equally endogenous claimants. The Zionophobic mindset is always "Me, Me, Me," unable to smell the racist stench of that posture.

11.17.2020 11:11pm - (1/n) (Replying to @DilijanTrails and @IntuitMachine) Agree with @IntuitMachine . Saying that "underspecification [is the] key reason for these failures" is like saying that "missing data is the key to all problems", because alas we can't find the needed answer in the data. The key problem, imo, is that 40 PhD's at Google ... 1/n
11.17.2020 11:25pm - Replying to @yudapearl @DilijanTrails and @IntuitMachine) (and other ML centers) continue to talk the language of data-fitting, training, expected predictive performance, inductive bias, etc., instead of glancing at the Ladder of causation and asking what it takes to snap out of Rung-1 What a tectonic shift:

11.17.2020 10:35pm - If you treat causality as if it is only "in the mind" you would end up thinking it is "in the data", which is much more dangerous than thinking it "in Nature", not "in the mind". Beside, it is in Nature.

11.17.2020 10:29pm - (Replying to @saramagliacane) Thanks for clarifying, @saramagniacane. It is still important to remind Dawid that without SCM there is no explanation, no mediation and no personalized decisions. Though your context, understandably, was discovery.

11.17.2020 10:18pm - Many thanks @mcelentano , @EconometricaEd (Guido Imbens) and the entire interviewing team for giving me a chance to discuss the past, present and future of the causal inference enterprise, as seen from the lopsided lens of an innocent bystander.

11.17.2020 9:38pm - (Replying to @FenCameron) In my corner of the woods ALL processes are causal, and need no theory beside themselves. I hope this does not make economists uncomfortable; they have enough on their plate already.

11.17.2020 9:24pm - (Replying to @dylantmoore and @FenCameron) Honored to be on this thread but, frankly, much would get brightened if you specify what rung of the ladder you are at. eg. Is Yitzhaki formula an exercise in identification or estimation? Why not average over heterogenous effects? Over all: Speak Ladder, not 'good economists'.

11.17.2020 12:51am - (Replying to @quantadan) If you can energize Health/Social science people to write the much needed "Glossary of terms and distinctions" so much the better. I am still hoping leaders of ML will contribute their wisdom, since they are offended by utterances such as "its all just correlations"

11.16.2020 5:42am - As one who urged Zoom to cancel Khaled speech, one who knows something about terrorism and its glorification under the rubric "resistance", please read an article I wrote in 2009: "Daniel Pearl and the normalization of evil" @alicesperi piece depresses me.

11.16.2020 4:23am - (1/n) A paper that epitomizes the cultural barriers between CI and ML is Having been immersed in counterfactual reasoning since 1993, I still can't figure out what the paper says. It aims to infer "individualized causal effect" using Rung-2 assumptions, 1/n
11.16.2020 4:23am - (2/ ) (Replying to @yudapearl) which CI folks have learned to be impossible. It then proposes the use model-blind methodologies such as "representation learning" and "adversarial training" to lift us from Rung-1 to Rung-2 which, again, is known to be impossible. We need a team of translators, from ML to CI, 2/
11.16.2020 4:23am - (3/3) (Replying to @yudapearl) to bridge these unbridgeable gaps of understanding, perhaps by creating "A glossary of terms and distinctions" for both cultures to enjoy. Let me start: 1. "individualized treatment effect" is NOT "covariate-specific treatment effect", the former is Rung-3, the latter Rung-2. 3/3

11.16.2020 3:11am - Another illuminating paper at Stanford's CI seminar was which offers a comprehensive perspective of causal discovery methods invoking both observational and interventional data. Phil Dawid rebuttal: "I do not like SCM" was deflected
11.16.2020 3:11am - (Replying to @yudapearl) by Mooij with: "SCM helps us deal with cycles and latent confounders". Which puzzles me. Latent confounders are perfectly modeled in Causal Bayesian Networks (Rung-2). We need SCM to deal with counterfactuals, explanations, mediation, & going from population to individual level.

11.16.2020 1:42am - (Replying to @EdKwangl) The saying it nonsensical if the alg. is fed additional information, beside correlations; it's no longer nonesensical if ALL you feed it is correlations. An awareness of this extra information is missing from the ML literature, because it prides itself on being model-free.

11.15.2020 11:32pm - (Replying to @dfoosher) I appreciate the honesty with which you describe the ML culture and practice: "describe the program and its I/O. Anything else leads to chaos." But if the input is correlations and the output is "fairness" it is only natural that guardians of ethics would feel uncomfortable.

11.15.2020 11:18pm - Agree. I also hoped they would unveil the causal roots of the epi/econ discontent, as I tried here but, evidently, they felt more comfortable swimming at the superficial level; root causes take a shovel to expose. #EconTwitter #epitwitter #causaltwitter

11.15.2020 9:47pm - (1/n) ML philosophers can help avert this confusion by explaining the difference between "correlation-based" and "data-driven"; the former they shun with resentment, the latter they embrace with pride. Concretely, they should explicate in what way ML definitions of "fairness" and 1/n

11.15.2020 9:47pm - (2/2) (Replying to @yudapearl) other ethical terms do NOT "operate on the basis of correlations". Undoubtedly, some of them (very few) do go beyond correlation, but this crucial step is not obvious to readers (like me) trying to find it in articles laden with arg-max equations of the data-fitting culture. 2/2

11.15.2020 8:30pm - (Replying to @dacio_ferreira and @JuhoPiironen) Extremely easy. As attested by this bystander, a student of miracles and other scientific wonders.

11.15.2020 1:39pm - There is more to it. DAGs tell you that, given this state of uncertainty, your problem is unsolvable by ANY method.

11.15.2020 6:59am - (Replying to @pablogerbas and @PHuenermund) The same goes for hiding your DAG under tons of ignorability assumptions or newly invented Deep Learning tricks.

11.15.2020 6:51am - (Replying to @CloudsWithCarl and @zacharylipton) Could be but, for me, it's a reason for hesitation.

11.15.2020 4:34am - (Replying to @quantadan and @zacharylipton) I believe Greenland argues that, to live up to statisticians' ideals are arbiters of scientific thought, statistical thinking should have included causal thinking. But did it? Does it? See my doubts:

11.15.2020 3:53am - A gorgeous example of "domain adaptation". Next is a synagogue in Mecca.

11.15.2020 3:40am - Sharing past talks at Stanford's CI seminars:, where I'll be interviewed on Tuesday 1 pm. Among my favorites is Betsy Ogburn's slide #4 "What constitutes a rigorous justification?" which should be studied carefully and taken mighty seriously by ML folks.

11.14.2020 9:53pm - It's a good discussion indeed Whenever a reader mentions it, I re-read a portion, and am amazed at the clarity of arguments that we rarely see in the literature; powerful cultures and paradigms that stifle scientific progress yet remain below the radar.

11.14.2020 8:46pm - (Replying to @rmarcilhoo and @zacharylipton) What chip? What shoulder? Please help this innocent bystander.

11.14.2020 8:42pm - (Replying to @TownesZhou) The "semantic relations" that you ascribe to KG are rung-1, noted by "What is" (eg. "Is there a...".) Causal relations have their own semantics, captured by "listens to..." and it would be nice indeed to marry the two.

11.14.2020 7:26am - (Replying to @AdiShavit) I am not aware of any such study, though I have posted an assessment of why it is absent: "10 years from now historians will be asking: How could scientific leaders of the time allow society to invest almost all its resources in data-fitting technologies?

11.14.2020 6:52am - Back to the courts. How many man-hours will be saved if we could automate the generation of court’s view in civil cases, including of course all pertinent facts and plaintiff’s claims. Here is a causal perspective of the problem: Counterfactuals r needed.

11.14.2020 6:26am - For the cognitive scientists among us, here is a paper dealing with the "explaining away" effect (also know as "collider effect"), how people perceive it, and how they deviate from the dictates of its formal model:

11.14.2020 6:00am - I have never been bothered by the tension between causal ecumenism and causal exclusion, the topic of this paper: I am glad nevertheless that philosophers are beginning to understand that Bayesian Networks is the proper arena for discussing such issues

11.14.2020 5:19am - How do we model an agent's epistemic state? I was fairly content equating it with SCM, the only model we have that manages all 3 rungs of the ladder. But this paper proposes to add an explicit "knowledge operator", as in "Joe KNOWS he can't do(x)".

11.14.2020 4:20am - (Replying to @omokasha) This is not politics but identity. Zionism is more central to Jewish identity than the belief in afterlife. Part of waking up to reality is accepting this fact, and re-interpreting history in its light.

11.14.2020 4:11am - (Replying to @EyalShay2 and @awhillas) Kuddos for trying to fight mud with facts.

11.14.2020 12:43am - (Replying to @awhillas) Holy Goodness, you sound like a Zionophobe: Throw mud at someone you don't know then ask: "Aren't you a bit dirty?" Relax! This "rightwing" phobia is in your imagination; there are more progressives in the Zionist movement than you can find among AOC followers. Relax!

11.14.2020 12:08am - These three ladies are my 2020's heroines. I have had the privilege of communicating with Rose Ritch, when USC professors issued their Open Letter, and I have been inspired since to continue acting on behalf of students under Zionophobic assaults .

11.13.2020 6:42am - No comment necessary

11.13.2020 3:47am - (Replying to @omokasha) An important component of an honest dialogue is to let each side define itself. Arabs hate it when European Orientalists try to define them. Similarly, let Zionists define themselves; don't judge them by inaccurate definitions and malicious accusations contrived by Zionophobes.

11.13.2020 2:00am - MY! my! This is the greatest interfaith article I have read since I started reading them. Finally, young Muslims who understand that the fiercest animosity between the two faiths revolves around anti-Zionism, not around anti-Semitism. Finally, a Jewish leader who dares say: ++
11.13.2020 2:00am - Replying to @yudapearl "Let's first deal with the hardest core issue, anti-Zionism", thus defying most Jewish leaders who are afraid to utter those words for fear of offending their Muslim friends. "Naming a disease is half way towards curing it" (Anon), so I am more hopeful today than I was in years.

11.13.2020 12:29am - (Replying to @edzaizv) Sure, see the "new napkin problem" #Bookofwhy page 240, and many more.

11.13.2020 12:04am - (Replying to @RaulMachadoG) Conditional and Stochastic interventions are discussed in Causality pp 131-132 in both 1st and 2nd editions. (2000 and 2009). The semantics of both are defined in terms of atomic interventions and inferences are governed by the do-calculus. sigma-do-calculus goes further.

11.12.2020 5:59am - An interesting adaptation of causal ideas to image interpretation, viewing inferences in compositional recognition as finding “which intervention caused the image?”.

11.12.2020 5:08am - I believe it was Erekat who was conflating cultural and genetic continuity - I made the distinction crystal clear. While there may be some trace of Canaanite genes in Erekat's DNA (I doubt it) there is ZERO cultural continuity here -- the only thing that counts for a "narrative".

11.12.2020 4:08am - Belated welcome to the Causal Data Science Meeting 2020. Next year you will be able to drop the "Causal" from the title; "data science" will mean "science", not "data".

11.12.2020 1:59am - It's reassuring to know that the ancestral linkage of modern Arabs to ancient Canaanites was invented 20-30 years before Erekat. Western "scholars" swallowed it like discovering New Atlantis; none asked: Do your children know anything about Canaanites? A hero? A fable? A poem?

11.12.2020 6:12pm - I met Erekat once, Natania 2008. A master of deceit. He could tell you the Netufian narrative as if Palestinian children celebrate Netufian holidays twice a year and recite Canaanite poetry since Kindergarten. Simultaneously, that he is for a 2-state solution, if only Israel ....

11.12.2020 5:37pm - Erekat will be remembered as the inventor of the "Netufian" narrative: “I am the proud son of the Netufians and the Canaanites. I’ve been there for 5,500 years before Joshua. ” Hence "Palestinians could never recognize Israel as the nation-state of the Jewish people." Logic-101.

11.12.2020 3:28am - (Replying to @hubertpaulo) For "human thought" to be "governed" by causal logic does not mean that this logic governs ALL human thought. It means that whenever causal relations are at stake (eg 'helps', 'harms', 'affects') causal logic is AVAILABLE for the mind to distinguish the plausible from implausible

11.12.2020 2:28am - Among the many, my favorite evidence is Simpson's paradox, ie, finding a drug that helps men, helps women and harms people offends our intuition, although it does not violate any law of probability. Same w/ Sure Thing Principle

11.11.2020 4:23am - (Replying to @3asangoham) I beg to differ. We are not "clueless" but "clueful". Whenever we find a logic that accounts for a new aspect of thought, we leverage the progress and go to unchartered territories. Having one that covers associations-interventions-counterfactuals is a major victory.

11.11.2020 3:43am - (Replying to @hubertpaulo) If a statement violates your intuition and does not violate the laws of probability, it means that your intuition is governed by some other logic that proves the statement "false". What is it? The only logic I know, is the logic of causation. Any alternative?

11.11.2020 3:36am - (Replying to @hubertpaulo) I did not say "all prob models should be causal". I said: "behind every STATISTICAL model there is a causal one" In addition to probabilistic models, Statisticians invoke domain understanding, choice of measurements, design of experiments, etc. all based on causal thinking.

11.11.2020 2:28am - Among the many, my favorite evidence is Simpson's paradox, ie, finding a drug that helps men, helps women and harms people offends our intuition, although it does not violate any law of probability. Same w/ Sure Thing Principle

11.11.2020 1:01am - My reasons for tweeting: "behind every statistical model there stands a causal model begging to be expressed," reflect irrefutable evidence that human thoughts are governed by causal, not probabilistic logic and that statisticians, by all available accounts, are humans.

11.10.2020 12:59pm - The funny thing to watch is how each of these tricks becomes a "framework".

11.10.2020 7:53pm - (Replying to @lawrensack) The difference is that the distribution does not need to BEG; we have the language to express it and we do. The causal model, in contrast, needs to BEG because stat textbooks do not provide a language to express it, which prevents practitioners from articulating it.

11.10.2020 7:42pm - (Replying to @analisereal @djinnome and 6 others) To make the question concrete, how do we check that your tree represents our firing squads and not the kind they use in Kamchatka, where the captain carries a pistol to be used whenever he sees fit. Let's walk through the tree and verify step be step that we are not in Kamchatka

11.10.2020 10:31am - (Replying to @jacobmbuckman and @danijarh) I am not sure what interpretation you/we are are aiming at, but if it is a causal interpretation, I would quit right here. Entropy is a probabilistic notion, so entopy-regularization or any probability-based regularization will not enable us to go from rung-1 to rung-2

11.10.2020 1:20am - (Replying to @jfvs41) Ask those who tried it.

11.10.2020 12:22am - Your words echo those of my spiritual mentor, Rabbi Jonathan Sacks: "Through being what we alone are, we give to humanity what only we can give." See AOC and her Squad howl "me! me! me!" while we say "we! we! we!".

11.9.2020 11:55pm - I once tweeted: "behind every statistical model there stands a causal model begging to be expressed," which readers probably dismissed as "causal poetry." Hearing it from Sander Greenland:, statisticians and data scientists should take a closer look.

11.9.2020 9:49pm - From Taipei to Santa Monica If you want to see the amazing musical talents currently hatching in your own backyard, tune in to World Music Day, Saturday, Nov 14, 3pm,

11.9.2020 7:19pm - Personal Reflections on Rabbi Sacks’ Life and Wisdom via @jewishjournal

11.9.2020 3:19am - (Replying to @NeuroChooser and @AOC) True. But data scientists consider not only surface data but also the ropes behind the data. How? They observe the behavior of someone who knows those ropes more than anyone else, Joe Biden, and ask why his reluctance to be visibly associated with AOC.

11.9.2020 12:32am - (Replying to @desai_pratik) That paper on complexity vs. credibility (1978) was written before Bayesian Networks (1982), while trying to anchor philosophy of induction in computational models, and discovered the Vapnick-Chevonenkis theorem. I still enjoy reading it, especially when Occam's razor comes up

11.8.2020 9:51pm - (Replying to @futureiscoming9 and @AOC) I would never judge a person's character, or the characters of those he/she listens to, unless I know something about the issues involved. I do not know much about climate change, but I know something about PM Rabin and the BDS bigots. AOC decided to listen to the latter.

11.8.2020 9:43pm - (Replying to @futureiscoming9 @anilaali and @AOC) That does not negate the theory that with visible ideological linkage to AOC, Biden would have lost the election.

11.8.2020 9:39pm - (Replying to @jacobmbuckman) Only if the preferred functions are more prevalent in nature than those discriminated against, eg additive gaussian noise,

11.8.2020 6:58am - (Replying to @HectorFromMX and @AOC) What is it about @OAC that Latin American immigrants found objectionable?

11.8.2020 6:41am - (Replying to @shoepergirl @AOC and @mehdirhasan)
11.8.2020 2:39am - (Replying to @schrepfler and @AOC) I beg to be excused from the misnomer of associating the adjective "progressive" with someone (eg @AOC ) who backs out of Rabin Memorial to appease BDS cronies. Such a deviant is either a bigot or a fool, not progressive.

11.8.2020 6:15am - (Replying to @shoepergirl and @AOC) "Progressive" does not mean "extremist". @AOC means extremist, if not worse.

11.8.2020 3:38am - (1/n) Commending you on: (1) using toy examples to demonstrate principles and (2) Starting with SCM, to make the principles transparent. I am not too happy with evoking notions such as "complexity" or "entropy" which carry an aura of magic as I argue here 1/n
11.8.2020 3:38am - (2/n) (Replying to @yudapearl) Concretely, I prefer to account for the asymmetry that you discovered in terms of the limitations imposed on the set of functions considered (as in Hoyer and Shimizu etal) which prevent them from generating ALL P(x,y) distributions. I think this perspective would be more fruitful

11.8.2020 3:03am - (Replying to @schrepfler and @AOC) I dont blame you for following the news, but it turns my stomach to see such debasement of the word "progressive." I believe even the news makes a distinction between "left wing" and members of "The Squad".

11.8.2020 2:39am - (Replying to @schrepfler and @AOC) I beg to be excused from the misnomer of associating the adjective "progressive" with someone (eg @AOC ) who backs out of Rabin Memorial to appease BDS cronies. Such a deviant is either a bigot or a fool, not progressive.

11.8.2020 12:53am - (Replying to @christatistic and @AOC) You talk to people, all Biden supporters, and ask if there is anything about him that makes them hesitate. Try it. The answers I got were, invariably, that he won't have the backbone to reign in the extremists in his party.

11.8.2020 12:19am - It just does't occur to her, @AOC , that Biden won because he pretended she does't exist.

11.7.2020 11:17pm - (Replying to @ZoubinGhahrama1 and @Corey_Yanofsky) NN does not fit x-->y or y-->x, it fits pairs (X, Y) drawn from some distribution P(x,y). Now, since ANY P(x,y) can be generated by x-->y as well as by y-->x, there is no way of telling who the generator was, unless we make some assumptions on how (in)flexible the generator is.

11.7.2020 9:40pm - Rabbi-Lord Sacks addressing the House of Lords.

11.7.2020 8:59pm - This picture of Rabbi Sacks is from my dialogue partner Dr. Akbar Ahmed It reminds me of how the 3 of us visited Jewish and Muslim schools in London 2005 and he told the kids the story of Pharaoh daughter, who rescued Moses, to become a heroine to us Jews.

11.7.2020 4:32pm - Baruch Dayan Emet. Rabbi Sacks was a true pillar of wisdom and humanity for our generation. I had the fortune of meeting him in January and handing him a copy of #Bookofwhy. He told me something that I'd never forget: They'll follow you, if you're true to yourself. I'll miss him.

11.7.2020 3:40pm - Thanks for sharing these photos from the Daniel Pearl World Music Day @Taipei . Delighted to see the spirit of Daniel empowering music lovers who refuse to give up on shared humanity and global optimism.

11.7.2020 3:30pm - (Replying to @TomTomGDN and @Ojdadana) The majority of contemporary readers (at least on this forum) know that Poland is not in Germany, and that the reality signaled by Krystallnacht applied equally to all European countries.

11.7.2020 5:05am - (Replying to @bjh_ip) Just tell me what you are trying to do, what information you have and I can ask my oracle (SCM + calculus) if it is doable or not. This is what oracles are for.

11.7.2020 5:00am - It was two days after Kristallnacht, Nov. 9, 1938, that my mother sent a telegram from Tel-Aviv to her parents in Poland: "Sell everything and come!" It was too late; the British Gov. bent to Arab riots and closed the gates, never to see them again.

11.7.2020 4:29am - (Replying to @dataengines and @djinnome) We need to distinguish "statistics" as it was practiced in the past 120 years from "statistics" as it is often presented to the public: the study of what we can learn from data. CI belongs in the latter. CI expands the former.

11.7.2020 4:22am - (Replying to @dataengines and @djinnome) The "advantage" you probably have in mind, context-specific dependence, is representable in SCM. There remains no advantage to Prob. Trees that I can see, and many disadvantages. Most notable: not seeing your own assumptions presented explicitly.

11.7.2020 4:10am - (Replying to @bjh_ip) Temporal information helps of course. But even having such information, the presence of unmeasured confounders can make two different models indistinguishable, while delivering opposite answers to a causal query. See Causality p.383. or

11.7.2020 3:33am - (Replying to @dataengines and @djinnome) CI = statistics of causality ? God forbids, I would prefer: Statistics = CI w/o causality. Namely, Stat = Rung-1 of the Ladder of causation.

11.7.2020 3:10am - Good question. Note however that (absent temporal information) any data fitted by y=f(x,eps) can be equally fitted by x=g(y,eps') . From the graph we see that neither model imposes any conditional probability constraint. It's also clear in linear models:

11.6.2020 1:42am - (Replying to @usuallyuseless) I actually used it in my last rebuttal. Haven't heard back from the nasty reviewer yet; I hope he/she is examining his/her last "real-life-data" paper saying: How true!

11.6.2020 1:32am - (Replying to @VladicaV) Isn't the causal-calculus we have a good breakthrough in math?

11.6.2020 1:29am - (Replying to @yudapearl @causalinf and @EpiEllie) "...reading Epi's papers, that my professor never cited nor read, even telling us that whatever is worth reading was authored by an economist?" @EpiEllie 's article would please the professor, while Kincaid's article would answer the student (and, hopefully, advance eco.)

11.6.2020 1:17am - (Replying to @causalinf) Thanks for the link, and kudos to @EpiEllie for the JEP article. It should I hope convince econ. that they share aspirations with Epi's, so why not be nice to each other. It's important. But if I were an inquisitive econ. student, I would ask "Why should I spend my time...1/n

11.6.2020 12:58am - (Replying to @DaniloJRezende and @rosemary_ke) Toy problems is were you learn if you are on the right track. Non-toy problems is when you hide you dont know which track you are. Would be curious about the former.

11.6.2020 12:23am - (Replying to @quantadan) It is deep: "Learning only occurs when the learnable offends the data less than its competitors". I was very happy it came up that way from the key board, after trying other ways of explaining that, lacking testable implications you can't rule out a model. Deep=worth retweeting.

11.5.2020 11:09pm - This question annoys ALL students (and professors) of ML, but they are afraid to ask. Thanks for raising it in this "no hand waving" forum. Take two causal diagrams: X-->Y and X<--Y, and ask a neural network to decide which is more probable, after seeing 10 billion samples. 1/n
11.5.2020 11:09pm - (2/n) (Replying to @yudapearl) The answer will be: No difference; each diagram scores the same fit as the other. Let's be more sophisticated: assign each diagram a prior and run a Bayesian analysis on the samples. Lo and Behold, the posteriors will equal to the priors no matter how we start. How come? 2/n
11.5.2020 11:09pm - (3/ ) (Replying to @yudapearl) Isn't a neural network supposed to learn the truth given enough data? Ans. No! Learning only occurs when the learnable offends the data less than its competitors. Our two diagrams never offend any data, so nothing is learnable. Aha! But what if our data involves interventions? 3/
11.5.2020 11:09pm - (4/ ) (Replying to @yudapearl) Now we begin to see some learning, and this is precisely the role of experimental data and randomized trials. The causal diagram is nothing but a parsimonious representation of how the environment responds to all possible interventions and their combinations. Learning 4/ Isn't a neural network supposed to learn the truth given enough data? Ans. No! Learning only occurs when the learnable offends the data less than its competitors. Our two diagrams never offend any data, so nothing is learnable. Aha! But what if our data involves interventions? 3/
11.5.2020 11:09pm - (5/ ) (Replying to @yudapearl) the best such representation from a barrage of interventions and observations is an exercise studied under the rubric "causal discovery", doing so with neural nets is an ambitious task, considering the size of the search space, but is not undoable, especially if we leverage 5/
11.5.2020 11:09pm - (6/6) (Replying to @yudapearl) the tools of causal discovery. It is hard to find a needle in a hay stack, true, but it helps to know what a needle looks like, and how it differs from the hay around it. That is why causal diagrams should be part of ML education. More on this here: 6/6

11.5.2020 8:39pm - (Replying to @nathankallus) Yes. c_i-specific effects are rung-2 and P(ind. i will be helped) is rung 3. However, it is not MY framework, it is a universal distinction. The former can be estimated from experimental studies, the latter can't. It ain't mine, its Nature's -- can't be dismissed as fiction.

11.5.2020 5:57pm - The concept of "c_i-specific" effect is extremely important, no doubt, where c_i is the set of attributes characterizing individual i. Still, within the subpopulation C=c_i it is important to know Pr(Yi(1)>Yi(0)) = Pr[individual i was helped by treatment] ="individualized effect"

11.5.2020 7:40am - Dorian Khan, the author of this article wrote to me that she had our son, Daniel Pearl, in mind when she wrote it. Genius Yasser Arafat knew already in 2001 that Western universities will eventually accept terrorism as a legitimate cultural expression.

11.5.2020 7:09am - (1/2) I have been reading several papers recently where the term "individualized treatment effect" is wrongly defined by E[Y(1)-Y(0)| C=ci] and ci is a set of characteristics associated with individual i. See Warning: This is still population-based 1/2
11.5.2020 7:09am - (2/2) (Replying to @yudapearl) treatment effect, for subpopulation C=ci. To be distinguished from truly individualized effect Y_i(1)-Y_i(0) as is treated (and bounded) here: See also Causality section 11.9.1. Watch out for possible confusions.

11.5.2020 6:29am - (1/2) Papers in the philosophy of economics normally tell us what economists have written, said or done, not what they should write, say or do. This paper by Harold Kincaid is different. It alerts economists to something they have been trying to deny for 2 decades:
11.5.2020 6:29am - (2/2) (Replying to @yudapearl) "Enormous progress has been made on causal inference and modeling in areas outside of economics." Moreover, the assertion is substantiated with examples and data. @causalinf , @PHuenermund , @econometricaEd

11.5.2020 4:24am - (Replying to @InciteDecisions) I wasn't able to, unfortunately, but my blessing goes to all participants, users, authors, and developers.

11.5.2020 12:00am - My top question would be: Tell us if the effect of X on Y is positive or negative based on data on X,Y,Z1,Z2,Z3.... generated by a Simpson's machine, Figure 3 in

11.4.2020 4:35am - (Replying to @numbersman77 @frejohk and 10 others) What's the question? What is assumed known? What kind of data you have? With these three inputs the problem is well defined and can be submitted to mathematical analysis (algorithmitized) to give you an estimand (or a warning: impossible). No hand waving.

11.3.2020 3:40pm - (Replying to @danijarh) I did not know that the "world model" people call themselves "deep RL", perhaps because I never met one. How do they model the world? Can you link to a typical sample? They should meet what we call "shallow CI" people to share notes.

11.3.2020 1:52pm - (Replying to @danijarh) I would put it a bit differently. RL agents can help Causal agents build their models from scratch, if none exists. Then, if an agent wants to learn from another, it needs to ask which var is an indicator and which manipulator, namely, it needs a causal graph,ie be a causal agent

11.3.2020 11:12am - (Replying to @danijarh) I grew up with learning systems called humans, not one of whom was ever caught trying to alter a barometer. Yet they all understood that the barometer is a good predictor but not a good manipulator. That kind of understanding we call "causality" and its laws are called Caus. Inf

11.3.2020 5:45am - (Replying to @neuroprinciples @aliceschwarze and @bschoelkopf) This overview received some nice feedback. Its missing some of the recent stuff though, transportability, missing data and more. #Bookofwhy

11.3.2020 4:13am - I am not surprised that RL-driven dogs would try to alter barometers. What is puzzling to me is why you say: "Causality within RL is beginning to receive mainstream attention" and not "RL within Causality should receive some attention." RL is a technique -- Causality is nature.

11.3.2020 2:40am - We must admit though that, if it were not for the seriousness with which the Jewish masses took the Balfour Declaration it would have remained a forgotten greeting card in the dustbin of history. As BG said: "Its not what they think that counts, but what we do with it.

11.3.2020 12:55am - I owe my life to the Balfour Declaration. If it were not for Balfour, my father would not have been able to leave pogrom-infected Europe (1923) and build my home town, near Tel Aviv, and my mother would not have been able to escape holocaust-bound Poland 1935 and have baby me.

11.2.2020 11:26pm - Our grandson, Adam, will be voting tomorrow in his first American election. Ruth and I feel as proud as he does.

11.2.2020 8:07pm - What a war tired people, on both sides.

11.2.2020 4:27pm - My faithful calendar rings again: Today is November 2nd, and I owe my debt to the Balfour Declaration which, in 1917, signaled to my grandfather that his prayers may have been answered. This was my tribute to Balfour:, on its 100th anniversary

11.2.2020 4:26am - Great lectures by Jonas Peters. Thanks for the link. I am especially intrigued by the reactions of MIT students who are exposed to causal inference for the first time in their lives. Fascinating.

11.2.2020 3:47am - Great idea - Kaggle competition for ML experts. I would start with this question: Given the data obtained here:, should Joe switch schools or not? I would then go to "patients of greatest needs"

11.2.2020 1:56am - (Replying to @omaclaren) That does not tell us why the mathematics covered in Primer would fail. Namely, what is it about their setup that necessitates new mathematics. Does it necessitate new principles too? I hope not.

11.1.2020 11:38pm - A thought. If I were to re-write Primer today, I would use this example: "How many COV-19 deaths would have been prevented by a smarter policy?" which so clearly depends on what we've learned (about the virus, the public and the system) from the outcome of the current policy.

11.1.2020 11:09pm - The NYT may be right; tantrum may seem controlled by bending to it. From a spiritual viewpoint, however, the idea that religions do not have a monopoly of holy values is a refreshing statement from a Western leader, and will pay off eventually, God willing.

11.1.2020 10:28pm - (Replying to @CasualBrady @mat_kelcey and 2 others) A valuable contribution to the toughest educational effort of the century. When you are done, share a tweet with us on what, in your view is the key difference between the ML and the CI perceptions of reality.

11.1.2020 10:17pm - I will never get tired of recommending Primer to folks wishing to lift themselves from ML culture. Honestly, where can you get an explanation of counterfactuals as in that clearly distinguishes Rung-2 from Rung-3 of the Ladder?

11.1.2020 4:57pm - Bowing to peer pressure, I am happy to share the ppt slides of my talk "The Silent History of Cause and Effect"

11.1.2020 1:11pm - (Replying to @DKedmey and @DavidDeutschOxf) It would be interesting to try this criterion on a causal model, bc such models provide miniature laboratories for "scientific state of understanding". But we need to further explicate what "hard to vary" means; who is doing the "varying" etc. Perhaps @DavidDeutschOxf can help.

11.1.2020 5:56am - (1/5) For readers waiting my assessment of the #DeepMind paper on "causal reasoning in probability trees", I now have a provisional judgment: I'm unable to find merit in using probability trees (PT), and can see many problems. I'm ready to change my mind if shown how PT's 1/5
11.1.2020 5:56am - (2/5) (Replying to @yudapearl) can be used to model the firing squad example (#Bookofwhy ch. 1). By "how" I don't mean a link to an algorithm but step by step instructions, how the tree is set up, what judgments are needed to quantify the transitional probabilities, how the tree confirms facts we expect 2/5
11.1.2020 5:56am - (3/5) (Replying to @yudapearl) to find in the story, eg., that seeing one rifleman informs us about the other, that intervening on one would not affect the other etc. etc. This paper has strengthened my belief in the power of toy examples to distinguish algorithms from substance. I hope we get a glimpse 3/5
11.1.2020 5:56am - (4/5) (Replying to @yudapearl) at one such example before we hear folks rejoicing: "Who needs a model? @DeepMind algorithm can do everything in probability! i.e., from data alone". Also doubly appreciated is the power of the Ladder of Causation to tell us, at each step, what information is absolutely needed 4/
11.1.2020 5:56am - (5/5) (Replying to @yudapearl) In particular, the Ladder tells us that causal reasoning needs "causal assumptions", and a symbolic representation to carry these assumptions. Looking at a PT, we do not see such assumptions (except temporal order) so, they must be hidden someplace. Can they be explicated? 5/5

11.1.2020 4:16am - Good luck, and a hopeful November.

11.1.2020 2:59am - In legal setting, the difference between "Sam killed Lee" and "Sam caused Lee to die" may cost Sam years in jail. But philosophers of language find other aspects of the difference to be puzzling. This paper tries to resolve the puzzles using causal models

10.31.2020 10:15pm - (Replying to @prem_k) Agree, and ready to confess ignorance of the "Invisible Oriental History of Cause and Effect." At the same time, I'm afraid it will remain invisible unless someone shows us how it can solve toy problems - say that no drug can be good for men good for women and bad for a person.

10.31.2020 9:06pm - Readers asked if my talk at the History and Philosophy of Science Meeting was recorded. Unfortunately not, but I can share the slides, titled "The Silent History of Cause and Effect":

10.31.2020 8:46pm - (Replying to @DaveBrady72 and @JennieBrand1) The DAG theory tells us when we can and when we can't tell whether W is a confounder or a mediator. Depending of course of what kind of substantive information we have, e.g., temporal order, other confounders, etc. Feel free to use them in class, meetings and in whatever occasion you are called to defend the proposition that History was unjustly silent.

10.31.2020 8:20am - (Replying to @mattshomepage @AgEconomist and 3 others) This is true for CBN (which are acyclic) but obviously not true for SCM, where Z can activate and shut-off arbitrary functions between X and Y. Off hand, I envy not a user having to express causal knowledge in "probability trees". But I'll keep on trying to find one advantage.

10.31.2020 7:20am - (Replying to @DavidDeutschOxf and @DouglasCarswell) Curious. What is your argument for judging this philosophical theory *false*. My arguments are articulated here:, so I am curious if they intersect with yours.

10.31.2020 4:49am - (Replying to @KilroyIsHeir) It is precisely to prevent legitimate complaints like those of @DoubleDownNews that I would prefer to see @Corbyn ousted on a more serious offense, one he cannot deflect: Zionophobia - An obsessive struggle to eliminate Jewish self determination.

10.31.2020 3:33am - It's a pleasure of course to eulogize #Corbin, but I wish he would be expelled on a stronger charge - Zionophobia - one he cannot deflect with "I am not what you think I am". I would rank him the most dangerous confessed Zionophobe since Ernst Bevin (1881-1951).

10.30.2020 10:24pm - (Replying to @RWerpachowski and @rravi) Come to UCLA. Or USC. And you wont find a trace of it in the public court.

10.30.2020 9:21pm - (Replying to @NelsonDaleSmith and @rravi) Every Israeli I know would give up 50% of his/her salary, for the next 100 years, to pay back the debt owed US, if only they can secure their sons and daughters' safe return home, not having to defend their families' lives, freedom and hopes for a day of normalcy.

10.30.2020 9:12pm - (Replying to @quantadan @AgEconomist and 3 others) I am inclined to trust things coming from deepminds, because they have read #Bookofwhy, and probably tried their algorithm on the firing squad example. What concerns me are the words "probability trees" because the Ladder tells us you can't get counterfactuals from probability.

10.30.2020 5:54pm - (Replying to @AgEconomist @PHuenermund and 2 others) This looks like an interesting paper to read over the weekend, with one caution: When speaking "causal reasoning" look not at the algorithm, look at the input information, and ask what judgments it carries. Probabilistic judgments, be they trees or forests, are insufficient.

10.30.2020 4:46pm - (Replying to @rravi) I don't see any parallel here. Islamophobia is tabooed by a cultural consensus that it is socially unacceptable, while Zionophobia is encouraged by a cultural consensus that it is COOL because, vocally, one minority has been more "maltreated" than the other.

10.30.2020 5:01am - Watch this student-faculty initiative carefully. The words are chosen wisely and correctly and, if traditionists do not spoil its momentum, it can turn into a game changer in the fight against Zionophobia on US Campuses. Wishing you good luck, a safe honeymoon and bon voyage.

10.30.2020 2:56am - Philosophers of statistics would no doubt be interested in this Zoom talk by Sander Greenland, November 12, 4pm, at UCLA.

10.29.2020 12:34pm - The link I have to the paper on missing data is I hope it works for you

10.29.2020 11:30am - (Replying to @MariaGlymour) So X* is not a selection node, it is what you measure when you really want to measure exposure X which remains invisible. Now I see the reason for the title "differential measurement error." I got it, thanks.

10.29.2020 6:23am - I love your motto. But I did not realize Epi folks are using "differential measurement" instead of the more traditional "selection bias". Perhaps we should change our title here to share with Epi's ways of recovering from such nasty "differentials".

10.29.2020 1:25am - This thoroughly researched essay by Martin Kramer reminds us of Nov. 2nd, and the 103 anniversary of Balfour's Declaration -- the first international recognition of the Jewish people's right to a homeland. I wrote a piece about its historical significance

10.28.2020 11:59pm - (Replying to @UCSF_Epibiostat @MariaGlymour and 5 others) Watch out, @MariaGlymour , you got a severe case of selection bias on your board and, as they say in #epitweeter, you can't humor it with ignorability talk.

10.28.2020 9:09pm - For all music lovers who happen to pass by Taipei next weekend, the Daniel Pearl World Music Day invites you to Taiwan's longest running outdoor concert: I'm often thinking: What if China invades Taiwan tomorrow, would they let it continue?

10.28.2020 3:00pm This picture, in remembrance of the Pittsburgh massacre's, strikes a personal note. Joyce Fienberg (named on one of the signs) is the wife of professor Steve Fienberg, who was a good friend, and a leading statistician at CMU, (Co-author of "Discrete Multivariate Analysis, 1975).

10.28.2020 6:10am - Hilarious! Worth every word! I bet some of Beinart's disciples will take it seriously and will ask Peter to confirm publically that He, only He is the real author of that Purim Shpiel in Jewish Current, and that he actually believes that Israel created Arab oppression. Hilarious!

10.27.2020 4:29pm - (Replying to @heikalkhan and @Sahil1V) Thanks for this concise compilation of recent papers on "counterfactual explanations"; time will tell if the warnings issued will decrease or increase the confusion between "explaining the world" and "explaining the data-fitting system"

10.27.2020 3:53pm - It is a beautiful work, agree. Which may partially explain why the priests of traditional missing-data analysis fight so hard to keep the topic esoteric, inaccessible to ordinary folks.

10.27.2020 11:24am - (Replying to @attilacsordas Statistical Time is a fascinating topic that deserves more attention than I could give it. Given a movie of Brownian motion, can we tell if it runs forward or backward in time? See p. 337.

10.27.2020 11:08am - (Replying to @richard_landes and @EinatWilf) There is another episode you should not miss: Haim Shur interview in Maariv, June, 2001, where he describes in heart-broken confession what his Palestinian "peace partners" did to him. I touch on it here:

10.27.2020 10:38am - (Replying to @Sahil1V) My concern is that tomorrow, when I present the Ladder of Causation to ML audience, they will ask me: "Counterfactuals? That's easy! You don't need a model! We (ML) can do it from the data alone, so it's Rung-1."

10.27.2020 4:54am - (1/ ) I am receiving a flood of philosophical papers discussing DAGs and SCMs. I think it is a sign that philosophers are becoming aware that we now have a new and powerful tinker-toy to play with - a computational model of an agent's "state of understanding". While not exactly
10.27.2020 4:54am - (Replying to @yudapearl) a "philosopher stone", SCM is certainly a flexible and accessible laboratory for trying out philosophical ideas about "belief", "evidence" "action" "explanation" and much more. I wish someone could explain to me what Kincaid finds missing from SCM.

10.27.2020 3:58am - (Replying to @hildeweerts) Agree but note that these so called "counterfactual explanation" systems are used to explain "decisions", not merely predictions, which makes it hard to distinguish.

10.27.2020 3:37am - (1/ ) When I see a paper on explainability, first question I ask is: "What does it explain?", the data-fitting strategy of the fitter? or real-life events such as death or survival. I believe this paper is mostly about the former, as can be seen from the
10.27.2020 3:37am - (2/3) (Replying to @yudapearl) equations and from the absence of any world-model. While it is sometimes useful to explain the data-fitting system (eg. for debugging), it is also important to distinguish this kind of counterfactual explanations from the kind generated in the causal inference literature.
10.27.2020 3:37am - (Replying to @yudapearl) Beware, a model-blind system might conclude that the rooster crow explains the sunrise. It might also explain that your loan was denied because you are a male, and would also have been denied if you were a female. I wonder how ML folks would debug this system.
10.27.2020 3:37am - (Replying to @yudapearl) Beware, a model-blind system might conclude that the rooster crow explains the sunrise.

10.26.2020 8:09pm - Is missing-data a causal problem or the other way around? The final version of a JASA submission demonstrating the former view is now posted here:, having survived the wrath of a nasty reviewer from the other camp. Enjoy, and pray for non-missing-data.

10.26.2020 5:05am - (Replying to @maliniw90th That was my impression too, skimming over dozens of ML papers. The idea of supplementing data-fitting with models of data generation is somewhat "out of policy", or, more recently, a wishful yet sloppily guided ML aspiration.

10.26.2020 4:51am - @richard_landes and @EinatWilf , I wonder why your writings about obstacles to peace do not mention the 2001 "Trojan Horse" doctrine of Faysal al Husseini which, to many of my friends, meant the final nail in the coffin of the Israeli peace camp.

10.26.2020 4:21am - A must read for anyone who wishes to understand why John Kerry failed, and why the key to peace is the understanding of its obstacles. R. Landes' thoroughly-documented and profoundly-interpreted account of the Oslo Agreement and its prolonged death.

10.26.2020 3:56am - A welcome survey of mediation analysis, written for economists. To which I would add that natural direct and indirect effects did not hatch out of syntactic manipulation of formulas, but capture meaningful concepts of necessary and sufficient explanation:

10.26.2020 1:46am - This 10 min recording shows machine-learning (ML) folks how to join the age of causal inference (CI) with minimal effort, and teaches CI folks how to estimate their hard-earned estimands using ML techniques See also

10.25.2020 3:50pm - (Replying to @michaeljcurry1) If you are (like me) doubtful of things that seems obvious, the first thing I would do is to check whether those who use statistical definitions got their PhD's before or after the causal revolution, and where; some stat (and ML) departments are still in the stone age.

10.25.2020 1:01pm - (Replying to @michaeljcurry1) "Logical impossibility", agree, may seem strong. It is a strongly needed way to call attention to the inadequacy of correlation-only measures of fairness, regardless what names they carry.

10.25.2020 1:02am - (Replying to @richard_landes @HananyaNaftali and @EinatWilf) I could not. Can you link us to a pdf version? I am eager to read it.

10.24.2020 5:52pm - (Replying to @Grady_Booch @ben_golub and 2 others) You mean "vastly disparate and correlated"? Sure! And that is when we appeal to the logic of causation and ask: Should we spend time trying to find causal rel. or not?

10.24.2020 3:13pm - All conjectures about relationships between causation and correlation should be submitted to the scrutiny and verdict of the logic of causation and correlation. We no longer have the luxury of playing with conjectures and counter-conjectures.

10.24.2020 7:00am - (Replying to @Sam32895821 @HananyaNaftali and @EinatWilf) I would like to believe that Kerry does not wish the demise of Israel and that his misconceptions were the result of naïve ignorance and wishful thinking, like those of many Europeans: Agree on borders and, OOPS, children in Ramallah will stop singing "Tel Aviv is stolen land."

10.24.2020 4:08am - For "algorithmic fairness" folks, here is a comprehensive survey & introduction to causal-based notions of fairness I am only wondering whether the words "causal-based" are justified, given that non-causal notions of "fairness" are a logical impossibility.

10.24.2020 3:42am - I've found a photo of Tel Aviv, 1909, to add to your thread. Here it is No street cars, not even streets, but you can tell by the faces under those hats that the future is inevitable.

10.24.2020 3:32am - Why do bad things happen to good people? via @ManuelFCasanova

10.24.2020 3:22am - #Sudan to normalize relations with Israel!!! Pundits will soon analyze the strategic/economic implications of this deal but, to Israelis, Sudan is also symbolic of the "Three NO's" Resolution (Khartum, 1967)-- the 1st knife in the back of our peace camp. Time to reassess symbols?

10.24.2020 2:10am - An incredible thread !!! To which I would only add Tel Aviv, 1909, which looks very much like this: --. . . . . . .--

10.24.2020 1:40am - To admire Herzl (1860-1904), you need to read my oped on the 1897 Basel Congress: and discover how profoundly true was his line: “Zionism is a homecoming to the Jewish fold even before it becomes a homecoming to the Jewish land.”

10.24.2020 1:16am - I hope you include the beautiful result of a purely non-parametric analysis of Instrumental variables, which goes under the rubric of "surrogate experiments," or Z-ID,

10.23.2020 1:22pm - (Replying to @examachine1) Legal philosophy that is not algorithmatized is pre-scientific and will be an inspiring source of dozens PhD dissertations in the next decade.

10.23.2020 5:52am - I have always suspected that the legal definition of "intent" will some day be formalized using causal models. This paper attempts to do just that: The soft link is in modeling the epistemic notion of "knowledge", which is not standard in SCM.

10.23.2020 2:32am - Sensitivity analysis ushered by graphical models endows Medelian Randomization with a sense of robustness.

10.22.2020 11:49pm - (Replying to @CGische) So why is it that whenever I ask economists: "Which parameter in YOUR OWN model is identifiable?" they sneer at the question and say: But we can solve "real-life" problems.

10.22.2020 10:13pm - It sounds like a clip from Herzl's speech at the Zionist Congress, Basel, 1897. As you note, the only thing missing is the adage: "equally legitimate and equally indigenous" which I learned in kindergarten, and which our neighbors can't pronounce. They need help.

10.22.2020 6:44pm - (Replying to @AngusReynolds94) Theoretically, no. You can put down all feasible factors and ask: find most predictive subset. Practically, however, you would like to organize the huge set of factors in a form that would facilitate debugging, in case you get counterintuitive answers. Intuition = Causality.

10.22.2020 6:34pm - (Replying to @RevDocGabriel and @matvil) You make it sound like #causality is some sort of esoteric magic. Its actually easy for students who have not taken any class in statistics or ML. Here is a do-it-yourself tool kit:

10.22.2020 6:15pm - (Replying to @DKedmey) Curious. Who wrote it?

10.22.2020 6:06pm - No comment, waiting for a better news.

10.22.2020 6:05pm - (Replying to @patrick_s_smart and @Nate_Cohn) I wish pollsters would read #Bookofwhy and get to understand that (1) statistics can't answer "what if" questions, and (2) the logic for answering them is simple.

10.22.2020 5:59pm - (Replying to @FOchsenfeld @DaveBrady72 and @JennieBrand1) Interesting question: How best to decide if a variable W is a mediator or confounder. Bounds won't help. I tried to look at your paper, but alas, it had no graphs, just labor-related variables. Is there a synopsis for ordinary folks?

10.22.2020 6:25am - (Replying to @CGische) Can techniques from "econometrics toolbox" tell us which parameter is identifiable and how? Can they recognize parameters identifiable by OLS? Others?

10.22.2020 2:33am - The shame is on University leaders who await @zoom_us philosophers to tell right from wrong

10.22.2020 6:27am - (Replying to @SilvioZaina and @zoom_us) The etymology of words is not determined by their literary translation but by their usage in society.

10.22.2020 1:53am - (Replying to @richard_landes and @Immort4l_Legacy) To many "journalists" the BBC is a prep school for fat salary jobs with Al-Jazeera; they would perform miracles to be worthy of their next job.

10.22.2020 1:38am - (Replying to @osazuwa) Even logic gate cannot be expressed in Boolean algebra. Take the logic gate called "identity" X-->Y, Boolean algebra will conclude not-Y ==> not-X, which is not sanction by the gate. Negating the output (by intervention) does not negate the input.

10.22.2020 12:02am - When you have a linear model with many variables, you might find it convenient to do causal inference using matrix algebra, as shown in this paper: assuming, of course, that you can do the identification part correctly (eg, using graphs).

10.21.2020 11:51pm - (Replying to @isacdaavid and @DKedmey) Use Physics and switch to computer science when you need to, or when you want to understand what a robot feels like.

10.21.2020 11:44pm - (Replying to @NandoDF) Sorry, dont remember Vancouver. I wish I could.

10.20.2020 10:52pm - (Replying to @jacksonwpitts and @DKedmey) And my professor used to scream: "Wrong! It all depends on whether you have a voltage source or a current source"

10.20.2020 5:26am - Many thanks

10.20.2020 12:44am - (Replying to @TMoldwin) Would appreciate a reference or a link.

10.20.2020 12:40am - (Replying to @RaphaelWimmer) Here are a couple of links.
July 26, 2020 Radical Empiricism and Machine Learning Research
July 7, 2020 Data versus Science: Contesting the Soul of Data-Science

10.19.2020 11:14pm - (Replying to @FunkoUnko) But the neural architecture behind the quest for "understanding" is vastly different from the one behind the quest for food that I doubt we can learn much from the latter to the former.

10.19.2020 10:53pm - Humans act like flies? Doubt it. Babies, unlike monkeys, engage in playful manipulations that are reward-neutral, just "out of curiosity", just to get to a state of mind they can call "understanding".

10.19.2020 6:56am - In natural discourse, the word "implication" carries causal connotations, however, since Boolean logic is incapable of expressing causal relations, material implication offers the best approximation: If I see p, I conclude q.

10.18.2020 11:35pm - (Replying to @jacobmbuckman and @danijarh) I believe you are wishing away the possibility that model-blind neural networks simply do NOT generalize, see Why is it so hard to internalize? Certain tasks require external models, others do not.

10.18.2020 11:21pm - (Replying to @jacobmbuckman and @danijarh) Thanks for the genuine attempt to submit the paradox for RL solution. However, notice that the same conclusions can be derived on the basis of the data alone, model free, w/o assuming H & M or any variable. Is there any work in RL on combining experimental and observational data?

10.18.2020 9:55pm - (Replying to @zhaphod) Running away from two fears: 1. Fear of boredom, hence a commitment to strong-AI, so that I will never find myself without a challenge. (2) Laziness, hence a commitment to tackle only easy problems, ie, with a good chance I could solve them.

10.18.2020 9:02pm - (Replying to @TeresaWatanabe and @Foodaism) I was going to write to you too, but Rob is always ahead of me. Are you planning to follow up with a story on the 43 professors who wrote this Open Letter ? That unprecedented Letter will have greater impact on campus life than the Rose-Tijani feud.

10.18.2020 7:35pm - Happy to see Verma-contraints playing such important role in causal discovery. Worth mentioning, they also play a role in the "simultaneous identify-discover" approach of Zhang et al, see

10.18.2020 5:50pm - To the attentions of readers concerned with generalizing experimental results. (I wonder who isn't?)

10.18.2020 2:19am - (Replying to @jacobmbuckman and @danijarh) But you cannot use data to query: what would be the consequence of action Z in those cases where action A led to consequence S1. Try to apply RL to recommend a school choice in this example:

10.18.2020 1:53am - I am retweeting this track because the place of RL in the causal hierarchy continues to be enigmatic to readers with good intentions. Correction: The example appears in Causality pages 35-6, not 25. For a more compelling example, see What would RL do?

10.18.2020 12:26am - (Replying to @jeasinema and @danijarh) You are right! The rooster crow precedes the sun rise.

10.18.2020 12:23am - (Replying to @shadihamid and @nytimes) HMM. Who is the victim? The beheader? The teacher? Poor NYT; caught between facts and PC vultures.

10.17.2020 9:42pm - Your picture of the "Copenhagen Business School" came right after another picture from Copenhagen, 1943:, telling the story of how Danish Jewry was rescued. Hats off, Denmark.

10.17.2020 8:55pm - (Replying to @revprez) Not necessarily. "causal thinking" is any thinking that successfully navigates the Ladder of Causation, ie. answers associational, interventional and counterfactual questions coherently. DAGs happen to be a very efficient way of facilitating such thinking. Any alternatives?

10.17.2020 8:04pm - Fascinating new paper on incorporating causal thinking in robotic actions.

10.17.2020 7:00pm - (Replying to @examachine1) I met Solomonoff a few times in the 1970's and don't recall him claiming a "complete solution to Occam's puzzle". The puzzle I describe in has evidently also puzzled V&Ch, Valiant, Blumer etal, and more, so "case closed" must have left a few gaps.

10.17.2020 6:47pm - (Replying to @examachine1) It is impossible by model-free ML methods. The examples in show that the same shift in probability may require two different repairs, depending on which structural change caused the shift. Would be thrilled to see a solution to Fig. 3. #CausalTweets

10.17.2020 3:05am - (Replying to @larosaandrea) Until I re-read Breiman's paper, a week ago, I used to say both ML and stats are "data driven", because stats models are assumptions about the data (Rung-1). Now I see that Rung-1 is split into "population-minded" (stats) and "data-minded" (ML) cultures, happily cohabitating.

10.17.2020 1:27am - (Replying to @un1crom) Footnotes is where even nasty reviewers become open-minded.

10.17.2020 1:14am - CI stands for Causal Inference, which I dare label "next generation ML" (see In CI, the idea of "distribution" allowed us to prove, for example, that all ML's efforts at achieving "transfer learning" are not-achievable (see

10.17.2020 12:54am - (Replying to @lacker) Not really. When we sample from a distribution, even a known one, it spits at us data that were never seen before. This is the mystic that iid sampling captures. Things get much more mystical when the distribution is unknown.

10.17.2020 12:42am - The puzzle of Occam's Razor (ie., are simpler theories more trustworthy?) which I've presented and partially formalized here was pursued and given a fairly complete treatment by Blumer et al (1987) in

10.17.2020 12:01am - (Replying to @NandoDF and @kareem_carr) If your goal is to model intelligent behavior you can't dismiss the causal vs association distinction as "a distraction", because our insatiable curiosity as to "how does the world works" is an indispensable pattern of intelligent behavior and is incompatible with associations.

10.16.2020 11:25pm - (Replying to @matloff and @lacker) I'm not setting up this competition, I am wondering whether there is one or not, ie, whether the whole idea of "sampling from a distribution" is just a rhetorical device that we/stat could do without. The CI style is: "prove that it can work assuming iid, else dont try".

10.16.2020 4:58pm - (Replying to @lacker and @matloff) The question to be unfolded is whether the attitude of "see if you win", also known as "it works", is in competition with the traditional attitude of "prove that it works assuming iid". I know a few cases were the latter says: impossible, and the former says: "let's try and see".

10.16.2020 3:08pm - Not sure about being "first", but vividly remember my excitement upon discovering V&Ch theorem, which I labeled "The Bernoulli theorem for the hind sighted scientist" here:

10.16.2020 2:57pm - (Replying to @jiafengchen42 and @akarp) This is true if we seek analytical bounds. But if we settle for empirical bounds, based on splitting existing data, we dont need to assume iid ~ F.

10.16.2020 6:41am - (Replying to @TPA_Debray) But the existence of some distribution is essential.

10.16.2020 6:38am - (Replying to @lacker) Indeed, I find the paradigm of AlphaZero (no sampled population) to dominate ML conversations. And I am wondering whether this paradigm is liberating, when compared to statistics, in which sampled population is axiomatic.

10.16.2020 6:32am - (Replying to @VC31415) Even the second stage requires that beliefs be represented in some form, so that the question of "accordance" can be decided.

10.16.2020 5:12am - (Replying to @VC31415) The additional structure does not come to do ID, it comes because it resides in the analyst's belief about the world. If so, how is it represented? If in structural equations, let it come forward. If in some other representation, it has to be a transparent one.

10.16.2020 4:36am - Not assuming a specific model is forgiven - you dont want to introduce unwanted bias. The question is whether not assuming ANY probability distribution penalizes you. In CI it would be devastating, for it would prevent us from proving that some model-blind tasks are impossible.

10.16.2020 3:50am - (Replying to @akarp) The assumption is the background, no doubt; this is how we were trained to think about data. But is the assumption used explicitly in the analysis? Does it appear in the equations? Does it appear in the performance guarantees reported? e.g., "it works under these distributions"

10.16.2020 3:42am - Stop Being Shocked. An important, must-read article by Bari Weiss: "American liberalism is in danger from a new ideology—one with dangerous implications" -- not only for Jews.

10.16.2020 3:11am - (Replying to @BakirGhb @richardtomsett and @matloff) Vapnik came from statistics and, before 1980, statisticians had only one oracle: "the joint distribution". They could not represent "the process" even if they understood it. Today we see that you don't have to be God to understand and reason about the process. It's easy.

10.16.2020 2:57am - My reflection on Breiman's paper has generated a lively discussion at UCLA stat dpt, now posted as ADDENDUM here It also touches on how we can tell when a model is useless w/o trying it, and why the notion of "distribution" is needed for doing it.

10.16.2020 2:03am - Moreover, in my lectures on CI, I consistently called the Joint Distribution "The Santa Clause" of statistics, or "The Oracle of all Oracles". See eg. here In CI, this oracle was replaced by SCM. Has ML anointed a new oracle too? Who? The data?

10.16.2020 1:37am - While celebrating the diversity of ML, it's intriguing to explain why we find so few (relatively to stat) references to "population" or "distribution" in the ML literature and to ask whether this trend is liberating or limiting. Opinion? "Distribution" is/was the ORACLE of stat.

10.16.2020 1:19am - (Replying to @questionsin2014) My, My! Zionophobes will never stop to surprise homosapiens. But what on earth does flat earth theory to do with the dept of Architecture?

10.16.2020 12:59am - (Replying to @questionsin2014) I am not aware of the funambulist letter. Still, Cornell is wrong for rejecting it on the grounds of "multiple viewpoints". The rejection should have been on moral grounds.

10.16.2020 12:52am - Thanks @questionsin2014 for reminding me of this oped, where the word Zionophobia first made it to the LATimes. They would not allow it today, since they have zero interest in what's going on US Campuses, eg. USC

10.15.2020 6:36pm - This is one of the most profound statements I've read in the past year: "stat people assume the data come from a population, while for ML people, the data are the data, period." Has any philosopher (of science) analyzed the pro and con of the "population" assumption?

10.15.2020 1:26pm - (Replying to @the_engi_nerd) But if models are abstractions of reality then, by definition, they leave out details and, by definition, they err on the details they leave out, so, by definition, they are "wrong" where they err. Isn't it what we should expect from every model?

10.15.2020 5:50am - What are simple examples of Class X and Class Y for which we have an answer to the question you posed?

10.15.2020 5:28am - Totally agree, especially because "purpose" is encodable as a query: e.g., "What's the effect of X on Y, etc". But what about the model? What features of a model would tell me if (once used) would advance me toward answering my query.?

10.15.2020 5:16am - All replies I got on this tweet are thoughtful and creative but remember folks, I am a dumb robot (ie. automated scientist); someone hands me two models that I've never seen before, and I need to decide which is more useful. What features do I examine, and with what algorithm?

10.15.2020 4:57am - This permissiveness, however, carries the danger of encouraging sloppy modeling; after all, if they are all wrong, why bother?

10.15.2020 4:49am - (Replying to @MaccormickIan) Thanks, I wasn't aware of this addendum, though it still borders on the trivial.

10.15.2020 4:45am - The relation to "value of information" (VOI) is interesting, because it raises the question: What do we need to know about a model (prior to actually using it) to decide if it's useful. BOI is defined on information sources, not on models, but perhaps it can be generalized.

10.15.2020 4:34am - Two comments: 1) An instrumental variable, is NOT conditionally independent of the outcome given the treatment. 2) How do you reconcile your result with the completeness theorem of:

10.15.2020 3:25am - I have always felt that G Box's statement: "All models are wrong, but some are useful" is trivially true but hardly useful. As one of the most quoted aphorism in statistics, it ought to have given us some clue as to what makes one model more useful than another - it doesn't.

10.14.2020 11:54pm - What you are describing is the story of statistics from its birth. Except the "though not formalized" is no longer justified, because the main hindrance to formalization, lack of mathematical language, is not longer valid, and it is a friendly language too, albeit suppressed.

10.14.2020 5:05pm - Remember Leo Breiman's influential paper: "Statistical Modeling: The Two Cultures" (2001)" ?? Well, I re-read the paper last week, and posted "Causally Colored Reflections" on it: Modern statisticians are still debating what "statistical models" are.

10.14.2020 6:18am - Some racists can't get it through their skull that "being middle eastern" is a historical-cultural, not a racial identity. Gal Gadot can sing dozens of mid-eastern Hebrew songs, the lyrics of which is 2,500 years old. @Partisangirl can't sing even one Assyrian song.

10.14.2020 1:38am - (Replying to @voidmstr and @newsycombinator) I prefer realizable dreams to virtual reality

10.13.2020 11:59pm - As we are trying to convince economists to listen to commonsense, those who speak Portuguese could do it in their mother tongue, in this podcast with Carlos Cinelli. I have a strange feeling that the future of econometrics is moving towards Brazil.

10.13.2020 8:32pm - (Replying to @Jabaluck) We are getting close to the breakthrough. What methods are available to 21st century scientists that allow them to go from assumptions about unobservables and "develop tests" on observables of whether the assumptions hold? Look how hard it could be:

10.13.2020 8:15pm - (Replying to @Jabaluck) My question is: Say something meaningful about the testability of your assumptions, without referring to me. Can you?

10.13.2020 8:08pm - (Replying to @Jabaluck) You listed 3 assumptions in your paper. Presumably you made them because 1)You needed them to get the answer. 2) You believed they were reasonable. Forgetting about what I propose, what does a red blooded economist say at this point about testability? We heard they care about it.

10.13.2020 7:55pm - (Replying to @Jabaluck) Not to me, who happened to know nothing about insurance. I see 3 symbolic assumptions in your paper, but no name given to the method through which you deem your assumptions testable or untestable. Does it have a name? Can you share the method with us? It would be a "breakthrough"

10.13.2020 7:41pm - (Replying to @Jabaluck) Since "directed graphs" are features of all structural equation, and since d-separation hold in cyclic graphs, economists will be missing these two spaces. I am missing only non-linear cyclic structural equations, for which, ttbomk, no one has a systematic logic of testability.

10.13.2020 7:10pm - "Why does BBC promote my daughter's murderer?" asks my dear friend and comrade Arnold Roth: Anyone there #bbcnews have any clue? or here @BBCWorld ? or @BBC ? or @BBCBreaking ?

10.13.2020 6:20pm - (Replying to @haig) Could be, but I have heard the phrase "the problem of causality" used again and again, mostly among philosophers, and I wish they would organize a conference on "Is there still a 'problem of causality' ?"

10.13.2020 5:57pm - No, I honestly do not know what "the problem of causality" is and what generations of philosophers meant by it, or what they would consider to be its "solution". An intriguing question: Could they be convinced that we have such a solution today? I wish I could try.

10.13.2020 5:00pm - (Replying to @DKedmey @Econ_Marshall and 2 others) An "economist that embraces new formal tools" will have to fight an uphill battle against a tyrannical conservative establishment. He/she will win at the end, at some price, but with history's gratitude. HMM, how about 2025 Nobel?

10.13.2020 4:24pm - (Replying to @agpatriota and @LucianoCSilva) We can't "solve" a problem unless we define it first, causality included. Curious: what IS "the problem of causality"???

10.13.2020 4:20pm - Pearl said economists are extremely concerned about testability, singing to it chapter and verse, relentlessly. He also added that, given this song, their rejection of an eyeglass that detects testable implications in 90% of their models constitutes a new definition of "narrow".

10.13.2020 3:33pm - (Replying to @agpatriota) It is still "reality" that made you add that arrow. Reality brought about your belief about the causal connection. Then this, together with your commitment to make your model reflect your beliefs left you no choice but to add that arrow.

10.13.2020 6:01am - (Replying to @dynamite_ai and @agpatriota) Good point. Except that we do not "Know" reality, only some of her features (eg symptoms do not cause diseases) and we are leveraging data to help quantify things we wish to know (eg. causal effects). It is not "inability" to do statistics, but a need to do what statistics can't.

10.13.2020 5:45am - (Replying to @valeritweety) Dying to hear more about this miraculous transition. Looking back, what do you think your professors should have known and didn't? What kind of questions they left unanswered in their students minds?

10.13.2020 2:56am - (Replying to @elikesprogramm1) No way! He who can't walk around the house can't run around the block. This becomes immediately clear to anyone who tries to solve a problem both ways. Unfortunately only DAGs folks can do it, and that does not leave us many PO enthusiasts. Try it! Dont be a bystander

10.12.2020 11:39pm - (Replying to @Jabaluck) I am looking back at my tweets and can't find one "overclaim" or less than "correct". I must conclude that you are using "Trumpian" tricks of discrediting those who ask you simple concrete questions: e.g.,"How do you tell if your model is testable?" Why do you have to do it?

10.12.2020 11:00pm - (Replying to @nyarlathotepesq and @Jabaluck) "economics classic", "parlor game", "sophomoric", "textbook" are common expressions I am hearing again and again, but no solution to my question: "Take your model, which confounder can you add without spoiling identifiability?" Same with "endogeneity". "Classic" with no solution.

10.12.2020 10:50pm - (Replying to @agpatriota and @eliasbareinboim) Relations among random variables are "statistical models" and they occupy just one Rung in the Ladder of Causation. (see #Bookofwhy) . Terms like "causal effects" used to be controversial until 3-4 decades ago. Luckily we are in the age of causation: see

10.12.2020 10:41pm - (Replying to @rodakker @FelixHill84 and @DaniloJRezende) It's an interesting question but I would not call it "The" question. The beliefs we possess are product of long experience with reality, some direct and some indirect, conveyed through education, culture and perhaps even our genes. Does the aspirin really care?

10.12.2020 10:10pm - (Replying to @Jabaluck) How do you know that this question (about consumers knowledge) is not answerable within the confines of graphical models unless we first cast them in such format and, using theory, decide that the answer necessitates parametric models, or countefactuals or other assumptions?

10.12.2020 9:57pm - To "verify it" means to spot what testable implications the theory has, and test them. This is precisely what economists resist; they refuse to teach their students how to glance at a model and see what testable implications it has. Strange reality, but verified again and again.

10.12.2020 9:42pm - This is precisely how economic education is seen from the outside: Econ "already know" the models they want to work with, because they can solve them - the rest is "peripheral". Any slight change (eg. new observation or confounder) and they are totally lost. No theory of repair

10.12.2020 9:29pm - Replying to @agpatriota and @eliasbareinboim Was the aspirin--->headache clear enough example of how reality imposes an arrow on your model?

10.12.2020 9:22pm - (Replying to @FelixHill84 and @DaniloJRezende) If you strongly believe that, in reality, aspirin affects headache, then this reality OBLIGES you to include an arrow Aspirin---->Headache in your model. You are not at liberty to remove this arrow for the sake of identification or publication or cultural loyalty. #CausalTweet

10.12.2020 9:07pm - If economists were fluent with DAGs they would be able to tell which arrows can be added without sacrificing identification, and which ones are crucial, not to be crossed out unless supported by solid scientific arguments or confirming data.

10.12.2020 8:56pm - Important question. "Reality" is what gives you a stomach ache when someone tells you "here is my model of reality". Reality is what makes you scream: "How can you justify this or that assumption?". It is a set of assumptions about Nature, you are not willing to give up.

10.12.2020 7:27pm - Models are not there to be cherry picked by analysists; they are imposed by reality, regardless if the analyst "care about" them or not. So, given ANY model imposed by reality, can you tell us which of its parameters is estimable, and whether it is falsifiable? #causaltwitterModels are not there to be cherry picked by analysists; they are imposed by reality, regardless if the analyst "care about" them or not. So, given ANY model imposed by reality, can you tell us which of its parameters is estimable, and whether it is falsifiable? #causaltwitter

10.12.2020 6:58pm - (Replying to @Jabaluck and @autoregress) "I do it in all my papers [link..]," like "economists do it all the time [link to Haavelmo]" is back to nowhere. Concreteness behooves us to name the steps, going from a model description to a decision: Yes, parameter theta is estimable if the model have this or that properties.

10.12.2020 6:45pm - If "useful" means "it can be credibly mapped to an important real-world problem" then, surely, a method that takes such real-world map and tells you which parameter is estimable, or whether it is falsifiable, should be deemed of highest priority, and appear in every eco textbook.

10.12.2020 6:17pm - (Replying to @autoregress) Is there any doubt in your mind that deciding whether one's model is falsifiable would be helpful to a practicing economist? Any doubt that arithmetic comes before building bridges?

10.12.2020 6:09pm - By all means: "Let the marketplace of ideas decide!" It does not hurt, however, to pause and take tally once in a while, and remind Econ. students: "Look how far behind Econ is." No push, No rush, just look for yourself.

10.12.2020 5:49pm - "Most important insights are already known to economists"- I have been hearing this since I met Heckman (1995). I am still searching for an Econ-PhD who could answer: "In your own model, (1) which parameters can be estimated by OLS? (2) Is the model falsifiable? #CausalTweet

10.12.2020 4:31pm - There is no symmetry here. Toy problems have agreed upon solutions. Complex real-life problems, remains complex, disputable, and incomplete, with no objective measure of "solution". So, which should come first?

10.12.2020 2:25am - (Replying to @mattmcd) The wikipedia articles are wanting, for they talk exclusively about "parameter identification". Practical economists are interested in policies, not parameters, so they ask whether the effect of a given policy is identifiable. #CausalTweeter

10.12.2020 2:13am - I have listed a dozen or so toy problems here:, and challenged economists friends to solve them. Here are two: Given a linear recursive structural model, (1) which of its parameters can be estimated by OLS? (2) Is the model falsifiable? #CausalTweeter

10.12.2020 12:39am - Poetry in motion. This is where I've learned to swim. The waves have not changed a bit - I did.

10.12.2020 12:28am - As economists are anxious to hear who the 2020 Nobel Prize winner is, I venture to re-tweet my earlier proposal: The Nobel committee should suspend the Prize until such time when 50% of econ PhDs understand indentifiability and testability of (simple) econ models.#CausalTwitter

10.11.2020 12:43pm - (Replying to @sdaPCA and @rogerwaters) Racism is not a "political view" but a "moral philosophy". If there's one thing that throws a Zionophobe off balance, it is proving him racist within his own moral philosophy. Roger Waters thinks that calling others "racists" is all it takes to stop being one.

10.11.2020 9:22am - Nice exposition! One comment: The v-structure test is a Theorem not a definition. So "Two causal model DAGs are said to belong to the same equivalence class.." should read: "It can be shown that two DAGs belong to the same equivalence class ..." #CausalTwitter

10.11.2020 4:27am - An intriguing and unknown story. Wondering: Does Roger Waters have any grandchildren? @rogerwaters ??

10.11.2020 4:12am - (Replying to @EyalShay2) If @MarkRuffalo is anything close to the sincere image he was trying to portray on @mehdirhasan show, he would take you on your offer. I wish I could join you two on that spiritual-scientific trip.

10.11.2020 3:12am - And Roger Waters will continue to prove to the world that, given colorful rhetoric, even a chronic racist could pass for a Nelson Mandela.

10.11.2020 11:22am - (Replying to @sdaPCA and @rogerwaters) Racists come in many shades, Zionophobia is one of the worse, and the question is whether his grand-children will some day reclaim the family honor.

10.10.2020 10:08pm - And no birthday of Danny can end without music. So thank you Todd Mack for continuing the tradition of Daniel Pearl World Music Days with your unique "Music in Common" initiative. Watch and listen to the Youtube clip below:

10.10.2020 9:07pm - It's impossible to end this birthday without one of Danny's jokes. From Eritrea (~1997) he wrote to us: "Hey mom! Today I have discovered my African roots. So many people are named "Daniel", so many have a Jewish nose, and so many are upset-minded. Bye, Danny."

10.10.2020 10:01pm - We miss you, Danny

10.10.2020 7:05pm - Note however that all the principled methods we have for fusing data rely on some assumptions about the data-generating process, namely, on properties of the DAG. Once you take this away, all we have are good people expressing needs, but no principles of data combination.

10.10.2020 3:53pm - A greatly needed initiative!! But, given my limitations, I would only be able to appreciate the results if the principles of combination are explained on a toy example, so we can see with our own eyes what is generalizable across problem instances.

10.10.2020 3:42pm - Thank you Raelle, and all who join us in rememberance on this virtual birthday of our son Danny. I wish I could believe that he is working overtime up there, trying to repair the world, or whatever's left of it.

10.10.2020 3:33pm - How true! I must add though that critics of specific assumptions started only when graphical models made the assumptions vivid. Prior to that, no one could tell what the assumptions mean, so they were welcome merely by licensing favorable estimation routines. #CausalTwitter

10.10.2020 7:02am - Most importantly, the Kurds, in sharp contrast, recognize their neighbors' right to equally legitimate states.

10.10.2020 6:27am - (Replying to @osazuwa and @eliasbareinboim) In general it will depend on x. I once computed such a limit (or related), probably here, on the mediation formula in the continuous case:

10.10.2020 2:05am - The magic words, that Columbia's President Lee Bollinger has still not shared with the campus. Will he? Or will he appoint another "task force" instead?

10.10.2020 1:37am - An unfashionable, yet urgent call from JFK - Justice For Kurds - and its full page ad in the NYT.

10.10.2020 1:07am - We need a theory of RWE to understand how it can compliment clinical trials. We still do not know how to combine cheap observational studies with sparse RCT findings, even when we have a believable causal model. We need to develop the scientific basis of RWE. #CausalTweeter

10.10.2020 12:52am - (Replying to @JoshWashIBSI and @MarkRuffalo) The real ignoranimus is @mehdirhasan . Watch his set up: "You are speaking about UNFASHIONABLE issues like Palestine, How come?". Unfashionable? Come on...Have you been to any campus lately? The ignorant wouldn't touch an issue if it were not fashionable.

10.9.2020 5:30am - (Replying to @daniel_bilar @MiriamElman and 2 others) Beg to differ. If you knew what it takes to get 10 Jewish professors out of their closets, overcome real and imaginary fears, and say in public: Enough with BDS, you would consider 75 signatories to be a monumental achievement. Next campus will see 200 signatures.

10.9.2020 5:01am - (Replying to @AkberKhan) I would never give BDS the benefit of being labeled "anti-semitic" thus minimizing their more significant contribution to world's racism - #Zionophobia. Here is why:

10.9.2020 4:39am - My second blessing goes to our academic colleagues at Columbia University. Let's hope this positive momentum does not end up like so many others, in a "task force" or a "training program" led by scholars who can't spell "Zionophobia".

10.8.2020 8:54pm - (Replying to @EvanCull @brentdg2 and @elderofziyon) A "binational state" experiment requires even more trust-building efforts than "2 states", hence the "equally indigenous" utterance becomes even more essential and pre-requisite. Yet Saib Erekat cannot not say it, and every antenna in Israel is tuned to Jericho.

10.8.2020 8:02pm - There is another point which perhaps has not been acknowledged in @chipro ML account of "data shifts". You can't see the difference between "data shifts" and "model shifts" if all you have is data. It is only when you formulate a model (eg, DAG) that you can see its importance.

10.8.2020 7:48pm - Faculty involvement is crucial, and I bet the petition will garner 150 signatures. However, Bollinger is still not being asked to do the one thing that would cure Columbia from its madness: Tell the campus, honestly, what makes BDS racist, as President Polack did, at Cornell.

10.8.2020 7:22pm - (Replying to @EvanCull @brentdg2 and @elderofziyon) "Political representation", even "statehood", was given many times, eg. 1947, before Beinart was born, and we remember what happened. So lets try an alternative, much less risky. Let one (just ONE) Palestinian leader utter the words "equally indigenous" and see what happens.

10.8.2020 6:38pm - (Replying to @EvanCull @brentdg2 and @elderofziyon) "2 states for 2 peoples" is a big lie when separated from the second part of the Zionist credo: "Equally legitimate and equally indigenous!". You will never become a good Zionist unless you can pronounce this part. Can you? Try! Our neighbors couldn't. Never tried.

10.8.2020 6:24pm - The ropes behind the "data drift" vs. "model drift" distinction is not just a metaphor. Simple examples show that two different "model drifts", leading to SAME "data drift" may require two DIFFERENT adjustments for adaptation. See - diagrams make it clear.

10.8.2020 6:11pm - (Replying to @EvanCull @brentdg2 and @elderofziyon) This is what Zionism is all about "2 states for 2 peoples, equally legitimate and equally indigenous" -- Welcome to the Zionist tent!!!

10.8.2020 12:51pm - (Replying to @brentdg2 and @elderofziyon) Do you think Beinhart believes for a moment that "peace" on his term is even remotely possible? Namely that Israelis would give up sovereignty short of loosing a genocidal war? There's nothing "earnest" about an irresponsible youngster playing cute with lives of millions.

10.8.2020 5:04am - (Replying to @elderofziyon) On Dec. 9 1666, the Rabbis of Constantinople excommunicated Shabbetai Zvi, a self-proclaimed messiah who led a sect of Jewish followers to act against the core values of the greater Jewish community. The dangers currently posed by Beinart are probably greater than those of Zvi.

10.8.2020 1:48am - The distinct features of such drifts cannot be described, diagnosed, or cured, in the language of probabilities, which are just surface phenomena. Drifts in the ropes behind the data require models of those ropes, as is done e.g., here:

10.7.2020 5:23pm - To celebrate the launching of #CausalTweet I am sharing a new post: - "A collage in the art of causal reasoning," based the wisdom of compiling ideas under the same umbrella: Causal, Casual and Curious.

10.7.2020 5:08pm - (Replying to @PhilPapag @gokhanyu and @Grady_Booch) Eratosthenes would not have invested that much energy in measuring the radius of the earth (240 BC) had he taken the spherical shape of the planet to be merely a "fruitful metaphor" instead of "metaphysical true".

10.7.2020 4:59pm - I am urging colleagues and friends at Columbia University to sign this petition to President Bollinger He tries his best to curb antisemitism on campus but stops short of addressing the root problem -Zionophobia - to restore civility on moral grounds.

10.7.2020 3:44pm - This blue sky dream is what keeps my burning with excitement when I am not Twitting. It is lightly sketched in the last chapter of #Bookofwhy, and will I hope come to fruition by some followers of @CausalTwitter.

10.7.2020 4:07am - Why is this important? Because the degree of effect modification, as well as z-specific effects, can be estimated from observational studies only when the latter when identification is certified by do-calculus.

10.7.2020 3:36pm - (Replying to @Scorpil) Locky I am, to still be naïve enough to view with breathless amazement AI's achievements in the last 3 decades, at least those with whom I am somewhat familiar. #CausalTwitter

10.7.2020 2:58am - (Replying to @DavidAOliverJr and @TWilliamson55) Strange, and I missed it completely. I bet Causality would have turned out differently if I hadn't. Is it too late?

10.7.2020 1:27am - (Replying to @NotTriggerAtAll) Both are true. DL's objective is "strong ai" but, if it continues along its data-centric mindset, the limitations of the Ladder of Causation will soon surface to halt its progress.

10.7.2020 1:15am - (Replying to @ShorrTirza) The algorithmization of PRIDE should be part of the Turing test for consciousness. I'm working on it.

10.7.2020 12:44am - The argument was not "there is plenty of space" (that was a fact), it was "equally indigenous on historical grounds." You won't find this argument in any of your classic texts of colonialism, hence you can proclaim your classics irrelevant and the word "colonialism"- a fiction.

10.7.2020 12:15am - (Replying to @gokhanyu and @Grady_Booch) The computationalistic paradigm is mighty productive, the anti-computational paradigm is barren.

10.7.2020 12:11am - (Replying to @AndrejSpiridon4) The trouble with the "what-computers-can't-do" folks is that they won't change their mind even seeing a computer discover a new mathematical field, tell funny jokes or run for President. They would keep on saying: "this is not true consciousness, because...[slippery phrase]..."

10.6.2020 11:43pm - (Replying to @gokhanyu) The sentence is unfalsifiable without defining consciousness or, at the very least, proposing a Turing test for consciousness.

10.6.2020 11:12pm - Another winner of the 2020 Nobel is Roger Penrose who is remembered in AI circles by his 1989 book "The Emperor's New Mind", arguing for the impossibility of strong AI. I wonder if he changed his mind in light of new advances in AI, especially deep learning and causal reasoning?

10.6.2020 10:47pm - Congratulations to my UCLA colleague, Andrea Ghez, for winning the 2020 Nobel Prize in Physics. Though I have not had the chance of meeting her in person, the fact that our offices are just 300 meters apart makes me feel mighty proud.

10.6.2020 10:37pm - Moreover, the Arab leadership in 1920-30 admitted that the country could absorb 5-10 millions new immigrants without dispossessing a single Arab peasant. See

10.6.2020 10:24pm - Formally, the question: "How well is my problem formulated?" means "how close is my model to reality?" and, since the model includes causal assumptions, the question cannot even be asked in the language of statistics. The ladder is a theorem, not an artistic metaphor.

10.6.2020 6:40pm - (Replying to @pentagoniac @CasualBrady and 9 others) What? Trustworthy-AI work at UCLA? It must be in the Department of Speculative Philosophy!

10.6.2020 3:37pm - (Replying to @andre_bida) "Is there a statistical test to quantify how well a problem is formulated?" You are too timid in saying: " As far I know, there is none. Yet." I would be honest and say: "We can prove mathematically that no such test exists". Why shy away from mathematics? Is it "cancelled"?

10.6.2020 12:45am - (Replying to @RWJE_BA) Did you mean "IV is a poor tool" or "IV is a good tool"? If the latter, it does not negate what I said, because there is no statistical test for instrumentality (save for bounds); to determine if a variable is a good IV we must make non-statistical assumptions, eg. exclusion.

10.5.2020 6:14pm - I can't get over the expression on AOC's face; how lovingly accepting, revering, respectful, approving and curious she is of a couple of Zionophobic nuts, dressed in Kaffias, praying for the end of Israel, with whom she hopes to fool Jewish voters. Motherly love, heavenly match.

10.5.2020 4:58am - (Replying to @ashkephardi and @Claire_Voltaire) Love it! @ashkephardi . My next oped: "How to become a Jewish leader?", then comes: "How to anoint Jewish leaders". Love it. p> 0.5.2020 4:39am - Replying to @andre_bida) o idea. I do not see a tread. p> 0.5.2020 4:19am - Replying to @HolgerSteinmetz) In which case it is time to take "moderation" and do onto it what was done to "mediation" and "modification". Not that the SEM community would celebrate right away, but it should be done at some point, and now is the best time, while they'r still digesting "mediation." Volunteer?

10.5.2020 3:43am - (Replying to @HolgerSteinmetz) I agree, of course, that ALL relationships must be classified and defined as causal or statistical. What is not clear to me is whether "moderation" has an agreed-upon definition, or is still at the mercy of traditional confusions that the SEM community has been suffering from.

10.5.2020 2:44am - The discussion about "effect modifiers" should benefit from an important observation: Identifying "effect modification" is an exercise in do-calculus, hence it can be done algorithmically. This is missing from Epi literature which, ruled by Harvard, lumps Rungs 2 & 3 together.

10.5.2020 2:30am - (Replying to @HolgerSteinmetz) In Epidemiology, Tyler VanderWheel's distinction between "effect modification" and "interaction" became standard in the field. Is "moderation" formally defined in Psychology? Last I checked it was still lingering in linear SEM tradition, unable to escape.

10.5.2020 2:13am - I don't know what qualifies me to be on the email list of Quora but, here, it is yr 2020, and I get this interesting question: What is the statistical test to establish that event B is caused by event A? I had to disappoint the readers: There isn't any.

10.5.2020 12:22am - Stimulated by epidemiologists fascination with "effect modifiers", I recalled an orphaned paper: which, having appeared in Social Methodology, skipped the attention of Epi researchers. Note the simple examples of identified effect modification in Figure 1

10.4.2020 6:36pm - What aspect of your experience at USC do you find most troubling? Would the new "Zionism Reclaimed" movement, alleviate those problems?

10.4.2020 6:18pm - It is always gratifying to see DAGs continue their penetration into epidemiology research I would only substitute the timid words "DAGs formalize concepts","illustrate exceptions","clarify phenomenon", "examine problems", with: "provide a solution to..."

10.4.2020 5:46pm - For music lovers everywhere, the Daniel Pearl Memorial Concert will be starting soon, Sunday, 7:30 pm, SPT. See you here:

10.4.2020 12:45am - Highly illuminating talk. Gratified by the connection you found between this work and causality. But I couldn't quite understand what you meant by "unit testing", what are the "units" in CI, and what are we "testing"?

10.4.2020 10:53am - (Replying to @soboleffspaces) How can "Probability and Conditional Expectation" capture the Empirical Sciences when they are stuck on Rung-1 of the Ladder? My ladder says: impossible.

10.3.2020 12:06am - (Replying to @paoladm) Identities are not defined by dates of names, nor by genetic lineage, but by historical continuity of narratives. If you show me a collective singing Hittite songs, or celebrating Hurian holidays, I would say you are lucid. See what I asked Saib Erakat:

10.2.2020 9:04pm - Please join me in saluting & supporting Zionist professors at USC, who are demanding an explicit recognition as bonafide members of the campus community. TTBOMK, such public demand is the FIRST in the history of Western Civilization. Interesting times.

10.2.2020 6:29pm - (Replying to @SheridanLGrant and @daniela_witten) Well put, but watch out. Saying "don't teach Simpson's "paradox" without causality" amounts to advising 95% of statistics professors to go against their textbooks -- a dangerous, if not fatal move.

10.2.2020 3:35am - (1/ ) An important theorem about propensity scores (PS) that I have not seen PS analysts acknowledge, goes: Let ATE(C) be the treatment effect estimate obtained by PS with covariate set C, and let ATD'(C) be that obtained by adjustment on the same set C. Then, as the sample size
10.2.2020 3:36am - (2/ ) (Replying to @yudapearl) increases, ATE(C) => ATE'(C) for all C, regardless of whether ATE(C) is biased or not. The proof is here:, which is built on the beautiful theory of "confounding equivalence" (Pearl and Paz 2014) Recommended for beauty lovers.6

10.2.2020 3:12am - This year, you need not visit Stanford to attend the Daniel Pearl Memorial Concert; it takes place on line, on Sunday, 7:30 pm, as choreographed here:

10.2.2020 12:10am - (Replying to @RaulMachadoG)

10.2.2020 12:07am - I have no objections to propensity scores; I object to using them blindly assuming that they would help you reduce bias regardless of what covariates you measure. They won't.

10.1.2020 11:59pm - (Replying to @soboleffspaces) Can you point me to a probability-theoretical book that uses "model" in that sense? This would help me convince statisticians that they ought to know what "models" are.

10.1.2020 11:50pm - I heard about do-calculus, I know what it is, and I still like your short video very much; highly recommended to all students of CI. I would only augment your course with a problem that cannot be solved without do-calculus, eg. the 'new napkin problem" on page 240 of #Bookofwhy.

10.1.2020 8:15pm - Is the legal settlement reached by NYU a victory for civil rights? Yes and No! See my reservations here: Victories achieved by legal means are never true victories; too bad university administrators neglect their duty of fighting it in the moral arena.

10.1.2020 1:23am - Bravo @GavinNewsom for vetoing the California Ethnic Studies bill. I have two grand-kids going to California schools, and I got sick to my stomach reading how their education was about to be hijacked by "ethnic studies" exclusivists.

9.30.2020 1:58pm - (Replying to @jouni_helske) Sure! If you have hidden variables, everything is possible. For example, you can assign the answer A to your causal query as a hidden variable and, of course, the model has the capacity of answering your query. (Or, for that matter, the size of a Unicorn, assuming Identifiability

9.30.2020 1:48pm - (Replying to @aztezcan and @learnfromerror) The litmus test is simple: Can you construct the model if I give you the joint distribution of all observed variables? If yes, then it is not a causal model, because it does not convey causal assumptions, hence it cannot answer any causal question. A fundamental conservation law.

9.30.2020 1:43pm - (Replying to @widemberg and @noah_greifer) "Nothing to do with causality" means that PS can be constructed from the joint distribution, no causal assumptions are needed to define it. The causal assumption of "strong ignorability" is imposed later, to get the Theorem going, but it is not part of the PS definition.

9.30.2020 1:38pm - (Replying to @RaulMachadoG) Society's decisions demand it, and I do not know of any viable alternative.

9.30.2020 4:59am - (Replying to @omaclaren and @tmorris_mrc) I think if you provide a different summary of statistics you would deserve credit for a revolutionary summary. I am not sure of its accuracy though.

9.30.2020 4:52am - (Replying to @Shockley_MRX @noah_greifer and @widemberg) I wouldn't advise Yahoo to assess their Ad-Affectiveness using this methodology. First, confounding is wished away by the assumption of "strong ignorability". Second, "affectiveness" is a Rung-3 notion, not equal to ATE (Rung-2). See

9.30.2020 3:12am - (Replying to @tmorris_mrc and @omaclaren) I am merely reflecting the definitions given by generations of leading statisticians, as well as practices and concepts and models discussed by 95 % of stat textbooks. Examine my list (Causality 2009) of leading statisticians commenting on causation, have I missed anyone?

9.30.2020 1:32am - (Replying to @omaclaren and @philipbstark) You lost me. I don't see any relations to "inverse problems" or "hypothesis testing". We are talking causality, and how statistics leaders, one by one, have proclaimed it to be outside the province of statistics proper, some formally, and some through calculated avoidance.

9.30.2020 1:14am - (Replying to @learnfromerror) I am a humble wayfarer, asking merely for ONE model, not altogether. Please point to ONE stat model that, in addition to describing data, also describes the "data generation mechanism it models"? ONE please, just ONE!

9.30.2020 1:01am - (Replying to @BlueManifold) And I yawn the same when told "logicians have been working on causality since Aristotle," and reply: "Can you algorithmitize what you learned from Aristotle so that a Robot would not try to meddle with the barometer when asked to improve the weather.?"

9.30.2020 12:55am - (Replying to @peder_isager @david_disabato and 2 others) For the record, "d" in d-connection stands for "directional", because dependence along paths containing directed arrows requires its own rules. In your description of colliders you should add that M is observed.

9.30.2020 12:45am - (Replying to @Shockley_MRX and @kareem_carr) All good points with the exception of one: what the data scientist intends to learn is not a specific PATTERN in the data, but a specific property of the data-generating model. That sought after property may or may not have a "pattern" in the data.

9.29.2020 11:20pm - (Replying to @spiderjens and @learnfromerror) My conversations with Spanos have confirmed my rule of thumb: To gauge a person's understanding of causality, test his/her understanding of Simpson's paradox. See Spanos on Simpson's

9.29.2020 11:02pm - I am not criticizing statistics for setting its boundaries narrowly. I am merely explicating how those boundaries were defined by its practices, textbooks and leaders. See quotes If you want to widen those boundaries, fine; but admit it is a post-hoc job.
9.29.2020 11:02pm - (Replying to @yudapearl) Moreover, I agree that behind every statistical model there stands a mental causal model begging to be expressed. After all, statisticians are humans and the human mind is driven by causal, not statistical logic. Still, the shadows of those mental models don't make up a science.

9.29.2020 10:21pm - There is a stronger KEY here: to separate the two NOT IN YOUR HEAD but in the SYNTAX, so that the separation is not left to the mercy of interpretations.

9.29.2020 10:17pm - I think you missed the key ingredient of CI assumptions. These assumptions CANNOT be EXPRESSED as properties of the distribution of the observed data. The test is in the the syntax. Researchers can debate reasons and intents, but the syntax reveals the assumptions' true essence.

9.29.2020 10:03pm - (Replying to @RichmanRonald and @learnfromerror) You r probably referring to the necessity of modeling the process of generating X. Rubin calls it "treatment assignment process" but, lacking graphs, students find it wanting.

9.29.2020 9:57pm - (Replying to @noah_greifer and @widemberg) Agree. And glad you said PS is a way to estimate a statistical expression whose appropriateness is approved by causal considerations. PS does not contribute to bias reduction if X is chosen inappropriately.

9.29.2020 9:37pm - (Replying to @learnfromerror) This is news to me. Can you give us an example of a commonly used stat model that, in addition to description of data, also describes the "data generation mechanism it models"? I have a shelf full of statistics textbooks in front of me. Which one should I open for such a model?

9.29.2020 8:25pm - I was surprised that my innocent statement on the distinction between statistical and causal model has generated such a lively discussion. On a second thought, it is natural. Some people are motivated by the comfort of past habits: "Don't worry, causal inference is nothing new,..
9.29.2020 8:25pm - (Replying to @yudapearl) "Statisticians have been doing it for centuries, relax!" Others are motivated by revolutions "Its a whole new logic, encountering a new set of challenges...". Don Rubin appealed to the former: "It's just statistical analysis on missing data problems -- we know how to handle it!"
9.29.2020 8:25pm - (Replying to @yudapearl) I have been motivated by the latter: "It requires a whole new algebra for new operators eg. do(x), and a whole new logic for new entities, eg. counterfactuals..." If you seek comfort in ancient wisdom then, Yes, it's all in Gauss and LaPlace and you have a license NOT to change
9.29.2020 8:25pm - (Replying to @yudapearl) habits of thoughts. But if you find excitement in seeing streams of new results pouring out of fresh paradigms, you would just yawn if someone tells you: "Statisticians have been doing it for centuries", and say: You are a statistician, tell me something about Simpson's paradox."

9.29.2020 7:36pm - (Replying to @omaclaren) Instead of equating them I am very clear on what "theories" are just constraints on distributions and what "theories" are more than that, for they give you answers to causal questions. I haven't seen any of the latter in the statistical literature till ..Wright?... Neyman?? ...

9.29.2020 6:57pm - (Replying to @omaclaren) But the distinction between statistical and non-statistical models is as clear as a baby smile, we don't need Theta and fancy mappings. The former is any statement(s) that can be expressed in terms of a distribution of observed variables. See

9.29.2020 5:44pm - (Replying to @swalkertweets) Some statisticians are getting it and some (hate to name) never will. But ask a statistician what Gray-Box Models are available to us, and that's where you would see the distinction between the enlightened and the rest.

9.29.2020 5:25pm - (Replying to @VC31415) From structural functions we can answer all causal questions, true. On the other hand, if in "causal questions" we include counterfactual questions, then the converse is also true: structural functions can be specified as a set of counterfactual statements. See Causality ch.7

9.29.2020 5:08pm - "Propensity score" is the source of one of the greatest confusion in our generation, second only to the word "approach". For "understanding propensity score" see here:

9.29.2020 4:06pm - Not really. What statisticians call a "Model" (eg, Gaussian, Bernoulli ...) is not a causal model but a description of the data. A statistical model cannot tell you if X causes Y or the other way around. Some statisticians (and all textbooks) still don't see the difference.

9.29.2020 3:46am - (Replying to @andy_l_jones) Gee, how little has changed in the past 23 years. One thing I was not aware of in 1997 was the tight grip that PO has had on some areas of social science, including Angrist type of econometrics. Otherwise, I agree with almost everything.

9.29.2020 1:02am - Not really. The whole point of this blog discussion was to convince you that, even if some day you will be able to infer causal models from data, your students will have to learn how to manage hybrid models, namely, snap out of the data-centric paradigm.

9.29.2020 12:22am - Good start! Statistics is a prerequisite to do causal inference, but is useless without it, at least for problems that require policy evaluation or scientific understanding.

9.29.2020 12:16am - (Replying to @alexp1799) Yes, that's the one. Highly recommended with no dog in the fight.

9.29.2020 12:10am - (Replying to @dovgvlad and @roydanroy) Well put!!

9.28.2020 11:52pm - That's exactly what I am predicting. ML will soon go hybrid and celebrate: "we swallowed causality". But that will only happen after ML students learn to solve at least some toy problems in Primer and, as an educator, this is my goal, not swallowing.

9.28.2020 8:34pm - Of course Data-driven ML wont be replaced. As Rung-1 in the Ladder, it is an important component in any Causal Inference Engine. It will transform, however, from a data-exclusive engine to a hybrid {data,model}-driven engine. As to far-fetched futures, see
9.28.2020 8:34pm - (Replying to @yudapearl) It says that, even if some day you will be able to infer causal models from data, those who are not trained to manage hybrid {data, model} engines will be out of a job. Take note. Because this is a significant chunk of ML folks.

9.28.2020 7:32pm - The spirit of Yom Kippur has ascended upon Pakistan and, evidently, reminded the Supreme Court that actions have consequences, consequences are recorded, and history is watching. Saluting the Court and our Lawyer. Time to break the fast.

9.27.2020 6:56pm - (Replying to @lorisdanto) I did, but our lawyer suggested against, for fear it would compromise the non-profit status of the Daniel Pearl Foundation.

9.27.2020 3:31pm - The sun is about to set on California, marking the beginning of Yom Kippur. It is about to rise in Pakistan, where the Supreme Court will rule on our appeal: May the Yom Kippur idea of accountability for all our choices guide the Court's deliberation.

9.27.2020 2:51pm - Read #Bookofwhy at your leisure while trying to solve ONE toy example from Primer, using your favorite ML method. The rest will follow organically, thanks to the laws of commonsense.

9.27.2020 1:32pm - ML will not be the same in 3-5 years, and ML folks who continue to follow the current data-centric paradigm will find themselves outdated, if not jobless. Take note.

9.25.2020 6:06am - (Replying to @learnfromerror) Enlightenment is decided by the lasting transformation they introduce to their respective fields and, until we know whether it is lasting or not, we go by the shiny eyes of readers when they say: "How come I was never told..."

9.25.2020 4:42am - Max Planck said "Science progresses from one funeral to another". In our days: "Science progresses from one enlightened editor to another."

9.24.2020 2:25pm - I discovered that the latest enlightenment in Statistics Education comes from the new Editor, Jeff Witmer. See his interview with Allan Rossman here The only thing science needs now is for such editors to take over econometrics and machine learning.

9.24.2020 6:06am - (Replying to @DanielHGill) That was before he was stationed in London and, funny, I never heard him regretting the decision.

9.24.2020 4:52am - But think about the mentality of those editors. Where have they acquired this gullibility to buy into the Palestinian narrative of saintly victimhood? Ans. Come to UCLA and examine the history department textbooks (eg Gelvin's).

9.24.2020 4:05am - There seems to be a genuine push in Journal of statistical Education to introduce Causal Inference as part of introductory statistics courses. These two papers explain why and how: This is an extremely brave and encouraging move,
9.24.2020 4:05am - (Replying to @yudapearl) which economists should take note of; if students take stat-101 before econ-101 they would not sit quiet seeing a graph-blind econ professor struggling to explain what non-confoundedness is all about. Same with ML classes; students will demand to see data-generation processes.

9.24.2020 2:29am - Thanks for a well-documented report on @HuffPost and it's persistent anti-Israel coverage. But we lose credibility when we label them "antisemitic", they are merely careless, gullible and ignorant journalists, managed by cynical, and calculated editors.

9.24.2020 2:17am - (Replying to @vassalos) "Sad" and "wrong" are emotional expressions, which I respect. But they cannot replace reasons and principles.

9.24.2020 12:24am - When our son, Danny, was offered a job at the NYT and decided to stay at the WSJ, we jumped on him with:"Are u crazy?" His answer: "It's a better paper, and I am a reporter." Reading Lance Morrow (below) I understand what Danny meant. He was a reporter, not a "journalist."

9.23.2020 9:42pm - Thank you @HeleneSolomon for giving me a chance to tell all readers, Jews and non-Jews, how valuable and unique the book "I am Jewish" is. I don't know of another moment of history where 150 Jewish thinkers would stopped their busy schedule and try to answer a simple question:
9.23.2020 9:42pm - (Replying to @yudapearl) "When you say: 'I am Jewish', what do you really mean?". And where else can you get a panoramic view of how 21st-Century Jews, from all walks of life, define themselves? I especially like the essays by Shimon Perez, Amos Oz, Theodor Bickel and Daniel Kahaneman. Unique!

9.23.2020 5:35pm - A new oped I just wrote about the late supreme court Justice Ruth Bader-Ginsburg, her sense of justice and her sense of being Jewish:

9.23.2020 12:20am - As a victim of the anti-human terrorism symbolized by Leila Khaled, I salute Zoom for their decision to refuse her "teaching" platform at SFSU. The former President of SFSU was warned that hiring Abdul Haddi means granting an academic license to terrorist "teaching"-- he ignored

9.22.2020 10:55pm - (Replying to @analisereal @BetsyOgburn and 2 others) And whenever they are not present, their absence represents the assumption needed for point identification. No identification without causal assumptions.

9.22.2020 7:13pm - (Replying to @DieterCastel and @blair_fix) Nice illustration, agree, but I dont buy the conclusion that, if we have interacting causes it makes no sense to ask what the contribution is of each. Striking a match contributes more to the fire than the presence of Oxygen, as proven (not "argued") here

9.22.2020 6:39pm - (Replying to @goldstein_aa and @HarryDCrane) If the question is to me, then, Yes. I ended up reading Primer and, modesty aside, will recommend it again as the BEST introduction to causal inference that I have seen. Just judge for yourself: Any other book with questions like Examples 4.4.1-4.4.5 ?

9.22.2020 2:02pm - (Replying to @Undercoverhist) Misunderstanding. By "I endorse the endorsement" I was referring to MY endorsement of the book. I would also endorse your "appreciation" if I only knew which part of my endorsement you do not endorse. (It's getting too syntactically complicated, shall we use DAGs to unwind?)

9.22.2020 1:56pm - (Replying to @Undercoverhist) I love tompepinsky's last sentence "once students have a firm command of instrumental variables, DAGs are helpful for representing the assumptions implied by such designs" IOW: "first make assumptions, call them "design", then ask yourself if you understand what they mean." HMM

9.22.2020 12:15pm - I endorse this endorsement which, I hope, will not get you in trouble with the guardians of econometrics education.

9.22.2020 2:41am - Excellent essay by Hussain Aboubkr, with one disturbing after thought. The Palestinian mystique that once mesmerized the Arab mind seems to be gaining a grip on European minds and US campuses. E.g., only one EU country sent representative to the WH peace ceremony. Inconceivable!

9.21.2020 4:19pm - (Replying to @Yair_Rosenberg) I put the two together for your followers:

9.21.2020 3:56pm - (Replying to @Yair_Rosenberg) You can add this to your collection: An essay she wrote for the book "I am Jewish"

9.21.2020 1:59pm - (Replying to @CasualBrady and @analisereal) Sure. How else is Y(t) defined? It is certainly not a property of probability distribution. So what is it? A new variable? So, how do we compute its value once we know everything about the world? Rubin & Co. evade this question by appealing to intuition about RCT. half-cheating.

9.21.2020 1:36pm - This is precisely what the great historian Thucydides did in 426 BC: "The cause, in my opinion, of this phenomenon must be sought in the earthquake". It requires however that we reason from effects to causes, as outlined here: or

9.21.2020 1:05pm - I like your phrase: "we can read off ignorability from our models, rather than just assuming it", which tells PO folks what causal models can do for them. However, to ordinary folks I would say: "we can answer causal questions directly, skipping ignorability jargon altogether."

9.21.2020 4:28am - (Replying to @tmorris_mrc) In SCM formulation, where you start with scientific knowledge, counterfactuals are DERIVED quantities, and consistency is a THEOREM. In PO framework, where you start with counterfactuals as primitives, consistency must be treated as an "assumption". See

9.21.2020 3:58am - A new mediation paper has crossed my screen: I was happy to see a parametric mediation formula for the direct and indirect effects on a odds ratio scale. Unhappy to see PO folks still referring to consistency as an "assumption" instead of "property."

9.21.2020 2:09am - (Replying to @offaltube @Rezbab and @Claire_Voltaire) What's the point? That antizionism is only bad in as far as it is tainted with anti-semitism? That it is otherwise OK? No way! Anti-semites are victims of ignorance and irrational obsessions, antizionists know exactly what they are doing and calculating, and it's very very ugly.

9.21.2020 12:29am - Holy mackerel!! Our paper on "Generalizing Experimental Results..." was accepted to "European Journal of Epidemiology" Our anonymous reviewers asked so many questions, that the paper now goes beyond "generalization" and "identification". Enjoy the beyond.

9.20.2020 11:33pm - (Replying to @yiannis_entropy) I read someplace that she was not an "observant Jew", like the biblical Ruth, grandmother of King David, who defined Jewishness through her famous saying: "Your people are my people, and your God my God", namely, people-hood first, religion second.

9.20.2020 9:13pm - Correcting the link to RBG's essay: She will also be remembered by the tree words inscribed in her office: Tsedek, tsedek, tirdof (Justice, Justice thou shalt thou pursue).

9.20.2020 8:50pm - (Replying to @LukeRodriguez75) Thanks, I will correct.

9.20.2020 8:36pm - I can now post Ruth Bader-Ginsburg's contribution to the Book "I am Jewish". May she be remembered by the people she touched, the ideas she affirmed and the values she pursued.

9.20.2020 1:25pm - (Replying to @MonsieurRicardo and @Eglencross) Yes, it will be posted as soon as I get the permissions.

9.19.2020 11:16pm - (Replying to @jaketapper @irin and 3 others) I'll be watching. And, @jaketapper , take a quick look at what RGB says here:

9.19.2020 11:00pm - (Replying to @StephReflects) And stay tuned to reading how she defined it:

9.19.2020 1:00pm - (Replying to @analisereal @CasualBrady and @diogo_pernes) Nice! We can even remove one of the 2 arrows from Z, and still have the same phenomenon displayed. I once used this example on a PO expert and asked if ignorability holds. Will spare you the rest. But try it on your professor.

9.19.2020 4:01am - (Replying to @diogo_pernes and @CasualBrady) X<-->W stands for X<--C-->W with C an unobserved variable. Some people call the mixture of directed and bi-directed edges "mixed model". But there is no need to proliferate terms when a bidirected edge is just a couple of directed edges and the middle node unobserved.

9.19.2020 1:13am - Supreme Court Justice Ruth Bader Ginburg is dead, at 87. When we asked her in 2003 to contribute an essay to the book "I am Jewish: Personal Reflections Inspired by the Words of Daniel Pearl", she did not hesitate for a moment, and sent us a gem of an essay, explaining precisely
9.19.2020 1:13am - what she means when she says "I am Jewish". I promised her that she will be remembered by that gem. Now that she is no longer with us, I will fulfill my promise by posting her essay for the general public. Stay tuned.

9.18.2020 3:59am - Shana Tovah! As we begin 5781 of the Hebrew calendar I wish all of you, readers, followers, discussants a HAPPY NEW YEAR. May 5781 be a year of peace and reconciliation, a year of improved understanding, both politically and scientifically -- a year of fulfillment of our dreams.

9.17.2020 3:29am - The distinction is crisp. Anti-Semitism = dislike of Jews. Zionophobia = Denying Jews the right to a sovereign homeland, regardless of whether you love them or hate them.

9.17.2020 11:06am - (Replying to @_Srijit) That would exonerate you from charges of discrimination, and genocide, but will qualify you as a denier of Palestinians rights -- an unforgivable cardinal sin.

9.17.2020 4:13pm - (Replying to @desai_pratik) I am trying, but can't understand the puzzle or the question.

9.16.2020 6:55pm - We should not waste time arguing which evil is greater, anti-Semitism or Zionophobia. What these 43 USC professors are waiting to hear from President Folt is whether they are welcome on USC campus or not. She wont say; scared to offend the Zionophobes.

9.16.2020 2:17pm - (Replying to @m_clem @analisereal and 3 others) Bias amplification is a peculiar phenomena, discovered by structural economists, then neglected by experimental economists. Intuitive explanation is given here: and here:

9.16.2020 6:44pm - (Replying to @m_clem @analisereal and 3 others) Thanks Michael for making me feel useful. My appearance in this thread was triggered by sensing your genuine interest in what we can learn from graphical models. There is probably a lot more to learn, so, join us in the fun.

9.15.2020 1:40pm - Seeing your work in the "Teacher's Corner" of a journal is a great honor, but seeing it in the Journal of SEM is a triumph for the causal interpretation of SEM, ending two decades of denial and confusion: The educational values of this paper should
9.15.2020 1:40pm - cross over to economics, especially the section on person-specific effects (PSE) which is a key contribution of the paper. Readers ask: Given linearity, why shouldn't we identify PSE using the general formula of Eq.(11.29) (Causality p.391)? Ans. PSE does follow from Eq. (11.29)
9.15.2020 1:40pm - as in Example 11.7.1, except that the causal effect T is not easily identified in models with latent confounders. The beauty of this paper lies in leveraging the linearity assumption to identify T despite those latent confounders. Crucial clarification: a "person" is (X,Y)-vector

9.15.2020 11:26am - Translation please? I grew up with Popeye and Little Red Riding Hood.

9.14.2020 11:25pm - Breaking News: It's today. The wheels of justice will pause for an hour or two to listen breathlessly to what layers say, argue or decide and, then, continue their rusty on-off movement, to the amazing approval of humanity.

9.14.2020 11:00pm - If you want to understand what tomorrow's signing of the Israel-UAE-Bahrain peace agreement means, both politically and conceptually, read what progressive Muslim women have to say, like Dr. Qanta Ahmed here: She's actually visited Israel, UAE and Bahrain

9.14.2020 1:54pm - An online course that introduces Obesity and Soda on the 2nd week must be heading in the right direction. Good luck.

9.13.2020 1:51pm - There is another metaphor that I occasionally use to explain what graphical models are: Simulators. Suppose you are asked to generate iid samples from a given distribution. How would you go about it? Each of the many algorithms that can be used to do this generation has a graph
9.13.2020 1:51pm - structure that depicts the step-wise process used, and all questions about interventions and counterfactuals can be answered from the generating algorithm chosen. Note: no RCT required for any of this.

9.13.2020 1:39pm - (Replying to @EdhanOmer and @Advill54) Zionophobes have no questions. They just howl what they were taught to howl and, thus reveal to the world how ready their teachers-leaders are to make peace in the middle east.

9.13.2020 3:20am - (Replying to @desai_pratik) Believe me, in my age, 2018 seems "in my youth," because I almost forgot what I wrote there and, upon re-reading, it sounded so fresh and young; I wish I could write like that today. Perhaps I still can ...

9.12.2020 9:09pm - No. Academic positions, like political statements, are determined by the voter-base one is less likely to offend. Note how the manipulability debate is clustered around the name of one's PhD advisor. Lucky me, I don't have an advisor demanding loyalty.

9.12.2020 3:24pm - Your question on non-manipulable causes led me to re-read the two papers below, and made me proud of the fact that, in my youth, I had the talent to write such clear papers on a topic that is so susceptible to confusions, both intended and unintended. Highly recommended!!!

9.12.2020 2:10pm - My favorite is Genesis story about Adam and Eve, and even more explicit, Thucydides’s story of the 426 BC tsunami: "The cause, in my opinion, of this phenomenon must be sought in the [non-manipulable] earthquake." But you are probably seeking a mathematical paper, here are some:
9.12.2020 2:10pm - 1. "Does Obesity Shorten Life? Or is it the Soda? On Non-manipulable Causes,"
2. "On the interpretation of do(x),"

9.12.2020 3:20am - (Replying to @desai_pratik) Believe me, in my age, 2018 seems "in my youth," because I almost forgot what I wrote there and, upon re-reading, it sounded so fresh and young; I wish I could write like that today. Perhaps I still can ...

9.12.2020 1:51am - (Replying to @EyalShay2) There are many Ex-Palestinians in Qatar, highly influential. And lets not underestimate the ideological influence of Al-Jazeera, which has hired top journalists with huge salaries to propagate the message of Arab rejectionism.

9.12.2020 1:35am - (Replying to @mkliger) Believe me, I know Hatikvah and I know Die Moldava. I also confided with the organizers after the event. It was Hatikvah by design and intent, to send a clear message to every Jew: "we know where your soul is, and we are willing to accept you"

9.12.2020 12:25am - (Replying to @EyalShay2) Qatar is a sad and perhaps hopeless case, firmly in the grip of Hamas-led operatives, Al Jazeera type. I spent a week in Doha (2005) and did not find a single person ready to accept Israel, yet the Emir himself told me: "we are trading with Israel". Me: Qatar will be the last.

9.11.2020 11:06pm - On a personal note, I knew that Bahrain means it already in September of 2017, in Los Angeles, during a visit by Prince Nasser. To my total surprise, the Bahraini orchestra started playing Hatikvah; not the usual religious tunes, but Hatikvah. For me, that was IT! A done deal!

9.11.2020 10:47pm - (Replying to @ewerlopes and @andreiformiga) I owe a lot to my computer-science education when I was struggling with the do-calculus and the ladder of causation. The CS molds of: semantics-syntactic-proof-algorithm-complexity were/are indispensable in managing a new language such as causation.

9.11.2020 4:25pm - 2 hours ago: Aren't we all waiting for Western intellectuals to tell us why some peace is really war and vice versa? And why "Fair is foul and foul is fair" (Macbeth)

9.11.2020 6:11am - (Replying to @againstutopia) You cannot be "oppressed" by a collective narrative that has existed for 2000 years. You can be "oppressed" by the consequences of trying to crush the organic culmination of that narrative: peaceful co-existence of two equally indigenous peoples, read: Zionism.

9.11.2020 5:59am - (Replying to @againstutopia) Beg to remind you that Anglo SA's history was not born is SA, they did not speak a language that was spoken there centuries before the local language, and did not pray 3 times a day (for 2000 yrs) "we will return to SA, in sovereignty"; connections that define indigenous status.

9.11.2020 5:40am - (Replying to @againstutopia) It is unfair, if not racist, to say that a concept "carries with it the weight" that its malingers fabricated to discredit it. Saying so reflects more on the sayer's susceptibility to propaganda than on the concept itself.

9.11.2020 4:08am - (Replying to @againstutopia) My conception of what "Zionist" means is not unique at all; this is what "Zionist" means to all Zionists, i.e., the vast majority of Israelis and sizable majority of American Jews. First principle of respect: to find out what "Pashtun" means, you ask Pahtuns, not their aligners.

9.11.2020 1:54am - (Replying to @againstutopia) Until such utopian time when linguistic, historical and identity differences are eliminated, Zionist students are asking to be treated with same respect as Pashtun students. Has anyone questioned your "fitness" to serve on student government on account of being a Pashtun?

9.11.2020 1:44am - (Replying to @zeemo_n) I cannot speak for "them". And it is dangerous today to speak for folks with professional pride.

9.11.2020 1:34am - (Replying to @VladicaV @robinsjami and @_MiguelHernan) I hope it meet your expectations. You can find the Ladder of Causation demonstrated already in chapter one, though with no ladder.

9.11.2020 1:30am - (Replying to @againstutopia) Care to take my litmus test for a Zionophobe? Say out loud: "The Jewish people has historical right to a sovereign homeland in the area that is now Israel." Try it! Forget all the populist slogans and accusations, just say those words out loud. No Zionophobe ever succeeded.

9.10.2020 11:51pm - Inside Higher Education had an alarming article today, on college anti-semitism: Although it references the USC Faculty Letter, it misses the main point: condemning antisemitism is seeking a license to do more of the same, namely,
9.10.2020 11:51pm - avoid naming the real virus behind campus bullying --Zionophobia

9.10.2020 8:36pm - (Replying to @VC31415) I would never use "effect" in predictive context - its sends a shock in my spine, perhaps because it reminds me of the dark days of regression analysis. I would use "trend", "slope", "conditional expectation" etc. not "effect".

9.10.2020 5:49pm - The expression "causal effect" is a relic of a period when people (mostly stats.) thought there are "non-causal effects". I believe it was Don Rubin who insisted on the distinction. Today, when we understand that all "effects" are causal, "causal effect" is redundant poetry.

9.10.2020 2:10am - (Replying to @EyalShay2) I am not sure that we are seeing the end of this genocidal strategy, given its enormous success among Western intellectuals, especially the gullible and misinformed.

9.10.2020 1:37am - Can't believe my eyes. "It contains DAGs!" is now a selling point for econ. textbooks! In 2010 I approached Jeff Wooldridge to do one jointly. He chose a DAG-free route, to the tune of a tormented generation of econ students. Are we seeing an econ-spring? #econbookclub

9.10.2020 12:43am - (Replying to @osazuwa @tdietterich and 2 others) You can't analyze "bias" if you can't express the quantity you wish to estimate, ie "effect". This is something mainstream stat could not express, until the 1970-80's.

9.10.2020 12:33am - Insightful observation. I've never looked at graphical models as a "language for representing measurement", but it is true, since "data generating process" is the same as "measurement process". Time to look at "measurement theory" and see if new results emerge, having a language

9.10.2020 12:01am - Who says things only get worse? I am old enough to remember the 1967 Arab League Summit, famous for its "Three No's"; No peace with Israel, no recognition of Israel, no negotiations with Israel. Yesterday's meeting of the Arab League sang a different tune:

9.10.2020 12:01am - For us, Israelis, the "Three No's" resolution was devastating; hopes for peaceful co-existence were shattered and a bleak fate of living on our sword to the end of time seemed inevitable. It is from the darkness of those "Three No's" that I view recent events in the Middle East.

9.9.2020 4:25am - (Replying to @rodakker and @mgaldino) From what I recall, Jaynes had no causal symbols. "No symbols no science" said Augustus de Morgan (1864).

9.8.2020 10:19pm - Holy Moses! Some truths take doing to get accepted!

9.8.2020 12:15pm - I was wondering too. I talked to McGrayne about it and it turned out the people she interviewed for her book were all Bayesian stats, namely, total oblivion to causality. They even claimed Bayesian Networks are not "Bayesian", because they do not invoke priors on parameters. 2/2
9.8.2020 12:15pm - 2/2 Bayesian stats do not even agree on "Who is a Bayesian", what is "inverse" probability, or on whether subjective knowledge is essential. IOW, lacking causal understanding, Bayesian stats don't agree on what Rev. Bayes has accomplished. See & #Bookofwhy

9.8.2020 4:44am - (Replying to @RobCalver5) Incredible. #DiscreteMathematics, of all fields? Mathematicians, so it seems, have a nose for new challenges.

9.7.2020 12:10pm - Gratified to see #oncology #Guoncology and #urology entering the modern era of causal analysis. Key quote (on limitations of DAGs): "These assumptions, however, are necessary for any research study, and the advantage of DAGs is that they make these assumptions explicit."

9.6.2020 10:52pm - Another antizionist Jew: "Mother, I hate you! The boys in school called you names and, unable to defend you, I joined them! I hate you for making me feel so stupid! And so guilty!" (After:

9.6.2020 9:56pm - (Replying to @VC31415) I dont think the stretch is such a stretch. Statistics has developed its own mathematical notation, based on properties of distribution functions (ie no do()) because it was developed before structural equations. It would be L3 speculation to ask: "what if history was different?"

9.6.2020 6:16pm - (Replying to @yudapearl and @VC31415) To stretch the analogy, one can argue: "Who needs statistics (L1)? Isn't it just a special case of L3 ? Yes it is! But it is such an important special case that universities create departments in its honor, and train thousands of PhD's who know nothing about L2 and L3.

9.6.2020 5:23pm - As I predicted in a previous tweet, the Open Letter of 43 USC professors about Rose Ritch resignation has become a turning point in the struggle against Zionophobia. This editorial of the Jewish Journal titles it "I am a Zionist - A New Frontier" :

9.6.2020 4:51pm - (Replying to @VC31415) We have an appropriate generaliztion of do() for L3, it is do(x) itself, which is the subset of sentences in L3 that can be inferred from population experiments. do-calculus merely recognizes that this subset of sentences deserves special attention and a special vocabulary.

9.6.2020 12:07pm - Why is L2 different from L3? The question keeps coming up every 2 weeks or so; glad your answer is ready for posting and re-posting. The reason people are still asking: (1) do(x) is ignored by the Harvard Epi group and (2) It's never been acknowledged by revered economists.

9.5.2020 4:25pm - (Replying to @ModernMaccabi @AndrewPessin and 2 others) I don't think so. Having read the students comments, they are just ordinary hungry alligators, emboldened by frightened university administrators and silent academics. But things may change tomorrow.

9.5.2020 4:19pm - (Replying to @JaapAbbring @VC31415 and @PHuenermund) I am not sure either. The only paper I saw on this connection is: It offers a platform for modeling an agent's state of belief which is different than yours, analyzing and rectifying causal misrepresentations.

9.5.2020 3:09pm - The village elderly told us the alligator only needs one baby each morning, which is not unreasonable, leaving most of us safe. Today they say: no more babies left, and it's our turn. We? Who have been so pro-alligators? Isn't it time we try an elderly once?

9.5.2020 3:38am - (Replying to @offaltube) Eizehu Gibor? Ha'Kovesh et Yitsro! (Hard at my age)

9.5.2020 2:46am - A thread in their memory -- 11 torches on the altar of the world's normalization of evil.

9.5.2020 2:34am - Thank you for all the birthday congratulations I received on this screen yesterday, especially the Twitter's balloons, a pleasant reminders of how blessed I am, at 84, with 36 more years of productive life, to have so many friends among seekers of truth and lovers of fun.

9.4.2020 4:36pm - (Replying to @deaneckles) You'll probably get lots of "see here"& "see there" replies but, to prove that the problem has NOT received due attention, remove U, assume linearity, now what? Do we know what if anything is gained by measuring M? See a beginning of an in-depth analysis:

9.4.2020 4:28am - (Replying to @GivingTools) The former does not depend on X, the latter does. The former is computable from observed data (+assumptions), the latter isn't. Averaging over X and limiting X to pre-treatment variables, might get you somewhere.

9.4.2020 1:27am - Some readers wrote that the PO framework as formulated in @VC31415 lecture notes is not totally helpless in dealing with sample selection bias, a problem that can be managed by "reweighing" methods. This is not exactly true, see:

9.3.2020 11:16pm - (Replying to @JaapAbbring) Please alert us when it is published and send us a link to the Editorial.

9.3.2020 4:56pm - I've found 5 interesting comments added to Dana MacKanzie's post (july 6 2020) on Race, Covid, Mortality and Simpson's Paradox: Just click on "comments" at the end of the post.

9.3.2020 4:39pm - (Replying to @JaapAbbring) Thanks for clarifying. The paper referenced did not have the title of the Special Issue spelled out. Perhaps it is not too late to add it on.

9.3.2020 1:42pm - Evidently, The Econometrics Journal had a special issue on synergies between structural economics, AI and ML, see The synergies, however, are not defined in terms of "input-output" capabilities, but added as an after-thought

9.3.2020 11:43am - (Replying to @SilvioZaina and @SamHarrisOrg) That's a valid question. Answer: Identity claims are not based on conquests of YESTERDAY but on collective memories of TODAY. Ask a Danish child: "What's E. England to you?" U'll get: "What?" Ask a Jewish child: "What's Jerusalem to you?" U'll get "We were born there".

9.3.2020 1:58am - (Replying to @VC31415 and @Julianaccp) Another point worth noting, the term "selection bias" has a totally different meaning in the CI literature, standing for preferential selection of subjects to the study (not to treatment). This bias is not removed by randomization; see for remedies.

9.3.2020 12:51am - The reluctance to address Zionophobia is not totally Jewish leaders fault; university administrators push them into fighting antisemitism instead. It gives them a cover up for inaction and an excuse for not addressing the real problem - Zionophobia, which takes moral courage.

9.3.2020 12:08am - (Replying to @Thani75) I've been called Zionist too -- a badge of honor! I hope no one ever calls you a Zionophobe -- the ugliest word in town.

9.2.2020 11:20pm - An opportunity to contribute. The 2020 NeurIPS conference is providing a Workshop on Causal Discovery and Causality-Inspired Machine Learning; submission deadline 10/10/2020. For details, see:

9.2.2020 5:03pm - Why "unprecedented"? How many professors in your university would risk their career and reputation to speak out for an unpopular cause they believe is true and just in the days of COVID-19 and cancel-culture? I could hardly count 5-7 such professors at UCLA.

9.2.2020 4:49pm - This is a beautiful example of mental causation. Some people take coffee not because it helps them wake up, but because it's in the morning's script, and our mind is a slave to scripts, habits and conventions.

9.2.2020 12:57pm - I was just notified that the "prologue" to Kahneman's book is copyrighted and has not been made public yet. Please refrain from disseminating it electronically.

9.1.2020 6:43pm - (1/2) An unprecedented letter by top USC professors demands an end to the harassment of Zionist students on campus. The letter is a game-changer. The anti-Zionism virus, which has long been ignored or considered untreatable, now demands a specific vaccine,
9.1.2020 6:43pm - (2/2) tailored to its own distinct chemistry. This long overdue letter is inspirational to students and faculty nation-wide, and I salute the authors for their courage and insight.

9.1.2020 10:50am - For months now readers have been complaining that the first issue of Journal of Causal Inference disappeared from the website. Today the publisher informs us that it is back on. Victory! See for yourself: I am especially fond of the "linear microscope"

9.1.2020 10:33am - This JRSM article is one of the clearest one I have seen on the COVID-19 pandemic, including retrospective epidemiological and social considerations

9.1.2020 10:15am - (Replying to @fmg_twtr and @JadePinkSameera) Fix the screw is what Kahneman would call "hygiene" - eliminating the source of the noise all together, as opposed to say compensating for each jitter individually

8.31.2020 6:52pm - (Replying to @paul_conyngham) I'll ask John Brockman, our publisher, if its available.

8.31.2020 12:59pm - I asked Kahneman why attention to cognitive bias has dominated research in the past. His answer: Being associated with a causal mechanism, Bias was deemed reparable. Noise, on the other hand is avoidable (by hygiene) but not reparable.

8.31.2020 2:14am - (Replying to @Zio_conspirator @Claire_Voltaire and @DeedsWylie) The descendants of the Bund have abandoned the Bund's socialist ideology and have formed mainstream American Jewry, which helped established Israel and a strong US-Israel bond. Those calling themselves Bundists in 2020 are proclaiming themselves an outcast of mainstream Jewry.

8.31.2020 1:20am - Daniel Kahneman is coming up with a new book, titled NOISE. Had a fun Zoom session today with a bunch of eager would-be readers, and received an illuminating prologue that I am happy to share: Truly Deep Learning.

8.31.2020 12:56am - (Replying to @_onionesque) No, I am not aware, but please alert me to it, in case I have been missing any. There is of course the works of Dreyfus and Searle on why AI will never answer any philosophical question. But as you can tell from my smile, and @Bookofwhy, I do not recommend this acrobatics.

8.30.2020 8:40pm - (Replying to @_onionesque) No, it was authored by Mathias Frisch (see bottom of the page). I should have mentioned it in my post.

8.30.2020 5:57pm - "Causation in Physics" : " is a new entry, just posted on the Stanford Encyclopedia of Science, which includes modern perspectives on the role that causation plays in physics. I still think it should be supplemented with my epilogue:

8.30.2020 2:23pm - (Replying to @dr_agc) Not clear what problem you refer to.

8.30.2020 2:22pm - (Replying to @SilvermanJacob) Sorry, I did not check the website, nor its reputation; I was struck by the implications of the DA's words.

8.30.2020 2:17pm - (Replying to @AkberKhan) Its not the riots which historians will find most interesting about Aug. 2020, but the way our leadership (DA's, WH, parties) missed opportunities to mend economic inequities by inflaming extremes on all sides.

8.30.2020 4:21am - I am retweeting this quote for archival purposes only. Future historians should have some idea of what civilization had to stomach in August of 2020.

8.30.2020 3:40am - (Replying to @WokeRabbi) And we prayed: U'Mipnei Chataeinu Galinu M'Artsam! Oy Lanu! Oy Lanu! Hashiveinu Adonai V'Nashuva, LeAn Adonai? LeAn?

8.30.2020 2:58am - (Replying to @TobiasWolbring) I hope we agree that we can't do the hard part properly without doing, or at least understanding the easy part.

8.30.2020 12:27am - (Replying to @Claire_Voltaire and @pedroborbon3) If anyone wishes to see where Bundists thrive and how they dominate Jewish education, come to UCLA. Our "Center for Jewish Studies" hasn't celebrated/mentioned Israel's Independence Day the past 20 years. I often wonder if their students know that Israel has gained independence.

8.29.2020 2:19am - An interesting generalization of structural causal models, allowing for interventions to alter the causal relationships acting in the future.

8.29.2020 1:58am - Another paper on "fairness": I understand and agree with the counterfactual definition, Eq (1), but I need help understanding the role of variational inference in the overall scheme.

8.29.2020 1:27am - Sarsour Says ‘Right-Wing Zionists’ Are Aligning With White Nationalists to Smear Her via

8.29.2020 1:27am - Sarsour is suffering from severe dementia. Her first exposure as a "fake feminist" was done by progressive Muslim women and super-progressive Zionist women. See And her Zionophobic-bigot title was given to her by a progressive male. None is "right-wing."

8.28.2020 7:07pm - I have received this paper: "A Causal Look at Statistical Definitions of Discrimination and confess to being unaware that stat definitions have ever been proposed, because such definitions are impossible - discrimination is a causal, not statistical notion

8.28.2020 12:34pm - Why can't we forgive INN - a bunch of confused juveniles who just want to say: "we dont know who we are and what we stand for, but we know we should have an opinion on everything; we sound so good as we sing empty slogans"

8.28.2020 12:02pm - You are right - these two papers are content-rich; they should have been read and cited by way over 30 sociologists. Strange kind of folks, sociologists. They write tons of papers on how important a certain problem is and, when you hand them the solution, they lose interest.

8.28.2020 12:54am - It is a tribute to Sewall Wright who stated this truth in 1920 and defended it all his life without having the mathematics to prove it in general. Today it is a mathematical theorem, e.g.,

8.27.2020 12:41pm - (Replying to @JohnKubie @GaryMarcus and 7 others) One should add that that classical logic is inadequate for reasoning with counterfactuals. Yet, the brain does the latter better than it does the former.

8.26.2020 2:35pm - See why the recent harassment of Zionist students at USC is more than a local scandal, but threatens to undermine academic freedom on all US campuses: I feel obliged to help colleagues who now find themselves unwelcome on a campus they have helped build.

8.26.2020 1:39pm - (Replying to @byronprecio) No, it is impossible to play with conditional probability and get the correct causal effect. For every conclusion you get there exist a world in which the same conditional probabilities hold and in which the opposite conclusion holds. This is why "probabilistic causation" failed.

8.26.2020 1:46am - (Replying to @yskout) Luckily, imagining is the top Rung in the Ladder of Causation.

8.25.2020 11:43pm - (Replying to @mmmDOMPH) no idea - I havn't been out of the house past 4 months.

8.25.2020 11:41pm - Thanks for the raving review - it made me take another look at the table of content and conclude, embarrassedly, that you have a point; the book casts a new light on many topics, including the traditional ones, like Bayes Rule, regression, RCT, IV. Funny! I enjoy reading it.

8.25.2020 11:14pm - (Replying to @offaltube @VinuSelvaratnam and @MakingSenseHQ) It's hard to imagine how a person could find regression to be simpler that causal BN. Unless of course that person got indoctrinated by Stat-101, which makes it hard to come back and think intuitively -- cause and effect. Five-ten years from now, Stat-101 will start with CBN.

8.25.2020 10:42pm - Yes, it is today, August 25. The paperback edition of the #Bookofwhy has made it to the bookstores: I am hopeful that its lower price and (possibly) greater availability might expose our book to a new audience. Behold! It's the #1 “new book” in baseball.

8.25.2020 3:56pm - (Replying to @Abel_TorresM @GaryMarcus and 16 others) Trying to understand, isn't a causal model itself an abstract representation of reality? So what is missing?

8.25.2020 2:17pm - (Replying to @neuro_data @tdietterich and 17 others) For motivation and vivid examples of the transition from Level-2 (intervention) to Level-3 (counterfactuals) I suggest and will never get tired of recommending For ML applications, mediation->robustness see

8.24.2020 10:17pm - What can I tell you? It gives me a sense of immortality to read words written three years ago and to stand behind them, despite the stormy days and the anti-anti cultures that shook our world since.

8.24.2020 10:02pm - The do-calculus transforms do-expressions into do-expressions, all in Rung-2, it does not accept counterfactual expressions like your d-->Y(d) which is in Rung-3. It is all explained so nicely in Primer, see for yourself: The quantity you want to compute
8.24.2020 10:02pm - E(Y(1)|D=0, X), Effect of Treatment on the Untreated, happened to be identifiable (for binary treatment) whenever the average causal effect is, see

8.24.2020 10:02pm - (Replying to @tdietterich @wooldridgemike and 17 others) Clarification: You can "learn about" Level 3 from lower levels, but you cannot UNIQUELY answer Level 3 questions. That is why we end up with BOUNDS for causal sufficiency and necessity, not with point estimates. As to innateness, L-3 sentences are DEFINED by L-3 representations.

8.24.2020 7:37am - (Replying to @YasMohammedxx) Don't fly, Don't stop Don't eat, nor drink A Tantrum attack is all that a Normali-phobe @cjwerleman can think.

8.24.2020 6:10am - (Replying to @DavidSalazarVir) Very nice. And this link would aid readers visualize the bounds.

8.24.2020 5:57am - (Replying to @ViktorHardarson) By calling her "antisemite" you give her the benefit of some hereditary disease or unfortunate upbringing, which she does not deserve. She is a calculating Zionophobe - a more dangerous form of genocidal ideology.

8.24.2020 5:38am - (Replying to @CasualBrady) Very nice, thanks for sharing.

8.24.2020 5:34am - Explaining my preference: Ask a statistician about Simpson's paradox and you can tell right away how far she is on the road to causal inference. The majority just don't get it, even the gurus (see here:, but the enlightened never go back to statistics.

8.24.2020 3:46am - (Replying to @Plinz @dileeplearning and @GaryMarcus) We can definitely do better, using filtered mutations, as I argue here: But, in order to do better, we need to know what we are looking for, and what we can do with it once we find it. That is why I study what to do with a DAG once we find one.

8.24.2020 3:17am - (Replying to @dynamite_ai @learnfromerror and @dataengines) No frustration. You take the most "principled" null-test that stat can provide and you apply it n times, where n is the number of missing edges in the graph. Plus, this is NOT the most imp part of CI; what you do with the DAG is not less important, see

8.23.2020 7:00pm - (Replying to @dynamite_ai @learnfromerror and @dataengines) Subjective? Not more subjective than standard statistics where some researchers use a cut-off of 0.05 (for the Null Hypothesis), some at 0.01 and some argue if the p-value is appropriate or alpha-value or more ...

8.23.2020 1:18pm - (Replying to @learnfromerror and @dataengines) Yes. "Statistical falsification" is the key here, though the precise criterion for falsification varies from researcher to researcher.

8.23.2020 1:14pm - (Replying to @dataengines) Yes, this is a good summary of the methodology, focusing on genetics.

8.23.2020 12:57pm - (Replying to @dataengines) If there was such a thing as "establish the validity of a DAG from data" we would not need the DAG, we would go to the DATA. Since DAGs have testable implications, we need to go backward and tease out a "set of DAGs" compatible with the data. This is called "causal discovery."

8.23.2020 12:49pm - (Replying to @F_Sammarco and @DanielePaliotta) I recommend Primer Which is essentially free here * primer page:

8.23.2020 12:22pm - Thanks for posting. I wasn't aware of so many applications, stories, and metaphors of Causal Bayesian Networks.

8.23.2020 5:34am - Explaining my preference: Ask a statistician about Simpson's paradox and you can tell right away how far she is on the road to causal inference. The majority just don't get it, even the gurus (see here:, but the enlightened never go back to statistics.

8.23.2020 5:12am - (Replying to @PHuenermund @VC31415 and @jermainkaminski) Agree with Paul. Simpson's paradox is the best way of introducing causality (and DAGs) to students, because its resolution rests on causal structure, ie DAGs. The Google example is striking, but students yawn over numerical examples. I prefer qualitative ways, scatter diagrams.

8.22.2020 11:06pm - Readers who resonated with my plea to have Holocaust memorials change focus, from the dead to the living, will find this incredible article (and video) by Ruth Wisse to be the cornerstone of a new era in Holocaust education. @USCShoahFdn @simonwisenthal :

8.22.2020 10:33pm - It makes perfect sense once we invoke the difference between explicit and implicit knowledge. Giving a machine the rules of chess it should "know" what the best move it from any position. Yet it makes perfect sense for a machine to say: "I can't predict who the winner is"

8.22.2020 9:43pm - Next step will be a deterministic Turing machine showing Searle that free-will and consciousness can be algorithmitized.

8.22.2020 3:13pm - (Replying to @nickchk @pablogerbas and 2 others) Beatifully done!!

8.22.2020 3:10pm - (Replying to @VC31415 and @PHuenermund) Here is an easy "after-Angrist" WOW : "As we have seen, deciding which variables to include in our analysis is not trivial, even supra-economists can't distinguish good from bad control. Today, I will show you a simple way of doing it, and much more. See

8.22.2020 2:49pm - (Replying to @BedecarratsF @VC31415 and @PHuenermund) I don't think it makes sense to cover topic X following someone who never used X. Imbens (2019) is a good example of someone talking ABOUT DAGS who never used one and, thus, fails to present their merits and limitations; I detail his failures here:

8.22.2020 3:40am - This brings us to the question whether there is anything at all that is more than just "a construct in the head of the observer" or, equivalently, whether @kaushikcbasu tried to imply that nothing exists but raw data, and/or nothing is worth talking about except sensed data.

8.21.2020 9:35pm - (Replying to @plevy) Relying solely on data is a temporary cultish phenomenon (read my lips!), as I argue here

8.21.2020 9:29pm - "A construct in the head of the observer" in not "nothing," but an important piece of knowledge when the construct is shared by many observers. It is in fact what AI is attempting to capture and what statistics can't, lacking an adequate language. Econ had a language - abandoned.

8.21.2020 2:14am - A nice summary of the current state of reinforcement learning (RL) in the context of Robot Learning and the tension between "innate structure" vs. "model-free" learning in the quest for robustness.

8.21.2020 12:16am - (1/3) CI is not looking for a home in related communities, it OFFERS a home. CogSci folks are already sold on CI because DAGs are the most natural representation of agent's "belief system", no competitor in sight. The reason I am banging on the gates of ML is to prevent them from
8.21.2020 12:16am - (2/3) squandering more of society's resources (including education and training) into dead alleys, from which it would be hard to recover. The ML community is under the tyranny of isolationism, which makes my effort seem Don Quixotic. But, as they say in the Mishna: "It is not
8.21.2020 12:16am - (3/3) upon you to finish the job, but neither are you free to desist from it" (Essays of the Fathers, Avot 2:21). I bet the next five years will see a tectonic shift in ML literature, vocabulary, thinking and training.

8.20.2020 2:33pm - To me, yesterday's court decision to keep the Damascus Crown in Jerusalem: is more than a spy-smuggling story. It represents the role of Israel's sovereignty as a custodian of Jewish civilization, a role that the "one state fantasy" would have to abandon.

8.20.2020 2:39am - (Replying to @AJCGlobal and @MortonAKlein7) Commitment to Israel and to Jewish people-hood is a valid certificate of conversion, more valid than any Rabbinical blessing.

8.20.2020 2:29am - (Replying to @ShMMor @EinatWilf and @OrenGross) Same for the argument that "statehood is not inherent to Jewish thinking".

8.20.2020 2:02am - (Replying to @IMourifie) I have learned to sniff rejection from 2 miles away, and I am an even better sniffer of Econ Top 5 rejectionism.

8.19.2020 11:20pm - Good news!! Google Scholar invites you all to a special celebration + party: The number of publications citing my works has crossed the 100,000 mark. There will be no beer this time, just music of appreciation to the many writers who made me feel useful.

8.19.2020 11:02pm - (Replying to @Nishit32755501) I am referring to the Twitter post you are currently on, which I regard as an "educational channel", due to its commitment to substantive education in the science of cause and effect.

8.19.2020 10:05pm - (1/ ) A devil advocate might argue: But why SEM? What if I want to formalize my economic theory in the language of potential outcomes (PO)?? Answer: (1) You can't. You wont be able to defend your assumptions. See (2) Detecting testable implications in PO is
8.19.2020 10:05pm - (2/ ) Detecting testable implications in PO language is still an art, the logic of it is hatching slowly, see:

8.19.2020 9:01pm - I see that our educational channel has garnered the attention of 35,000 followers. Thanks for the encouragement to continue our journey towards the scientification, algorithmization and a broad understanding of causes and effects. #Bookofwhy

8.19.2020 8:51pm - (1/ ) Being an "economic concept" does no exonerate it/us from the obligation to define it as a property of our economic theory. (Same as we define ATE or LATE) If our economic theory is formalized as an SEM, S, it behooves us to define a concept C as a property of S, namely,
8.19.2020 8:51pm - (2/ ) as a mapping from ANY S to the values of C. As to testable implications, "over identification" covers a tiny fraction of the testability landscape, missing for example, the vast territory of d-separation, equivalent models and Verma constraints. See

8.19.2020 3:20pm - (Replying to @VC31415 @JaapAbbring and @edwinleuven) My view differs: potential outcomes, hence Supply and Demand functions, are products of SEM, by definition. And I would not call SEM history "fine" if its practitioners can't tell, even today, what the testable implications are of a model, or which parameter is estimable by OLS.

8.19.2020 6:51am - (Replying to @JaapAbbring and @VC31415) Footnote 10, page 216. "I presented this problem to well over a hundred econometric students and faculty across the United States....none managed to answer question 3"

8.19.2020 6:32am - (Replying to @JaapAbbring and @VC31415) The Goldberger's model is linear, so it does not accommodate rationing. Still, as I report in the footnote, I haven't met an economist who could solve the counterfactual question. Perhaps they can do it now, 20 yrs later. Not sure.

8.19.2020 6:01am - (Replying to @IMourifie) I was asking because my motivation for looking into the instrumental inequality was a statement in some econ. Journal that IV has no testable implications. Plus, I was wondering if your paper was rejected first by one of those journals.

8.18.2020 7:22pm - (Replying to @deaneckles) Taking Gelman's quote seriously it works both ways: "Theoretical CI is the theory of applied ML." Or, as I put it, solving one toy problem from begining to end tells us more than blindly running 100 "real life" problems. It also explains why Gelman abhors CI toy problems.

8.18.2020 7:12pm - (Replying to @memosisland) Agree. But I would not call it "misunderstanding". We understand very well that the aspiration to learn everything from data alone has kept the ML community away from science.

8.18.2020 7:06pm - Nice paper, thanks for sharing the final version. Curious: Given that both authors are economists, why is the paper published in Biometrika and not, say, in Econometrica??

8.18.2020 3:11pm - I am also puzzled why ML folks tend to run away from the solution, or why the available solution is not satisfactory to them and, if so, what else they would like to have.

8.18.2020 8:11am - Agree. Causal Science cannot be "ML with a sprinkle of CI". It must start with causal models of the world and leverage ML as a means for interrogating the world.

8.18.2020 7:59am - Thanks for re-posting this talk, I believe I apologized for the slides synchronization, but I stand behind the content. I should also mention that participants received these posts as handouts:, and

8.18.2020 7:54am - The wheels of justice are turning slow, and rusty. But the rustier they get the louder their squeaks, and the greater our hopes. Thanks everybody for helping pull the ropes.

8.16.2020 7:59pm - Thanks for re-posting this talk, I believe I apologized for the slides synchronization, but I stand behind the content. I should also mention that participants received these posts as handouts:, and

8.16.2020 7:49pm - Two questions. (1) When do you think a ML textbook will be authored by one of ML revered figures, weaving CI and ML? (2) Should CI be woven into ML or the other way around, ie, starting with the Ladder of Causation, show how ML takes us from samples to distributions.

8.16.2020 12:42pm - (Replying to @zacharylipton) I hope you will. But, in addition, you know the big shots in ML. How about urging them to write one; they won't do it unless someone like you confess to them what you think the most urgent need is in ML education.

8.16.2020 12:37pm - (Replying to @zpardos) "data-accessible" is also good. Some claim that classical science is necessarily "data-assistive". But "accessible" connotes "big data" or even "unlimited data".

8.16.2020 6:10am - (Replying to @ajaydiv) Well put: "all consequential structure will just emerge from data". I've addressed this ideology here:

8.16.2020 5:52am - True, same indictment can be used against econometrics education, but ML should be held to higher standards; Most ML folks speak computer science and should therefore be more flexible in accommodating new (formal) languages.

8.16.2020 5:18am - (Replying to @rohan_virani) Can you link us to the story?

8.16.2020 5:14am - Two additional observations on skeptical ML practitioners: (1) They have not been taught that, mathematically, causal models are NECESSARY for certain tasks, and (2) They have not been taught how to USE causal models even if they were handed one. (1) + (2) drives ML skepticism.

8.15.2020 10:28pm - Man's craving for "data-driven" is as ancient as man's craving for "objectivity"; its blindness is as ancient as man's trust in the brainless naked eye.

8.15.2020 6:35pm - @careem_carr, thanks for sharing these valuable exhibits-- the psychiatry of hate is still in its infancy. I wish I could say: As a Jew I know how to handle it. But I can't, because I grew up in Israel, as free as a palm tree, convinced that the psychiatry of hate is obsolete.

8.15.2020 3:45pm - (Replying to @geomblog and @rcsaxe) No, No, please do not ask @yudapearl ; Ask mathematics what can and cannot be done and, see for yourself, that you cannot climb the ladder of causation w/o help from the rung you wish to climb. Btw, what is "latent knowledge"? A new kid on the block?

8.14.2020 3:42pm - (Replying to @zacharylipton) How about writing that forward-looking book, from an insider viewpoint, that will prepare ML practitioners for tomorrow's jobs?

8.14.2020 3:38pm - (Replying to @SheldonRanz @FJnyc and 2 others) BDS is not anti-Zionist???? Why do you think did Noam Chomsky call them "hypocrisy rising to heaven"? And what did their co-founder Omar Barghouti mean by: "They [Jews] are not a people"? See my take on "BDS and Zionophobic Racism"

8.14.2020 1:32pm - I thought ML practitioners are waiting for a book that would prepare them for future ML systems, i.e., systems that climb the Ladder of Causation, from data-fitting to data interpretation, along and

8.14.2020 10:50am - (Replying to @FJnyc @SheldonRanz and 2 others) Happy to see the word Zionophobic used to settle disputes that no other word could. Yes, Zionophobia stands apart from antisemitism. The latter has some excuses (eg mental disorder, improper upbringing, etc), the former none. It's the ugliest word in town.

8.14.2020 9:22am - Perhaps it's old age that makes me sentimental, but watching these two flags moving with the "winds of peace" compelled me to click "retweet with comment"

8.14.2020 4:27am - I bet many of our readers are not aware of the fact that some people on earth are still proud of owning a passport stamped "except Israel," like the one below. Moreover, the stamp was configured in 1948, not in 1967. The UAE-Israel agreement is flushing out interesting wrinkles.

8.14.2020 3:41am - Very insightful observation by @EinatWilf below. When I talk with Palestinian intellectuals they invariably tell me: you can "belong" but not "own". Normalization means we can belong and own like any other nation on earth.

8.14.2020 3:29am - (Replying to @emil_krabbe) The importance of knowing what to do with the model is down played by some economists who argue they do not teach students what to do because, in real life, one is never sure about the model. This is IMO a non-scientific approach that accounts for much of econ. students confusion

8.13.2020 3:01pm - (Replying to @dynamite_ai) It sounds like that, agree, but if we examine what "compromise" means in this case -- internalize Israel's existence -- then it sounds like a win-win option, in fact the ONLY option.

8.13.2020 2:35pm - The significance of Israel-UAE deal is further illuminated by @EinatWilf 's escellent book The War of Return which explains, facts and figures, why the Palestinian dream of two Palestinian states can't work, and why emboldening that dream is anti-peace.

8.13.2020 2:13pm - (Replying to @jcsahnwaldt and @The_RickMc) No, he is not a liar, just a deliberate exaggerator, for a mean purpose. Nor is he a "bad person", but he knew very well how his exaggeration ("my entire life") can and will be used by the deniers of his people's history and people-hood. Not someone to date.

8.13.2020 1:46pm - (Replying to @demirlenk92) Peace prefers corrupt Kings on saintly Ayatollahs.

8.13.2020 12:09pm - Why is this deal of historical significance? Because it sends a message to Palestinians: compromise or lose, to Arab states: what are you waiting for? To Iran: the coalition against you is being shaped. And to the world: You want peace? Here is one way to get it. (S. Rosner)

8.13.2020 11:43am - (Replying to @jcsahnwaldt and @The_RickMc) "Oh, by the way, there were people there" is an accusation of deliberate institutional deceit, not only of his teachers, but of the whole Zionist narrative and Israeli education. To this he added "my entire life", the poor boy... Deliberate malingers do not deserve your tears.

8.13.2020 11:15am - BBC News - Israel and UAE strike historic deal to normalise relations

8.13.2020 10:39am - (Replying to @jcsahnwaldt and @The_RickMc) Twain was not humoring when he wrote "we never saw a human being". He is describing the road from Jaffa to Jerusalem in 1867. Even Palestinian historians confirm his observation except, so they say, he should have climbed a few hills off the road to see the shrubs he missed.

8.13.2020 10:15am - (Replying to @jcsahnwaldt and @The_RickMc) There is a difference between "empty" and "desolate". Ben-Gurion counted 700,000 Palestinians west of Jordan in 1917 -- that is not empty. My father had to carry worm- infected water, daily, from 2 km away -- that's desolation.

8.13.2020 9:49am - (Replying to @jcsahnwaldt and @The_RickMc) Joan Peter's and Mark Twain accounts have elements of truth to them. Both the massive immigration of unskilled workers from Egypt and Syria and the desolation reported by Twain were also witnessed and documented by my father, who came there 1924.

8.13.2020 8:47am - (Replying to @The_RickMc) I am referring to these "broader narratives". The idea that Palestinians are "equally indigenous" to the land has been at the core of the Zionist narrative since day one, and at the core of Zionist education to this very day.

8.13.2020 8:34am - (Replying to @The_RickMc) My evidence? A strongly held assumption that Rogen's teachers used the same textbooks as my teachers.

8.13.2020 8:28am - (Replying to @The_RickMc) This is precisely my point (quote): "I do not believe for a second that Rogen was actually told Eretz Israel had no people there."

8.13.2020 8:12am - "knowing the causal model" may not be necessary, but knowing what to do with a causal model once we have it IS necessary, as I argue here:

8.13.2020 8:05am - Remember Seth Rogen's interview on Marc Maron's "WTF" podcast? and his heartbreaking confession about how he was "fed a huge amount of lies about Israel?" I wrote a piece on this funny episode, assuming that "Facts Do Matter."

8.13.2020 2:13am - (Replying to @jcsahnwaldt and @The_RickMc) No, he is not a liar, just a deliberate exaggerator, for a mean purpose. Nor is he a "bad person", but he knew very well how his exaggeration ("my entire life") can and will be used by the deniers of his people's history and people-hood. Not someone to date.

8.12.2020 3:39pm - There is a blip on the Babylon+UCL project in today's Times: I tried to educate the public on the difference between Rung-1 and Rung-3 inferences.

8.12.2020 6:28am - (Replying to @stephensenn and @learnfromerror) Of course it is very important. But this does not negate my description of what Data Science practitioners do.

8.12.2020 6:17am - (Replying to @tdietterich) I do resonate with Slide 12, 2nd bullet: "* Inferential thinking means (inter alia) – considering the real-world phenomenon behind the data – etc." Perhaps because I am familiar with a specific implementation of this kind of "thinking". I do not understand the rest.

8.11.2020 11:37am - But how do we get access to your "supplements"??

8.11.2020 8:42pm - (Replying to @RaulMachadoG) Thank you. Indeed, the supplemental material is there, containing most of the mathematical results.

8.11.2020 8:39pm - (Replying to @EveryDayBME) Not familiar with competitions. Please, lets talk by email about it.

8.11.2020 11:19am - (Replying to @colinkgarvey and @GaryMarcus) I beg to differ about overstating. But, as I argue here:, the ML --> CI transition requires a profound paradigm shift that has not taken place yet at the former's establishments. It takes time.

8.11.2020 10:42am - (Replying to @quantumciaran and @NatureComms) I would exonerate AI from your indictment, since I still see myself as an AI-er, and some of the major contributors to causal inference publish in that field. You are right about ML, an off-shoot of AI that is laboring to restore science into data, see

8.11.2020 5:50am - (Replying to @Kapoterike) The smartest investment in the future of AI

8.11.2020 5:42am - A Nature article reports a breakthrough in medical diagnosis. While most ML enthusiasts are busy mining raw data, a British team has taken counterfactuals seriously (do-calculus, probability of causation, twin networks, noisy-OR) with impressive results.

8.10.2020 12:55pm - I feel like a dinosaur - A 30 yr old paper of ours just got a "Classic Paper Award" from the AI Journal. The paper deals with temporal constraints, a hot topic in the 1980's. To see why, go to Example 1.1 here:

8.10.2020 5:39am - (1/ ) At first I thought that this paper, on Predictive Model Interpretation via Causal Attribution violates the Ladder of Causation, because "attribution" is not definable at Rung-1. Now I see that the authors define "attribution" differently,
8.10.2020 5:39am - (2/2) taking the data-fitted function as a new data generation process and defining "attribution" with respect to this environment. Not sure how problematic this substitution can be. For example, attributing diseases to symptoms, and crimes to ice-cream sales.

8.10.2020 5:14am - For economists immersed in the time-series tradition, this paper by Ken Hoover provides an interpretation of Granger Causality, cointegration, causal-order and other esoteric concepts in terms of DAGs.

8.10.2020 2:50am - (Replying to @Rob_Malley) And with an answer like this @Rob_Malley wants to be taken seriously. "All means all": the elephant and the mouse, rich and poor, merchant & chief. And there is no hidden agenda behind this charitable egalitarian analysis of a scene to which he claims expertise. Go on, Nasrallah.

8.9.2020 10:53pm - I would support this proposal only if Statistics is supplemented with causal reasoning, else we are back to "black box" data-fitting mentality, which always comes at the expense of "data-interpretation" - the main focus of data science.

8.9.2020 1:28pm - (Replying to @shriyamite) The paper that comes to mind on application of do() to legal problem is "Causes of effects and effect of causes" Most judicial question deal with individual-level causation, hence, counterfactual logic.

8.9.2020 7:19am - I've seen it with gullible left-liberal journalists. They get invited to smooch with other l-liberal journalists and, forget facts and morals, they fall for the theory that whatever Israel does right it does to deflect attenti

8.9.2020 3:46am - Moreover, given that "one-state" dreamers (eg @Beinart ) know perfectly well that Israelis won't give up their sovereignty no matter what, is it really inappropriate to call their campaign "incitement to genocide"? and their megaphones (eg @NYT ) "accomplices"???

8.8.2020 6:03pm - And it is anti-Zionism (ie Zionophobia) that is becoming the ugliest form of thuggery on US campuses:, in the best tradition of morally bankrupt ideologies.

8.8.2020 5:18pm - We had a tweeter discussion on whether it is ethical, given that reviewers assume a limited audience. I believe the consensus was: it is ethical. I further suggested that historians will find this material more representative of contemporary science than the paper itself.

8.8.2020 3:43am - (Replying to @l_gervasi) I am not familiar with KPI systems. What is the input and what is the output?

8.8.2020 3:38am - (Replying to @abdallah_fayed) I presume you meant "equally indigenous people" and that you have read the facts about the heart of the Zionist program. I happened to research it thoroughly:

8.7.2020 11:58pm - At long last, a major newspaper is calling on University administrators to address the ugliest and least acknowledged form of racism on US campuses -- Zionophobia -- Denying Jews the right to a homeland:

8.7.2020 8:59pm - Clarifying the night-driving paradigm: To make sensible choices when you doubt the structure of your DAG, first learn how to make choices for a given DAG, second, see if your doubts make a difference, finally, see if there is a choice that remains satisfactory despite your doubts

8.7.2020 8:33am - (Replying to @PophamFrank @frankdevocht and 5 others) On the relationships between DAGs and PO see: * Judea Pearl Interview by David Hand (on blog):

8.7.2020 8:27am - (Replying to @PWGTennant @frankdevocht and 6 others) "quasi-experiments" are just observational studies with supposedly credible knowledge about the causal relationships among certain variables.

8.7.2020 12:36am - (Replying to @l_gervasi) Curious what sort of "causal model" you have in mind, that would allow you to decide which of two sets of measurements you should send your patient to.

8.6.2020 11:29pm - A prerequisite for good night driving is knowing how to drive at day time, then compensate for darkness and other agnostic impediments. Using darkness as justification for quitting driving school is awkward.

8.6.2020 10:24pm - A very disturbing incident at USC What's more disturbing, the resigning student complains about "anti-Zionism" (8 times in her letter) and the President of USC apologizes for "antisemitism"-- the easy way out and the obvious way to ensure zero result.

8.6.2020 8:37pm - Controlling for a "broader set" makes sense when measurments have zero cost. But in most real life situations one cannot afford to control for EVERYHING that can possibly be measured and, then, cost-sufficiency-efficiency considerations demand some domain models, namely DAGs.

8.6.2020 6:03pm - And it is anti-Zionism (ie Zionophobia) that is becoming the ugliest form of thuggery on US campuses:, in the best tradition of morally bankrupt ideologies.

8.6.2020 5:18pm - We had a tweeter discussion on whether it is ethical, given that reviewers assume a limited audience. I believe the consensus was: it is ethical. I further suggested that historians will find this material more representative of contemporary science than the paper itself.

8.6.2020 4:42pm - If by "out of sample" one means "under data shift" or "under process shift" than a charitable interpretation can be constructed. Otherwise, @_MiguelHernan is right; the Ladder of Causation has its laws, and "nobody is perfect."

8.6.2020 4:31pm - (Replying to @JPedro_Monteiro @NateSilver538 and @mattyglesias) @NateSilver538 is a great predictor; I don't want to embarrass him when he is wrong. It ain't fair, if he has not read #Bookofwhy.

8.6.2020 4:04pm - I love this article @ShMMor , with one exception -- it does not name the offender. "Anti-Zionism" sounds like some sort of a benign political opinion, socialism, Maoism, anti-liberalism, etc. Zionophobia better expresses the racist and moral deformity of the Anti-Zionist agenda.

8.6.2020 9:54am - (Replying to @ScienceandLove) Why should we "put things philosophically" when we can "put things computationally"? Philosophers are moving from hand waving to computational models, why should we go backward?

8.6.2020 1:36am - Sharing a video of my talk at the CI-FAR workshop on causality: "Contesting the Soul of Data Science"
Apologizing for the occasional mis-synchronization with the slides (which I could not see in LA), due to a mischievous Zoom problem.

8.5.2020 9:54pm - Sharing an uplifting project: Mark O'connor Summer Music Camp, is one of the most meaningful memorial to our son Daniel, his love for music and his hopes for humanity.

8.5.2020 8:02pm - One unexpected consequence of Covid-19 is the unprecedented rise of public interest in reasoning under uncertainty. Here is another NYT story on Bayesian stat: . Their next story is, perhaps, "Why I am only Half Bayesian"

8.5.2020 3:08pm - Remember my post on machine learning and whether all knowledge comes from sense data? Well, in the wake of the CI-FAR workshop on causality, Yoshua Bengio sent me a couple of comments which I have since appended to the post. Enjoy, and join the dance.

8.5.2020 4:55am - (1/ ) Following our (re)discovery of the beautiful paper by White and Lu, here is another forgotten paper, this time for ML folks interested in mining "invariants", as in recent works of Y. Bengio. The forgotten 2001 paper, "Causal Discovery From Changes"
8.5.2020 4:55am - (2/ ) uses "shocks", or spontaneous local changes in the environment ("Nature's interventions") to infer causal directionality among variables. As is common in CI analyses, the paper unveils those features of the world that would "guarantee" correct discovery and those that won't.
8.5.2020 4:55am - (3/3) A more complete and general analysis of this problem is provided by Jaber etal's "Causal Discovery from Soft Interventions with Unknown Targets" (2020) Again, the emphasis is on the theoretical distinction between the "doable" and the "undoable".

8.5.2020 2:36am - (Replying to @IgnaGB26) None in mainstream Stat. Which is both a shame and an opportunity for an author to be "the first". But Primer would easily be handled by stat students/instructor. See

8.4.2020 6:55am - Thank you, America! Exactly sixty years ago, August 4, 1960, me and my wife landed in NYC as graduate students, seeking education and something meaningful to do in the land of unlimited possibilities. Today, looking back, we are grateful for the opportunity. Thank you, America.

8.4.2020 6:51am - (Replying to @eddericu) Yes, see Causality, page 81-82.

8.3.2020 6:34pm - (Replying to @orlipeter @ConceptualJames and @kareem_carr) Sorry, I took @kareem_carr post to be about science (I still think it was) and I suddenly find myself embroiled in an anti-culture war. I hope no one accuses me of taking sides.

8.3.2020 7:06am - Nice discussion and illustration of c-component identification strategy.

8.3.2020 4:20am - (Replying to @jenhelenmar and @trishgreenhalgh) I never understood what "evidence based" medicine is all about, so I feel unqualified to participate in this debate. Perhaps it would jolt "evidence based" philosophers (eg. Cartwright et al) to define what it is in causal terms.

8.2.2020 9:56pm - (Replying to @Isaac_Herzog and @Sethrogen) I don't buy it. @Sethrogen has all the microphones in the world, 10 times more than he had before making that idiotic "jest". He can stand up like a mench and tell the microphones what Israel really means to him. I have not heard him yet. I am listening.

8.2.2020 5:01pm - (Replying to @RaulMachadoG) Attempts? I know of only one attempt: causal diagrams. Would love to hear of alternatives.

8.2.2020 4:58pm - Every "approach" that is based on Rubin's potential outcome "framework" is guilty of bypassing graph construction and of mixing right and wrong. see and and In SCM, non-linearity is assumed a-priori.

8.2.2020 2:46pm - (Replying to @Jabaluck @causalinf and 8 others) The W&L paper cited is saying, if we commit to a class of DAGs with certain features, we get improved efficiency. To communicate what those features are, requires DAG language, eg. drivers, proxies, colliders, chains, etc. Often, a single dag suffices to encode those features.

8.2.2020 11:56am - (1/ ) It is for these reasons I consider myself lucky to be in computer science, where math is regarded as a man-made language in the service of thought, not the other way around. When G. Boole wanted math to serve logic he decide that 1+1=1. When S. Wright wanted math to serve
8.2.2020 11:56am - (2/2) causality, he invented diagrams. If you are still a statistician, ask yourself when the last time was that statistics has switched languages to cope with new type of problems.

8.2.2020 11:34am - (Replying to @Jabaluck @causalinf and 8 others) Sure, observe all the models shown in:; each summarizes trillions of models sharing certain structural features. The point is that you won't know whether OLS delivers causal effects unless you assume those structural features to hold.

8.2.2020 7:00am - I thank @eliasbareinboim for sharing his talk's video at Microsoft Research: . It explains the fundamental challenge of CI in a slightly different way from the way I (and the #Bookofwhy) present the Ladder of Causation.

8.2.2020 6:33am - (1/ ) Dearth of Econ examples could not possibly be the obstacle. If by "traditional means" you mean: "assuming conditional ignorability" then there were hundreds of Econ examples that started that way and ended of course with the perennial bewilderment whether the assumption holds
8.2.2020 6:33am - (2/2) for the selected covariates. Each of these examples could have benefited from the transparency of graphical models, especially in regard to efficiency considerations. (See "A Crash Course in Good and Bad Controls" So what was & still is the obstacle?

8.2.2020 3:21am - Sure. This paper by S. Greenland, J. Pearl, and J.M. Robins, "Causal Diagrams for Epidemiologic Research," has inspired 2638 citations, compared with 41 citations of White and Lu (2011). The shrill silence of dogma.

8.1.2020 7:19pm - (1/ ) I've just (re)discovered a beautiful paper by White and Lu It gives economists a simple introduction to graphical models plus compelling arguments why they're useful, if not necessary. My question to enlightened economists: How come this paper all but
8.1.2020 7:19pm - (2/ ) disappeared from the econometrics literature (only 41 citations). How come it failed to convince econometrics leadership, from Heckman to Angrist, that the benefit of seeing your assumptions crystal clear far outweighs allegiance to traditions that keep those assumptions
8.1.2020 7:19pm - (3/3) cryptic and indefensible. How come? @causalinf , @PHuenermund , @lewbel , @Susan_Athey @pedrohcgs @snavarrol @TymonSloczynski @tdetonberry @MacroPru . The question is interesting because, in other fields (eg. epi) it took only one paper to convince both leaders and masses.

8.1.2020 6:07pm - (Replying to @BenOgorek) The previews give away (essentially) the entire book. The only reason we labeled them "previews" is to keep the publisher from suing us. (Too old to sit in jail).

8.1.2020 5:22pm - I can't praise this book more than I already did but, having just read how economists introduce causality to college students, I am going to praise this book one more time.

8.1.2020 5:14pm - (Replying to @LahavHarkov and @Sethrogen) Moreover, see the facts about what Israelis knew about the Arab population in the land:

8.1.2020 5:11pm - (Replying to @benshapiro and @Sethrogen) Moreover, see He who has been fed lies all his life and swallowed them illiterately is hardly in a position to wake up one day and teach others how to tell lies from falsehood.

8.1.2020 4:57pm - (Replying to @davidharsanyi) Moreover, see

7.31.2020 9:05am - (Replying to @ScienceandLove) I do not see a distinction between "useful intuitive pump" and the way "human think about causality" and, moreover, I do not think it is relevant to ask how causal forces "really" act in the world. As a student of AI, capturing how human so it is already a great achievement.

7.31.2020 5:23am - (1/ ) It occurred to me that Seth Rogen's idiotic remarks about the establishment of Israel ("racist endeavor") are not totally his fault. This chapter of Jewish history has not been part of Jewish education in America. We should correct for it by revisiting
7.31.2020 5:23am - (2/2) (Replying to @yudapearl) and citing and some of those remarkable quotes by Ben Gurion, Weizmann, and Jabotinsky, to set the record straight. @MiriamElman @WarpedMirrorPMB @MattiFriedman @rabbisacks @EinatWilf @HenMazzig @AndrewPessin @ayidindixieland @GilTroy @drdivine @deborahlipstadt @marksjo1)

7.31.2020 4:08am - Eine Kleine Nachtmusic

7.31.2020 3:03am - A must-read for understanding the New Religion and its holy scripture.

7.31.2020 2:30am - (Replying to @mfkuepp and @eliowa) I recommend reading it in English, it rhymes like Mozart.

7.31.2020 1:57am - (Replying to @clovesgtx) Your class notes strike me as an upside down way of teaching causal inference. Why not start with structural equations then show how potential outcomes can be derived organically from those models, as is done here Otherwise, what's the role of graphs?

7.31.2020 1:45am - My readings into the phrase "appropriate study design" made me more cautious of the phrase than of what it intends to exclude.

7.30.2020 1:24am - This favorite lie repeated by Zionophobes, that early Zionists ignored the (equally) indigenous Arab population, prompted me to do some research on the topic. The results are published here: We can forgive Rogen's ignorance, but who were his teachers?

7.29.2020 10:59pm - Rogen is envious of all the media attention Peter Beinart got for saying: "Who needs Israel". Two Jews of discomfort: "Mother, I hate you for make me feel so inadequate. The boys in school called you names, and I couldn't defend you, so I joined them"see

7.29.2020 8:54pm - The greatest lie Rogen tells us is that he was fed lies about absence of Arabs on the land. Ben Gurion's first paper (1917)was a demographic study of those Arabs, his first book: "We and our neighbors (1933)". From kindergarten to college Israelis are told "our future neighbors"

7.29.2020 8:54pm - Seth Rogen: 'I Was Fed a Huge Amount of Lies About Israel My Entire Life' via @jewishjournal

7.29.2020 3:06pm - (Replying to @ThatMarkElliott) I did not realize registration is closed. Will check with the organizers about recording.

7.29.2020 2:03pm - Tomorrow, 9am PST, I will be giving a virtual talk at the CI-FAR workshop on causal inference, My title is almost the same as here But I will spent the 1st 10 minutes discussing Radical Empiricism

7.29.2020 12:22pm - Tomorrow, July 30 is the 9th of Av, the most calamitous day in the Jewish calendar. Please take a look at the golden candelabrum, or Menorah, carved in deep relief, top of this post: I never pass through Rome without paying a visit to Arch of Titus. Why?
7.29.2020 12:22pm - (2/ ) Because it never fails to dazzle me with the profound irony of history: A monument erected to commemorate the eternal end of Jewish independence turning into the most explicit and indisputable testimony of that independence, and the most inspiring symbol of its revival. I ...
7.29.2020 12:22pm - (3/3) I often catch myself wanting to say: Thank you Titus, only to recall that, after all, he was the destroyer of our 2nd Temple.

7.29.2020 11:48am - Kudos to @themuslimreform for joining the @Twitter Walkout. This Wiley gives me an unwanted superiority complex (Jewish privilege?) for having been protected from the cesspools of hate and ignorance in which he must have grown up.

7.29.2020 10:36am - Note that the courts upheld: "the ordinary meaning of 'because of' is 'by reason of' or 'on account of,' ... That term incorporates the but-for causation standard, formally defined using counterfactual logic, as in #Bookofwhy

7.29.2020 10:24am - Glad to be back with readers, having just ended our #48HoursSilence walkout to protest @Twitter role in amplifying the racist voice of Wiley the Conspiracy Maker. True, I usually react by walk-in, not walkout, but, in this case being counted was important.

7.27.2020 10:51am - In solidarity with our readers in UK, Jews and non-Jews, artists, authors and civic leaders, I am joining their call for a 48 hours Twitter Walkout I find it inconceivable that we, scientists, will remain unmoved, distant onlookers.

7.27.2020 10:22am - BBC News - Wiley: Anti-Semitism row prompts 48-hour Twitter boycott

7.27.2020 2:39 (Replying to @yudapearl and @skjask) ... while the bridegroom denies the bride was born.

7.27.2020 2:15am - (Replying to @zacharylipton) Some of them have escaped through graphical models to prepare the grounds for next generation ML:

7.27.2020 2:04am - (Replying to @skjask) Three arm-chair match-makers discussing the menu for the wedding, without asking the bride if she agrees to have one.

7.27.2020 1:47am - The Book of Why has a whole section on it, including the Kruskal-Bikel exchange, reflecting on its lessons. But no discussion of Simpson's paradox is complete without understanding its resolution and its post-resolution longevity:

7.26.2020 7:17pm - The example has a variant with a painfully bitter end, concluding with inevitable regret, and irredeemable remorse. See

7.26.2020 6:19pm - I understand that @smueller taught a causal inference class in high school (using Primer). He may have class notes to share.

7.26.2020 5:42pm - If all knowledge comes from sense data, why don't we just gain knowledge the way our ancestors did, by simulating natural selection on our data? Or the way ML folks do, by fitting functions to data? I tried to answer both questions in a new blog post:

7.26.2020 2:15pm - Rung-3 comprises inferences that require undoing past events, eg, " if only my plane had left on time..." It is explained superbly here:

7.26.2020 5:00am - I am posting this paper "Einstein's Greatest Legacy, Thought Experiments" because thought-experiments are acts of imagination, unique to Rung-3 of the Ladder of Causation, that are a necessary ingredient of any automated scientist.

7.25.2020 5:16pm - A new @JewishJournal story on Carnegie's "Great Immigrants" Award and on how Israeli-American immigrants can save American Jewry:

7.24.2020 10:24am - It's not enough for hundreds, even thousands professors to prove Zionism the most noble and inspiring experiment in human history; the crocodiles will continue to demand their daily meal, until the village elders say: "What an ugly and dangerous animal! No more!"

7.24.2020 9:07am - Thanks, Cyrus, for excavating this ancient 1987 piece from the archives of embryonic ideas. Reminds me of the days when we were truly wondering: Who needs graph? Some people are still wondering, amazing!

7.24.2020 1:35am - (Replying to @rorykoehler) The definition is from the dictionary, but applying it to people with whom you disagree turns you into a bigot, God forbid.

7.24.2020 12:33am - (Replying to @rorykoehler) Please note, your definition of "bigotry" does not express much warmth toward other opinions, say toward those who do find positive sides in certain forms of nationalism. One positive side, eg. is the cultivation of knowledge and experience transmitted across generations.

7.23.2020 12:46pm - (1/2) In solidarity with my friend, Professor Miriam Elman, who was attacked for being "openly Zionist": I would like to openly state that, in addition to being "openly Zionist" I am also openly disgusted with those who are openly Zionophobic and,...
7.23.2020 12:46pm - (2/2) as long as I live, I will try my best to openly expose their bigotry, hypocrisy and moral deformity.

7.23.2020 11:15pm - (Replying to @YablokoU) Zionophobia is treating Zionism as anything less than normative and organic. If you question the fitness of an openly straight person to teach a class on marital psychology then, yes, I will question your motives.

7.23.2020 10:59pm - (Replying to @YablokoU) Zionophobia is a moral deformity in that it denies one people the right to a homeland. Beside, I've never seen a Native American or a Canadian patriot being questioned for fitness to teach history.

7.23.2020 10:40pm - A surprised reader asks: Did you really mean to imply that Potential Outcome folks who do not know graphs do not know potential outcomes? Yes, I meant it. And I have explicated precisely what properties of potential outcome those folks are missing, e.g.,

7.23.2020 4:22pm - (Replying to @aminkarbasi) What permission do you seek?

7.23.2020 3:57pm - I don't know anyone knowledgeable in causal graphical models who is not also familiar with potential outcomes. The latter are derived from the former in 3-steps (see, Section 4.2.4) which tell us what potential outcomes are all about.

7.23.2020 3:04pm - (Replying to @rorykoehler) The positive side of it, yes!

7.23.2020 2:15am - (Replying to @itamarcaspi and @HebAcademy) Behatslacha!

7.23.2020 12:00am - (Replying to @rohan_virani) Hard to comment w/o seeing the transcript. But, off hand, a monkey too learns by playful manipulations of objects yet, unlike a baby, the monkey never grows up to invent a sowing machine. Why? Isn't it because it lacks a template to store its experience? i.e., a causal model.

7.22.2020 1:21pm - A breathtaking discovery in Jerusalem: World-orders are crumbling, new viruses mutating, yet nations are busy unearthing their historical infancy and, at times, the earth is generous -- I have known King Menashe and Hezekiah since 3rd grade (Kings 21:1).

7.22.2020 10:44am - (Replying to @itamarcaspi and @RazRaanan) How about "Bli-Avad", al mishkal "BeDi-Avad". But its more than just "bli", its "Negdi-Avad". My grammar teacher would have flipped, poor Dr. Gotshtein.

7.21.2020 12:53pm - What a beautiful satellite photo of Israel, with patches of lights marking Tel-Aviv, Jerusalem, Haifa, Amman and Cairo. Even Cyprus and Turkey can be seen in the distance, and the Sinai Peninsula amazes you again why it took 40 years for the Israelites to reach the holy land.

7.21.2020 12:19pm - Wonderful post about missingness-graphs, agree! How come we missed it thus far? Does it means missing-data folks are beginning to ask WHY, like the rest of us, and demand to understand what their assumptions mean, and what can be recovered from missing data? End to hopelessness!

7.21.2020 6:05pm - (Replying to @GaryMarcus) Thanks for re-posting, Gary. This came out July 4th, but it's always good to be reminded that I represent an ethnic minority of Israeli-born immigrants who are grateful for the opportunity to share our unique experience and add our distinctive color to the American tapestry.

7.21.2020 2:51pm - (Replying to @itamarcaspi) Eineni Yodea. An interesting question for Vaad Halashon.

7.21.2020 8:18am - How do we know it is a matter of kind, not of degree? Because we can't even write the question in the vocabulary of "what if I do?". Try writing formally: "What is the probability that the dead victim would be alive had the defender not shot". See

7.21.2020 8:08am - (Replying to @BenHanowell @PHuenermund and @eliasbareinboim) Think about causes of effects at the individual level. Was it my aspiring that caused my headache to go away, today, at 3pm, after I watched that movie, and listened to Mozart's serenade? What if my plane had arrived on time, now that my partner is angry and the deal is off?

7.21.2020 7:31am - This paper illuminates the hierarchy that I once drew in the sand with several new lights, more mature and compelling, making me ask again, this time more humbly: How can any data scientist ignore its ramifications?

7.20.2020 10:17am - One thing publishers don't tell you, which is probably more important than the cover or the price: This edition will have all the errata corrected, so you can run through it with no hurdles. Enjoy the run!

7.20.2020 7:48am - [Sorry for the broken tweets.] Its value is explicated here: and elsewhere. However, research questions handled by Tools 3-7 are outside the vocabulary of most "Data Science" folks today, and so are these questions:

7.20.2020 7:41am - Retweeting. I have explicated this value here: 3-7 are currently outside the vocabulary of most Data Science folks, and so are these:

7.20.2020 7:29am - The value of Data Science is obvious to me too, even under its current data-centric direction. I have explicated this value here:

7.20.2020 4:48am - For readers who, like me, grew up in the information age, here is a fascinating public lecture about the life and works of Claude Shannon, the man who brought information age to where it is today: AI folks should also recall his 1948 chess-playing program

7.19.2020 9:07am - (Replying to @willcfleshman) Why would answering questions at the bottom of the ladder be appropriate in your research question is about a higher rung of the ladder?

7.19.2020 8:59am - (Replying to @ShalitUri) The difference between "people w/diabetes" and "entire med. record" seems a matter of degree, whereas "group-level" vs. "individual-level" is a qualitative difference. If so, I am not entirely off claiming (in that the latter requires counterfactual logic

7.19.2020 2:33am - (Replying to @lukasvermeer and @learnfromerror) I would be delighted.

7.19.2020 2:24am - Alchemists had one advantage over data scientists, they could recognize gold if/when they were to produce one. Ask a "data scientist" if her latest neural net answers her scientific question. The answer I often hear: "What question?" This is why I contest the SOUL of Data Science

7.19.2020 1:18am - (Replying to @lukasvermeer and @learnfromerror) Your Goto 2016 talk is brilliant. Is there a transcript or a summary? I would love to taste more of that historical Sauerkraut. It reminds me of a parable I often use: "It is hard to find a needle in a haystack, it is almost impossible if you've never seen one."

7.18.2020 11:19pm - (Replying to @learnfromerror) What data science people aspire to is "Data Intensive Science", which is a lot different from traditional, pre-big-data "science". What they actually practice is a "Science of Summarizing Data" which is hardly a "science". See

7.18.2020 11:01pm - (Replying to @data4sci) I've always questioned the purpose of "working with data". Even when I first heard about "taking the average" I asked "WHY"? Why not go with the last sample? And only when told that the mean gives you a better prediction I was (partially) satisfied. (Proved in 1756).

7.18.2020 5:02pm - I like the name "Data for Science", which counters the current direction of "Data Science", as if Data in themselves are the target of scientific exploration.

7.18.2020 4:05pm - (Replying to @tytung2020) I will look into the alternative explanation of the horrific video that I've shared here. But please do not teach me about Guantanamo - the confessed murderer of my son sits there, still boasting in his accomplishment.

7.18.2020 3:55pm - (Replying to @CasualBrady) It is the same concept, yes. We tried to name the specificity by the characteristics assumed, or the variables which are measured.

7.18.2020 3:51pm - (Replying to @stephensenn) Regression slopes are still population-level properties. There is not "regression slope" for Mr. Jones' reaction to drug D, taken Wednesday 1 pm. Counterfactual logic does tell you something about this singular reaction, see

7.18.2020 3:40pm - A thrilling chapter in history - "20 years ago today": the Camp David Summit. An interesting story from an eye witness, encapsulating the core of the core of the Israeli-Palestinian conflict: MUTUAL ACCEPTANCE of historical legitimacy.

7.18.2020 9:46am - (Replying to @CasualBrady) "c" stands for "characteristics". That where the "c" came from originally (Shpitser and Pearl 200?). But you have a point.

7.18.2020 8:54am - (Replying to @nagpalchirag) Thanks for posting. I like the fact that you do not pose "ignorability" as an apriori axiom, but show how it is derivable from the model of Fig. 2. I recommend that you state explicitly what type of assumptions allow you to estimate effects, given the Z is latent.

7.18.2020 7:11am - Readers noted that my Moment Magazine piece on 1948 is chopped prematurely, and the part where I urge every holocaust museum to dedicate a "From the Ashes" exhibit is hidden away. Here is the hidden part:

7.18.2020 6:41am - Readers ask: How can you claim (in that going from population data to individual level requires counterfactual logic, when ML folks estimate Individual Treatment Effects (ITE) straight from group data (eg Answer: What they
7.18.2020 6:41am - call ITE is not really "individual" but average effects in a specific group that shares the measured characteristics of an individual. We call this average "c-specific effect." Individual-level effects indeed require counterfactual analysis, as is done in

7.18.2020 6:11am - Thanking @eliasbareinboim for sharing his illuminating slides and video on Causal Reinforcement Learning, and his methods of combining the unbiasedness guarantees of Causal Inference with the sampling efficiency of Reinforcement Learning.

7.17.2020 12:51am - Photos of roundups, ordinary folks, in ordinary civilian cloths, carrying ordinary household items, clinging to ordinary logic, are sometimes more horrifying than those of inmates in death camps. Watch the 1942 roundup of Paris Jews, 78 years ago, today:

7.16.2020 10:58pm - I tried to skip this footage, and move on to AI, ML and CI, but my fingers wouldn't let me. They force me to retweet, as wide as I can, because the appointed guardian of these brutalized victims, the UN Human Right Council won't - it's busy condemning Israel for being alive.

7.16.2020 9:19pm - (Replying to @genomixgmailcom) And I can't wait to hear how a hard-core regressionist reacts to my heretical conclusions.

7.16.2020 11:31am - As scientists, it behooves us to do more than point out how absurd the UNHRC has become, and how it stains the reputation UN as a whole. We have psychiatrists and social workers in our ranks, so, don't these delegates and their governments have some sense of right and wrong?

7.15.2020 8:38pm - My favorite beach. That's were I learned to swim, 80 years ago. I've changed somewhat, the waves havn't. I became impatient, the waves didn't. Look at them, endlessly trying to dare the beach into submission - I've given up.

7.15.2020 8:16pm - (Replying to @Claire_Voltaire) In my neighborhood Zionophobes use it more than any other group.

7.15.2020 9:34am - Sharing personal experience from 1948, the year when everything changed, the magnitude of which we, 11-year-olds could not begin to imagine.

7.15.2020 8:40am - (Replying to @michaeldickson @michaelduckson) , when was the last time you used the word "Zionophobia"? The only fighting word left to us, and our leaders and spokespersons are so careful not to use it.

7.15.2020 8:13am - Sharing a fascinating interview with Eran Zaidel - How Brain Scientists Think about Consciousness Closer to Truth, 13 July 2020 Eran Zaidel is a UCLA colleague, another former Israeli who came to this country on an international student visa.

7.15.2020 5:38am - (Replying to @jwbelmon) Funders are rarely evaluated, hence rarely criticized. Funders shield themselves from criticism by re-funding the old guard of PI's, to do more of the same, often under the hype of "new initiatives".

7.15.2020 4:32am - I believe the symbiotic relations between ML and CI is defined by the inference engine of as well as in the more general contest on the soul of data-science:

7.15.2020 4:12am - (Replying to @renatrigiorese) I think the "IN vs. OUT" partitioning can be viewed as compliant with Dennett's intentional stance. It is appealing to me because it is simpler and more operational.

7.15.2020 4:01am - A reader asks: "Some colleagues are insinuating that you subscribe to the probability interpretation of causation from the logic of your do-calculus..." God forbid! The probability interpretation of causation was an unfortunate blunder, discussed here:

7.14.2020 1:13pm - As I read her words: "Twitter has become the ultimate editor," I was thinking, to what extent this is true in our truth-based scientific bubble? To what extent signaling allegiance to a guru, editor or colleagues dictate how we phrase our findings or equations? Still thinking...

7.14.2020 12:54pm - I was shocked this morning to read Bari Weiss' resignation letter (from NYT): I have known Bari since her student's day at Columbia when she bravely stood up for students rights, and she has been on my hero's list since. Freedom has lost a real fighter.

7.14.2020 12:12pm - (Replying to @metadiogenes) The other way around. My issue with entropic formulations is that they DO NOT partition the world, but try to explain agent-asymmetries in physics itself, not in the observer.

7.14.2020 8:20am - (Replying to @metadiogenes) I am not familiar with this formulation, thanks for pointing, but my skepticism to any entropic formulation continues to dim my enthusiasm.

7.14.2020 7:11am - The reason I feel uncomfortable with the answers is that thermodynamic-based features apply equally well to a thermostat, a device we all agree has NO agency. The account I've proposed to explain these asymmetries is based on our habit of partitioning the world into IN vs. OUT
7.14.2020 7:11am - namely, a chunk of the world we provisionally consider the focus of our attention vs. the rest, which we consider "external" to the former (see p. 420 of Epilogue ) This arbitrary choice does not show up in the equations of physics.

7.14.2020 4:19am - Another #JewishPrivelege I cherish: Being bought up in a culture that looks at facts and exercises reason before falling for juvenile conspiratorial theories that Cannon evidently takes very seriously. What a mighty privilege!

7.14.2020 3:33am - This choir of Japan's House of Peace is singing in Hebrew: "When the heart cries". The choir apologizes for having to do it through youtube instead of their usual world-wide tour. "When the heart cries, God does listen".

7.14.2020 3:11am - Apologizing for the occasional mis-syncronization between my voice and my slides which I could not observe in LA, due to a Zoom Dragon acting up over the pacific ocean. But the message is still there: Data-Science should shift its focus from data to science.

7.14.2020 1:09am - (Replying to @MasahikoKim) Given your audience, I believe your team did a great job in jolting the bio-health informatics community to think causally, to leverage ML methods, and to ask themselves which rung of the ladder they wish to climb.

7.13.2020 5:12pm - I always welcome "computational models" in Journals like Congnitive Science but I can't see the "model" when authors refrain from using counterfactual notation to define "sufficient cause" as in or

7.13.2020 6:27am - What is an AGENT? The questions raised in this paper deserve discussion: (i) what makes a physical system an agent; (ii) the reason for agency's time orientation; (iii) the source of the information generated in choosing an action. Unsure about the answers

7.13.2020 1:07am - My personal #JewishPrivilege is that my parents concealed the holocaust from me till I was old enough to stomach it and, this way, ushered me and my generation to grow up into normalcy rather than victim-hood -- what a privilege!!!

7.13.2020 12:26am - I wish I could attend this ICML tutorial today (Monday) on causal reinforcement learning: -- I've always wondered why RL folks don't sit together with CI folks to create a synergistic ML agenda. Webpage:

7.12.2020 2:33am - (Replying to @WiringTheBrain and @DrYohanJohn) If we ask: Why is experience in the world so important for intelligence? How is it encoded in the mind after it is no longer experienced? How it is accessed, etc. we get closer to replacing that experience with a surrogate program, thus rendering that experience unnecessary.

7.11.2020 7:04pm - (Replying to @latinostats) Luckily, human intelligence refused to listen to philosophers speaking and has evolved beautifully in grey organic substance called "brain". Now on its way to experiment with silicons.

7.11.2020 4:48pm - The article is worth reading, I agree. I cannot share however its enthusiasm for, and resurrection of Dreyfus arguments which, IMO, aim to mystify, rather than understand human intelligence. AGI will be achieved by conquering one mini-Turing-test at a time, not by quitting.

7.11.2020 3:48am - (1/2) Now that Beinart has declared himself a "Zionist", thus stripping Jews of the one word that defines their historical identity, let me assure readers that the word Zionophobia still shines in its original clarity: "Denying Jews the right to a homeland of their own." So,
7.11.2020 3:48am - (2/2) So, Beinart may call himself a "Zionist" or a Giraffe, if he wants; the bigotry of Zionophobia will continue to haunt him unambiguously, with all its ugliness.

7.10.2020 11:19am - Jayathi Murphy, the Dean of our school (UCLA, Engineering) has authored a powerful oped on behalf of International Students. I am proud to share it on this educational channel.

7.10.2020 10:20am - (Replying to @daniel_bilar and @ShlomoArgamon) Thanks for clarifying.

7.10.2020 10:19am - (Replying to @Rondo2 and @kweansmom) Beg to differ. Words mean an awful lot, if you teach those words to your children, and reinforce them in schools and mosques.

7.10.2020 9:26am - (Replying to @Rondo2 and @kweansmom) They can easily prove that they are different: Acknowledge some historic Jewish connection to the land. I am waiting.

7.10.2020 8:59am - (Replying to @BlakeFlayton) The term "pro-Israel" is thrown around because we, Israelis and Zionists, have allowed anti-Israelism to become a debatable item, rather than mental disorder. Same with "proud-Zionist"; I've never met a Zionist who was less than proud.

7.10.2020 7:24am - Beinart's first victory = Linda Sarsour! A Zionophobe turned "Zionist".

7.10.2020 7:10am - Any Albanians among our readers? I salute you for your bravery and sacrifice during WW-2! In general, I am not too fond of "holocaust memorials". The best monument to Jewish survival is the rebirth of Israel, which is a living monument that breathes, creates and inspires.

7.10.2020 5:53am (Replying to @devon_l_sch) Happy Friday, Devon. I am familiar with the diagnostic system of Bayesia, and the contact I have is: . I hope it works for you.

7.9.2020 11:43pm - (Replying to @prem_k) Do you believe there is merit to this critics? NonParametric causal inference starts with "if then" specification of what we know (DAGs) and ends with "how much" answers to what we want to know (eg ACE). Gelman missed it, because he refuses to solve a toy problem start-to-end.

7.9.2020 5:02pm - (Replying to @ekirwin8 and @Larry_Svenson) Why get your head spinning in all those so called "frameworks". Just solve ONE toy problem, from beginning to end, and you are ahead of 90% of the spinners.

7.9.2020 3:23pm - (Replying to @kareem_carr) It is easy to agree with me (I am fairly agreeable person) The tough thing is to agree on the mathematics of causation. Do you think your fellow statisticians understand the two fundamental laws of causal inference, or, at the very least, the 3 rungs of the ladder of causation?

7.9.2020 2:11pm - Glad you brought up Gelman's "sort of" review. I always recommend it colleagues who ask about causal thinking among leading statisticians in the 2nd decade of the 21st century. But, I reiterate, one toy example teaches us more about causation than 100 debates among the gurus.

7.9.2020 7:17am - Do We Really Have to Read Beinart Again? via @jewishjournal

7.9.2020 7:17am - Another take I recommend about Beinart's heart-breaking plea for social acceptance.

7.9.2020 6:48am - What are "Jews of discomfort"? What is "indigenous"? Two thoughts pertaining to Peter Beinart's surrender to thoughtlessness.

7.9.2020 5:06am - What part of data analysis is or is not causal is not up to @yudapearl , but comes out of the mathematics of causality. See interview with David Hand on this issue:; tasks believed to require no causal thinking turned out to benefit from it. @kareem_carr

7.8.2020 4:40pm - The famous plea: "Give me Yavne and its scholars" was made when Jerusalem was in flames. @PeterBeinhart distorts it to wet the appetite of Jerusalem's enemies, while she is a thriving center of art & science, and a bastion of moral clarity. Beinart is a catalyst of destruction.

7.8.2020 4:11am - (Replying to @kenanalytics) What's the point? No one doubts the antiquity of Australian natives' culture. This does not negate however the Jewish quest for an "equally indigenous" status, a status denied them by "sole ownership" neighbors.

7.8.2020 2:53am - Another tiny difference: British settlers did not see Australia as the cradle of their civilization, Jews did. Having celebrated homecoming for eighty generations in poetry, prose, lore, holidays, and daily prayers, they saw themselves equally indigenous to the land.

7.8.2020 2:10am - Contesting the Soul of Data-Science. Below, an introduction to the Spanish translation of #Bookofwhy: It expresses my conviction that the data-fitting direction taken by “Data Science” is temporary, soon to be replaced by "Data-intensive Causal Science"

7.8.2020 12:56am - This Foodbenders Lady has taught Toronto a lesson in propaganda. Decent Torontonians should reciprocate: "Some anti-Zionists ARE Welcome in our stores, if they admit they know nothing about the conflict, and haven't realized how genocidal and racist anti-Zionism is."

7.8.2020 12:33am - Foodbenders Owner Backtracks, Now Claims Some Zionists ARE Welcome

7.7.2020 11:38pm - Congratulations from inspired readers remind me that we, ~500,000 Israeli-born Americans, make up an sizable ethnic group with a unique culture and a unique experience to share: Building a new nation (against all odds), now a world center of art, science and entrepreneurship.

7.7.2020 1:09pm - As promised, the Spanish translation of #Bookofwhy is finally out It gives me a tremendous pleasure to think that my playful thoughts will be read in the language of Cervantes, Goya and Picasso. So, read it slowly, and with a spec of forgiveness.

7.7.2020 11:00am - As one of the few survivors who participated in its inauguration (UCLA, Aug. 1985) I am proud to announce this year's Conference on Uncertainty in Artificial Intelligence (UAI) to be held FULLY VIRTUAL -- Aug 3-6, 2020. Details here

7.6.2020 9:51pm - "Race, COVID Mortality, and Simpson's Paradox." Dana Mackenzie, my co-author in #Bookofwhy, has discovered another case of Simpson's Paradox in COVID data, and has posted an interesting analysis of its implications here:

7.6.2020 12:34am - These are important papers that should be read by all ML folks who talk about "transfer learning," "domain adaptation," or "lifelong learning". I wonder if the Epi folks are aware of these general methods for "counterfactual predictions" @barbradickerman @_MiguelHernan

7.5.2020 11:29pm - The problem of external validity has many variants, depending of what data is available from each setting and how the settings differ from each other. Once we encode this information in a selection diagram, the solution pops up algorithmically, as in

7.5.2020 9:01pm - The problem of deploying one prediction model in a new settings seems to lie within the umbrella of problems called "external validity". Could selection diagrams be of use here, to denote the disparity between the two settings? eg. as in

7.5.2020 2:55am - A great post with which we can reflect perhaps on our educational and research institutions and how they manage to keep academics toeing the party line.

7.5.2020 1:22am - More on Simpson's paradox (SP). As Covid-19 stimulates greater public interest in data analysis, it also makes more people wonder about SP and its implications. Unfortunately, most published accounts of SP are flawed, for they skip its causal roots and causal logic, 1/3
see: Readers who wish to get straight to the point, should take a look at Fig. 6.6 (p. 212) of #Bookofwhy, and ask themselves two questions: (1) Why are you surprised? (2) What is the correct action? You will get a better understanding from this quick 2/3 7.5.2020 1:22am - see: Readers who wish to get straight to the point, should take a look at Fig. 6.6 (p. 212) of #Bookofwhy, and ask themselves two questions: (1) Why are you surprised? (2) What is the correct action? You will get a better understanding from this quick 2/3
7.5.2020 1:22am - exercise than from most explanations I can find in the literature. Same figure can also be seen here: (fig. 1.1 1.2, p.3-4), and requires no numerical tables, and no arithmetical operations to get to the main point. It is thus a "Critique of Pure Reason."

7.4.2020 11:38pm - I am echoing the beautiful words of Sarai, Miss Iraq, with whom I had the fortune of singing duet last year, who grew up in the same neighborhood as my wife, and who speaks for all freedom-seeking Muslim women, betrayed by their political "leaders" and hate-peddling activists.

7.4.2020 2:29am - RL vs. CBN. In Causality p. 24 a Causal Bayesian Network (CBN) is defined as a parsimonious representation of all possible interventions. This paper seems to reason in the space of interventions, not their CBN. Transparency is lost; what is gained?

7.4.2020 2:02am - (Replying to @EricTopol @CarnegieCorp and 2 others) Strange, when people label me an "immigrant", I feel mighty privileged, not as a member of an under-privileged under-represented minority. This is perhaps unique to US and Israel; two countries were differences are virtue. Happy 4th of July.

7.3.2020 12:14pm - This wikipedia entry on Savage's principle is OK, except that Jeffrey could not say "unaffected by the action". He did not have causal language and had to say "unrelated to the action" which is an overkill. See

7.3.2020 11:01am - (Replying to @AnnaPodolanczuk) B'Yeshimon U'Baarava we called this book in Hebrew. Then there were Pan Volodyovsky, Quo Vadis, The Deludge, By Fire and Sword. Wow, this guy really shaped our imagination. I ask my grandchildren if they read Scienkiewizc and I hardly get: "What?" Immigrants have more in common.

7.3.2020 10:37am - Great post! This is indeed a quote from the #Bookofwhy (p. 87) paraphrased and attributed to S. Karlin. Note, isn't it precisely what we are hearing from researchers locked into Big Data, ML, Deep Learning etc. "The answers are all in the data, including culture and evolution."

7.3.2020 6:46am - Video clips like this one make you pause and think, how many moments in history you've almost forgot and how many heroes of history you've barely ever known

7.3.2020 3:29am - My old Tweet may be relevant here: : "Who needs causality if all you want is prediction?". Ask any pollster what the secret is to good prediction and the answer will soaked in causal vocabulary. Why? Because integrating data from multiple sources (Fusion) is a causal exercise.

7.3.2020 3:02am - What a small world! Your picture is next to mine: and my mother is from Kielce, Poland, just 200km from your hometown, Kolonowskie. Moreover, I grew up with the books of Henryk Sienkiewicz on my lips, did you? Small world!

7.3.2020 2:35am - A slight correction. More dangerous than the untestability of the assumptions is their non-comprehensibility; e.g., the economist does not know if they are defensible or not, even given an unassailable economic theory. Try it on your economist colleague next door.

7.2.2020 11:56pm - My, My! And this was 15 years before Marilyn vos Savant's posed it in her column. So it was your father who first recognized that the problem can be turned into an intriguing guessing game. Please convey my admiration.

7.2.2020 12:07pm - Good news indeed. Easier sleep too. And thank you @AsraNomani for extracting this valuable information from the lawyers and the Press. May September bring us #justiceForDanielPearl

7.2.2020 9:36am - I could not agree more, this visualization is great for teaching. Plus, do not miss the sequel where it is applied to Covid-19 questions such as "patients in greater need"

7.2.2020 3:39am - (Replying to @riemannzeta) I tried to resonate to some of his ideas, but could not (though I liked his use of "Market Blanket", with which I am familiar). I think we are heading in opposite directions, Friston trying to philosophize computation while I'm trying to algorithmitize philosophy.

7.2.2020 2:06am - I have been hearing those same tired slogans since 1936, so where is the "rage"?

7.1.2020 9:09pm - Proud and privileged to be part of this creative group of immigrants, painting together the most colorful mural of the American journey.

7.1.2020 3:54am - (Replying to @s_kampakis) Amazing, this is the first Tweeted post with which I had no disagreement. Am I getting old? Wise? Forgiving?

7.1.2020 3:50am - In causal diagrams EVERY variable is understood to be a moderator, no need for special treatment. On the contrary, special consideration is given to NON-moderators, as in linear models. Can we tell them apart from data? See a beautiful, under-rated paper.

6.30.2020 11:09pm - (Replying to @neuro_data @RaiaHadsell and 5 others) Your manuscript promises to teach me all about "transfer learning" and why I was't able to do it from other manuscripts. My novice attempt to formulate the problem in statistical terms ended up with many open questions: Is there a non-empty interesection?

6.30.2020 2:49pm - (Replying to @desai_pratik) Whatever injustice is perceived by one side or another can be undone with one sentence: "Both peoples have the equal right to be masters of their fate in a state of their own in this land - a Jewish state, Israel, and an Arab state, Palestine.” Time for a new dawn.

6.30.2020 2:42pm - I did not know this 2013 picture with Carl Reiner existed. I will cherish it dearly. Thanks, and condolences to his family and to his many many admirers.

6.30.2020 6:18am - To our Irish readers: May you be served well by your new Government! May it be a new dawn! A new age of reason. Amen!

6.30.2020 6:08am - Tough decision for a leader: Getting his people a state of their own, on the one hand, and admitting that the past 72 years of denying Israel were a futile and unnecessary suffering, on the other. Tough decision.

6.29.2020 5:41am - Thanks you @AsraNomani and all readers who have been staying vigilant with us awaiting #justicefordanielpearl. May reason prevail in September.

6.29.2020 1:55am - I wish it was so easy but, today, saying: "I am against discrimination" automatically makes you an enemy of some people. E.g., The California legislature just voted to strike down the words "not discriminate" from its constitution. Orwellianism: You can't signal loyalty to anyone

6.29.2020 1:18am - Suppose it won't, then what? What if anti-Zionists come to truly love Israel-hating Jews like Neturey Kartha or Noam Chomsky, would that make them less ugly? Less dangerous? Would that offer identity-Jews any comfort? Would it make them more fit to fight injustice anywhere?

6.28.2020 11:53pm - Reiterating: Why does it matter if Zionophobia IS antisemitism or just another form of racism? The question is how genocidal it is in its aims and how deceptive it is in its tactics. To be effective, we must understand and target the unique characters of those aims and tactics.

6.28.2020 11:19pm - (Replying to @HenMazzig) This is just ain't true and insisting on this equation is weakening the fight against the more dangerous form of hate: Zionophobia. It's time that we quit the anti-Semitic debate and focus on the more pathological and more genocidal threat against the Jewish people - Zionophobia.

6.28.2020 5:00pm - It is not what one fails to consider that counts but the consequences of that failure. In the case of PO, what's eroded is the credibility of the conclusions; they might as well run regression. But PO sells itself as a framework for causal analysis, not regression analysis.

6.28.2020 6:43am - (Replying to @BachmannRudi @PHuenermund and 4 others) Sure. My next tweet explicates it:

6.28.2020 4:41am - I have my own theory on what happened to economics and why it has not escaped the PO darkness, but I was hoping to hear first from insiders who have had a chance to solve a problem or two both ways and compare. EG. @pedrohcgs @PHuenermund @causalinf @snavarrol @TymonSloczynski

6.27.2020 10:47pm - (Replying to @dlmillimet @pedrohcgs and 2 others) I have tried to learn from your sausage, but could not tell if you meant "how econ. sausage IS made" vs. "how econ. sausage SHOULD BE made". I am eager to see the "IS" and "SHOULD" applied to the toy example cited here:, so we can see every step.

6.27.2020 8:26pm - (Replying to @EnzoInman) Primer gets you straight into problem solving, using stories and examples (eg Causality spends more time on proofs and philosophy. The do-calculus, for example, is not used in Primer, nor do we compare DAGs with PO. But we you get curious to continue.

6.27.2020 1:48pm - (Replying to @Prof_Livengood) Statistics lack of uptake is easy to explain, given the purging of causality by two powerful founders, Pearson and Fisher: The real puzzle is economics, which pioneered some foundations of causality in the 1940-1970, then invaded by foreign ideologies...

6.27.2020 12:27pm - (Replying to @19Naranjito82) I cannot buy the "there is no right and wrong" argument, unless it comes from someone who tries the X->Y->Z example in and concludes: "no right and wrong". I have not met such a person yet. Economists avoid toy examples because they show the right & wrong.

6.27.2020 9:33am - I am waiting for someone to ask: How is it possible that a whole field of extremely smart people will remain in darkness for 2 decades, when flash lights are readily available? It is really hard to believe, but several explanations have been offered, none entirely satisfactory.

6.27.2020 8:59am - (Replying to @PHuenermund and @krisgulati) Agree! and highly recommended! Is it published/submitted now? Or waiting for an enlightened editor that recognizes merit?

6.27.2020 7:17am - (Replying to @krisgulati) Wow, you are in trouble! PO in 2020? Actually, you are already way ahead of your class (and your instructor). Just learn the translation to graphs and, whenever your professor blushes or hand waves, just help him/her out. Don't reveal how you did it, they won't appreciate it.

6.27.2020 7:09am - I have an example that can clarify my last Tweet day & night. Consider a simple causal chain X--->Y--->Z. Try to express it in PO notation, and answer some questions about it. It is done here, Eqs 6-8. Try it yourself, step by step, and see the darkness.

6.27.2020 6:10am - (Replying to @craigwpickett) If you want to learn to solve concrete problems start with Primer. If you want to philosophize and argue about who did what, when, how, and why not earlier, start with #Bookofwhy

6.27.2020 4:41am - A fair question, that comes up again and again. The darkness behind PO is the requirement to encode knowledge (read: assumptions) in a form that is cognitively formidable to the knowledge provider. See A toy example will speak louder than I can.

6.27.2020 1:55am - I will never get tired recommending Primer, while browsing through #Bookofwhy for fun. But be prepared for a traumatic paradigm shift upon learning that "causal statistics" is an oxymoron, like a "squared circle".

6.27.2020 1:28am - (1/2) Readers will enjoy seeing what happens when economists, bent on avoiding graphs, try to tackle data-fusion problems using PO vocabulary. The result is laid out here: and confirms what #Bookofwhy (p.283) describes as a collection of ignorability
6.27.2020 1:28am - (2/2) assumptions crafted to "justify available statistical methods" with no one left to tell when/if they hold or not. Compare to the way data-fusion problems are solved using graphs: and . Is light just another face of darkness?

6.26.2020 10:51pm - I am glad to see DAG's and colliders leveraged to clarify "systemic bias" and other hotly debated notions. Note however that the checker-board display is, again, statistical, thus concealing the causal forces behind the proportions. The cited paper by Knox etal shows both.

6.26.2020 10:36pm - A nation lacking nationalism is what we call "benign nationalism" which is, indeed, what makes some nations "great".

6.26.2020 2:07pm - This is the most close-up video of a car ramming scenario I've seen in my life. The body you see flying out of the kiosk is that of a female soldier, who suffered only minor injuries. The driver, through the Palestinian lens of truth, is on his way to a wedding. No words!

6.25.2020 11:05pm - The only way to curb this growing cult of inhumanity is to disseminate the video as broadly as possible; Shame-striken Arabic-speaking viewers will then realize what's being done to the reputation of their culture and heritage, and will break their silence.

6.25.2020 9:11pm - (Replying to @parijat_267 and @daniel_serman) That's an interesting take which, as a scientist, I would entertain only after trying my best to refute, hoping to capture formally at least most features of "fairness" that we sapiens all share. Recall, the "no universal" excuse was tried on "beliefs", then on "causes".

6.25.2020 3:59pm - (Replying to @ehudkar) Phrased this way, it makes perfect sense, and I would be the last one to label it "trivial".

6.25.2020 2:33pm - If there is one thing the ladder of causation tells us, it is that it is impossible to get "causal influence" from data + logistic regression or ANY kind of regression. The latter is rung-1 while the former is rung-2. So I cannot parse the question.

6.25.2020 8:40am - Why? I think this piece created a wonderful opportunity to show people what DAGs can actually do, and how misguided and disoriented their critics are. See my take on the Tale and on pluralism:

6.25.2020 8:23am - @arnoldroth is my good friend, and his daughter Malki is still my angel. This stomach-turning video (below) of her smiley murderer, boasting on Jordanian TV, should be played in every life-asserting march & rally, to the tune of: Malki Life Matters.

6.25.2020 1:21am - (Replying to @19Naranjito82) A criterion that sits behind a door sounds interesting already. Where can I hear more about it?

6.25.2020 1:04am - (Replying to @Faldict and @daniel_serman) The theoretical foundation for fairness analysis rests on mediation and counterfactual logic, eg.#Bookofwhy Ch. 9, or Recent papers building on this foundation are: and

6.23.2020 6:59am - Today, June 23, marks 25 years since the death Jonas Salk, the man who eradicated polio, and whose quote: “The most important question we must ask ourselves is, 'Are we being good ancestors?'” reminds me again why we erect statures in our town squares.

6.23.2020 6:33am - (Replying to @edsonedge) Allegiance to cultural taboos is one reason, others are described here:

6.23.2020 6:17am - Who would imagine that the Journal of Applied Economics could publish a paper using Causal Bayesian Networks to analyze food security: The authors are from China, where they have not been lectured yet on why graphical models are not useful in economics.

6.23.2020 2:29am - As a grad student who landed in NYC in 1960 with $56 in his pocket on a special "scientist quota," I promise all stranded grad students that we, American faculty, will do everything in our power, however limited, to help you circumvent the new draconian restrictions.

6.23.2020 1:29am - (Replying to @BlakeFlayton) Well put. And let's not forget that this sort of rhetoric is cultivated by many Jewish leaders, who would scream from the roof tops upon hearing "Benjamins" or "controlling the media" but would cut deals with their ideological comrades who just want to do away with Israel.

6.22.2020 9:40pm - (Replying to @primedivider I am always open to new ideas, and my email is public.

6.22.2020 9:36pm - (Replying to @analisereal and @f2harrell) I had the same difficulty understanding what the research question is/was.

6.22.2020 6:07am - (Replying to @victorstorchan) If I may, what is "predictive justice"?

6.22.2020 4:50am - (Replying to @garethrobertsza) I tried to respect egos and prides. But, frankly, examining the past decade, can you spot a conceptual breakthrough, or an out-of-the-boxish insight?

6.22.2020 4:37am - Great discussion! Agree! Thanks for posting.

6.22.2020 3:23am - For readers waiting for an insightful article by an economist, this one is the best I have seen in a decade: It offers a platform for modeling an agent's state of belief which is different than yours, analyzing and rectifying causal misrepresentations.

6.22.2020 2:22am - Congratulations to our esteemed colleague Daphne Koller who has just won the 2019 Allen Newell Prize,computer%20science%20and%20other%20disciplines. Daphne is the co-author of the book "Probabilistic Graphical Models" and the co-founder of Coursera Congratulations!!! Daphne.

6.21.2020 6:34am - (Replying to @DrBobGoldberg) Of course not. They only intimidate the ignorant and the gullible, not folks who may expose what they really stand for.

6.21.2020 4:04am - Paying my dues to history, and to collective memory.

6.21.2020 3:45am - (Replying to @ToddTheLinguist) By "leaders" I mean people like SF Mayor who was afraid to ask: "Are you crazy?", and only commented on "the cost of re-painting".There are also academic leaders, who rush to take down whatever can be stained with "racism", however remotely.

6.21.2020 2:47am - (Replying to @zhimin_xiao) Perhaps because I still have a dual American-Israeli citizenship, or because I have a visiting position at the Technion, Haifa. Or, perhaps they wanted to make fun of the BDS movement -- I hope it's the latter.

6.21.2020 1:45am - Just finished my Zoom talk at and my host says 165,000 attended, almost as many as stood with me at the foot of Mt. Sinai (remember?). Could China surpass the US in Causal Inference? They surely have a much smaller investment in data-fitting.

6.20.2020 11:56pm - (1/ ) I am always amazed at the depth of wisdom this verse radiates: "Thou shall not oppress the stranger, because you were once a stranger in the land of Egypt" (Exodus 22:21). Meaning: collective memory is the basis of ethics and values, personal experience is not rich enough. 1/2
6.20.2020 11:56pm - (2/ ) Corollary: Common memories inspire unity, unity then inspires willingness to share values; you can't have shared values among total strangers, because "value" is a distillation of accumulated experience. In summary, the DAG is: Memory--->Unity--->Value

6.20.2020 1:59pm - (Replying to @diydatascience @BHilbush and @tdietterich) True. It's just "one or two examples":, "they are only toppling statues", "its the cost of re-painting" (SF Mayor), and "It's just one Reinoceros!" (Ionesco,1959) I hope it ends here, but it's our duty to say: "it may end THERE."

6.20.2020 5:29am - With "Aunt Jemima", "Columbus Day", and even "Fisher Lecture" gone with the wind, here is a panoramic view, from Jerusalem, of America's new memory-smashing psychosis: Where will it end? A nation lacking memory is a nation lacking unity.

6.20.2020 4:34am - (1/4) Comments on your Front-Door paper: * The expression "a single, strictly exogenous mediator variable" is problematic: (1) Causality p. 82 defines FDC as "A set of variables", not "a single variable". (2) "exogenous mediator" is an oxymoron. I originally called it (1973):
6.20.2020 4:34am - (2/ ) "Mediating Instrumental Variables", best described as an "exogenously-disturbed mediator". * "The first application of FDC" sounds too pessimistic. Situations involving exogenously-disturbed mediators are at least as plausible as "exclusion-restricted
6.20.2020 4:34am - (3/ ) exogenous variables" (traditional IV's) which were introduced 70 yrs earlier, when DAGs were not around to invite scrutiny. Imbens comments reflect that absence * Why introduce FDC in the context of linear regression where ATE is identifiable by
6.20.2020 4:34am - (4/ ) Wright's rule The miracle of FDC stems from its non-parametric identifiability. * The do(X) operator was not introduced as a shorthand for "a variable that is randomly assigned", far from it, see Actions need no randomization.

6.20.2020 2:49am - Reflecting on @DerekHelbing's, "crowd wisdom" begins with scouts that travel independent paths. "Crowd madness" sweeps when scouts read the same newspaper and are afraid to report what they read elsewhere.

6.19.2020 4:15am - Speaking about fact and fiction, data and theory, science and opinion, this hardest of facts should turn every Israel-basher into a speechless pillar of salt, and every BDS rhetoric into a pitiful joke. ONE FULL DAY OF PEACE.

6.19.2020 3:41am - (Replying to @yudapearl and @JKugelgen) I believe your paper will be clearer if it starts with a formal expression of your target quantity, to convince us that it resides on Rung-3, not Rung-2, thus requiring SCM specification. Your second approach appears to yield a Rung-2 quantity, which leaves readers perplexed.

6.19.2020 2:18am - Good catch. It should have read: "the guillotine rule of 1793".

6.19.2020 2:15am - Would love to go back 5 years, undo my sins, and renew our dance in the socially-tight #epilove and #epitwitter family.

6.19.2020 2:06am - And we, data scientists, used to believe in the objectivity of data or, at the very least, in the protective shield of mathematical truths. No more! But who is to blame for the guillotine rule of 1973? The Jacobins? Or those who did not scream "Are you out of your mind?"

6.19.2020 1:27am - President Trump and @MaxBlumenthal !!! What a heavenly match!! One consumed by ego, the other by Israel-bashing convulsions. Read on Blumenthal and Elie Wiesel. Endangered species, so it seems, can attract each other from unimaginable distances.

6.19.2020 12:08am - His memory is a blessing. I will never forget his courageous stance against some pseudo-progressives who chastised him for being a staunch Zionist; he won't budge an inch. "Look at yourselves in the mirror" he said (paraphrased) "Is your humanity dead when it comes to Israel?"

6.18.2020 11:39pm - (Replying to @JKugelgen) Question: If we have the causal graph, we can get the causal effects on the actual group seeking recourse and, then, the best we can do is reach Tian's bounds (combining observational & experimental studies), as in Are you saying we can do better?

6.18.2020 10:44pm - A colleague was gracious enough to send me a picture of the Chinese translation of #Bookofwhy, Let's welcome our Chinese readers into the movement of redirecting data-science towards its scientific aims.

6.18.2020 12:52am - Glad the ML community is aware of the need of Causal Science. The next step is to train 20% of their PhD's to write down a formula for the sentence: "It wasn't the aspirin, but the good news that cured my headache" and 10% to model it.

6.18.2020 12:12am - (Replying to @sai_prasanna) Yes, this is what we mean by: opening up a M shaped path, the path X <- A - C -> Y does not exist in the graph, so we prefer not use this terminology, though some European writers do imagine such paths.

6.16.2020 11:39pm - (Replying to @resourcefulco and @analisereal) Good catch. Embarrassment saved. Happy sailing.

6.16.2020 10:30pm - Many thanks to @analisreal for sharing the 3-line solution to the new Napkin Problem (#Bookofwhy p.240), which explains why the baby was so happy, and why I am so confident that, in 3-5 years, every PhD in Epi, Stat, ML, or Econ. will be able to do it.

6.16.2020 7:28pm - (Replying to @analyticsinme @geoffreyhinton and 11 others) I doubt I would make it to this stellar list had they known my resistance to the data-centric direction "data-science" is taking. eg.

6.16.2020 1:04am - Extremely gratified to see the Change petition exceeding 10,000 signatures I hope the Pakistani Supreme Court takes notice and reverses the ruling to free the murderers of my son, Daniel.

6.16.2020 12:31am - Why is he so happy? Good question! But put yourself in his shoes: When was the last time you've found a three-line proof to a problem that would take the average PhD in your generation three months of hard labor? Note, this is the new Napkin Problem, #Bookofwhy page 240.

6.16.2020 12:05am - I am wondering, Ron, Were there other compromises suggested? The Fisher Lecture, from what I understand, has become an Institute all by itself, more stoney than a statue. Were there other ideas to save the Institute and stone the man?

6.15.2020 5:11am - (Replying to @AndersHuitfeldt @ESteyerberg and 3 others) From what I know about non-parametric causal models, they are all about effect measures, in the sense that, combined with data, they allow you to compute any effect measure you wish. Collapsability, however, is purely statistical notion, defined by conditioning, not by do(x).

6.15.2020 5:05am - (Replying to @ESteyerberg @tmorris_mrc and 2 others) Agree. Collapsability is a statistical notion, see (insensitivity to conditioning), and has little to do with "causality" or "effects".(see Causality p.193). A tenuous connection emerges under the assumption of "stability" (p. 194)

6.15.2020 12:04am - (Replying to @rodakker and @JudaPearl) And I wholeheartedly agree with your focus on the hypothetical nature of modern science. When folks object: "but I am not sure about the diagram" I answer: What if you were sure, would you know how to proceed? or what to test? Or how sensitive the answer is to what you dont know?

6.14.2020 11:54pm - (Replying to @Ron_Wasserstein and @AmstatNews) A suggested compromise: Keep the lecture name and change the honor. Replace "in recognition of the English statistician and biologist Ronald Fisher," to read: "in recognition of the scientific achievements of the English statistician and biologist Ronald Fisher." @daniela_witten

6.14.2020 10:10pm - (Replying to @Vitaly84938575) Publishers are notorious for deliberately hiding this information. One way is to look up our Errata Sheet and check if the corrections were done.

6.14.2020 7:28pm - Good news to all readers who have been waiting patiently for the paperback edition of #Bookofwhy. Our publisher says: and What the publisher hides is that this edition will CORRECT all errors that you have helped us detect. Enjoy!

6.14.2020 4:32pm - (Replying to @kareem_carr and @jodiecongirl) I would forgive the irony if it was only in usage of words, not usage of tools. But I think you are right that stat folks emphasize analysis of variance more than ML folks. Both are stuck, of course, at Rung-1, borrowing terminology from higher rungs, eg "explain" "attribute".

6.14.2020 3:19pm - (Replying to @kareem_carr and @jodiecongirl) What is so important about the "can shrink to zero" criterion that deserves a distinction between such two heavy words like "inference" and "prediction"? Will guessing the age of a skeleton amount to "prediction" just because we will never know it for sure? Hard to buy?

6.14.2020 3:04pm - (Replying to @TuanAIWS) Interesting. Has anyone summarized the causal rules of Buddhism in a Tweetable format?

6.14.2020 6:46am - (Replying to @jodiecongirl and @kareem_carr) I think by "inference" you mean "estimating conditional expectation", which embraces both prediction and redrodiction. Would guessing your parents income be "prediction" or "inference"? As we can tell from our Tweets, statisticians are still choosing their terms.

6.14.2020 5:52am - (Replying to @sometimes_data and @kareem_carr) Thanks for the pointer. Note that Efron does not use the word "inference" and that he borrows the causal notion of "attribution" to mean "statistical significance" (variance reduction), confirming my observation that statistics has remained causality-blind

6.14.2020 1:50am - I've found this new article to provide the most thoughtful narrative of the destruction of American news media and, by implication, the destruction of academia as well. I'm still optimistic, knowing that people with no salary to protect WILL speak up.

6.13.2020 11:58pm - (Replying to @Abhishe45860449) My colleagues in ML proud themselves on using NO MODEL. So, I do not know how to answer your question. The role of causality? To do better than: User: Why was my loan was denied? Explainable ML: Because we used Dirichlet prior

6.13.2020 11:47pm - (Replying to @jodiecongirl and @kareem_carr) Thanks for elaborating. But this is the first I hear that statisticians care about "effects" and "interpretability" . I thought these notions are not part of statistics proper. E.g. is estimating likelihood of a disease, given a set of symptoms a "prediction" or "inference" task?

6.13.2020 10:53pm - Counterfactual reasoning in the age of normalized tantrum.

6.13.2020 2:01pm - That's a good point. If we wish to compare natural talents, we should use the "per literate capita" metric. But to compare the ability and commitment of society to investment in education, the "per capita" metric would be more appropriate.

6.13.2020 3:10am - This is the first time I see the front-door formula estimated using Inverse Probability Weighing. Great to hear the wheels of progress squeaking.

6.13.2020 2:38am - (1/2) I did not come from academic family. My father toiled in the orange groves. But my teachers were giants, as I describe here (p.7) and here Oh, and one should not forget my home town, Bnai Brak, founded in 1924, which was named
6.13.2020 2:38am - (2/2) after the Mishnaic town which, in 2nd century AD became a Center of High Learning. Far fetched? Not to us, kids; we knew that those 2nd century sages were watching us daily and expecting us to follow the tradition of deep learning. No excuse.

6.13.2020 1:47am - (Replying to @rodakker) An ideal randomized experiment is still "ideal", and the uncertainties in practical RCT's are not truly "known", for they involve selection bias and other imperfections that are assumed small using "causal assumptions"; same assumptions that RCT's are supposed to escape.

6.13.2020 1:23am - (Replying to @EylonALevy and @EinatWilf) Interesting. Many Indians feel that Winston's status should go down because he was "directly responsible" for famine death of 3 million Indians. Luther was only "indirectly responsible" for European antisemitism and its resulting murder of ??? million Jews. Direct vs Indirect

6.13.2020 1:09am - (Replying to @EinatWilf) I can't resist the temptation of asking (hypothetically) how many of those bemoaning "destroy the two-state solution" would switch position had it been the case that a given act would actually "advance the two-state solution"

6.13.2020 12:29am - Beg to differ. You don't turn an apple into an orange by looking at it through colored eyeglasses. #Bookofwhy (pp. 143-150) shows why randomized trials are merely skillful interrogations of effects that exist with or w/o the trial. An apple is an apple when properly defined.

6.12.2020 11:54pm - My My!! He got it right!! The most inspiring photo I've seen on Twitter in the past year!!

6.12.2020 11:30pm - This is not merely good, it is the GREATER good. What we now need is an authorized list of people/movements to question, or mildly criticize, or less than deify; the hate list is way outdated.

6.12.2020 11:07pm - (Replying to @TuanAIWS) Wait a minute, @TuanAIWS , streets are named after dead people, and I still have 36 productive years left to explore the truly hard questions.

6.12.2020 10:52pm - OK, I confess, the joy is mixed with pride of seeing Israel rank 1st in # of Turing Laureates (per capita), where "pride" is given CI semantics, i.e., an assignment credit to a culture that encourages deep learning and independent thinking.

6.12.2020 10:32pm - (Replying to @BoredAlanTuring) You look awfully familiar! #honestly

6.12.2020 4:43pm - What a joy to see a street named after Alan Turing spelled in Hebrew and English, in Herziliya (named after Theodor Herzl), 5 miles north of my home town. May we live to be worthy of these legacies.

6.12.2020 1:03am - Your question (paraphrased): "Why not let model-free machine learning simulate evolution and reach human-level intelligence as our ancestors did?" was considered by Turing (1950). I touch on it here (~4 min) using the snake and eagle analogy.

6.11.2020 4:34am - This makes me admire the Hebrew Bible. King David is arguably the most admired of all its ancient heroes and, yet, the authors do not attempt to whitewash any of his weaknesses, including murders and moral atrocities that transcend time-sensitive norms. Time for every season.

6.11.2020 3:38am - (Replying to @pratikmehta94) This may be asking too much, I realize. At the same time you can't deny the British people, and their allies, the memory of those days in 1940-1 when they stood alone and prevailed. And (as I tweeted before) collective memory requires a leader's face to be encoded and transmitted

6.11.2020 3:19am - (Replying to @adabhishekdabas) DAG's bless us with a logic that protects us from fallacies such as "good for men, good for women, bad for people". This protection transcends data, bias and sample size. Dont you feel safer being protected from conclusions that our mind finds offensive?

6.10.2020 11:43pm - (Replying to @maliniw90th) When all the statues are down, the town squares are empty, and children books too, just dragons and super-women -- no historical figures, it is then when I will open my first science book, to play chess w/ Pythagoras and Archimedes, admiring their victories, forgiving their sins.

6.10.2020 11:21pm - (Replying to @likeavass) Please re-read my tweet. The words "allegation of blemishes" was used for Gandhi's critics. I hope I have not lost you after all.

6.10.2020 10:40pm - (1/ ) My stand on Churchill and his statue acknowledges the dark side(s) of the man, and my underestimating the intensity to which he is hated by people who were victims of his policies, from Indians to Poles. Still, statues are erected to glorify ideas, and hours of greatness in
6.10.2020 10:40pm - (2/ ) a nation history. Defacing a statue mocks a nation's hours of honor and, indirectly, mocks the lessons we wish our grandchildren to inherit from our experience. And, yes, we must continue to embody those lessons in a face of a person, not abstract symbols, because our mind's
6.10.2020 10:40pm - (3/ ) eye operates thus; we say "Einstein", instead of "relativity"; we say "Galileo" instead of "experimental methods". Faces make us feel part of the idea. So let us feel free to admire Thomas Jefferson for what he means to us, in front of an un-defaced Jefferson Memorial,
6.10.2020 10:40pm - (4/4) acknowledging his ownership of slaves. And let us admire Mahatma Gandhi for what he means to us, precisely as he is depicted in his monument in Delhi, and leave allegations of blemishes to scholarly articles. Our grandchildren beg us for clarity of ideas, cast in human face.

6.10.2020 1:29am - (Replying to @AvanteQuora) "Graphical" models provide the vocabulary for CI. But "probabilistic" models, by definition, remain in Rung-1, unless the instructor/author is enlightened and is aware of the opportunities to climb up.

6.10.2020 12:59am - (Replying to @Abhishe45860449) Mathematics tells us the we dont get a unique solution if we have 3 equations and 4 unknown. This is precisely what the mathematics of CI tells us: For every causal question, data alone are always compatible with both an answer and its negation. See Simpsn

6.10.2020 12:37am - The BDS clowns will stop their campus circus ONLY when each of their so-called "resolutions" ends up with the billion dollar question, and their true face is exposed to the entire campus. They feed on the ignorant and gullible, and cannot survive the light of truth.

6.10.2020 12:25am - Good question, which every ML researcher asks. The answer is 2-fold: (1) Mathematics tells us we can't go above Rung-1 w/o some causal assumptions. (2) To grasp how mild the assumptions, and what we can do once we make them, we must study the logic of CI, not statistics.

6.9.2020 11:59pm - UC Davis Student President Vetoes BDS Resolution And listen to the perpetual billion dollar question: "I can commit to supporting a Jewish and democratic state of Israel peacefully coexisting with it's Palestinian neighbor. Can you do the same?" - Nada!

6.9.2020 3:29am - (Replying to @raymondshpeley and @krysdolega) I was not aware of this side of Churchill, thanks for chiming in. Still, the statue defaced was not erected for the man as much as it was for his great hour. We need more such statues in our public squares for there aren't many great hours left to commemorate.

6.9.2020 12:54am - The defaced status of Winston Churchill is a microscope to the mentality and intellectual making of the BDS movement and its leaders. History will forever contrast the greatness of Churchill with the dishonesty of BDS.

6.9.2020 12:21am - (Replying to @PHuenermund and @JaminSpeer) The only thing I do not like about the US anthem is that it is almost always sung by a PERFORMER, not by the audience. Have you noticed? To me, growing up in Israel, this is more than strange, it defies the very meaning of "national" as well as of "anthem".

6.8.2020 6:15am - (Replying to @mddimick and @DaveBrady72) Sorry, I've never heard of "invariant contextual factors". Need a toy example (my weakness) no philosophy.

6.8.2020 6:11am - (Replying to @julianschuess and @DAGophile) I must confess that despite many efforts in the past I never truly understood what the "settable" system is, or what it offers. But Karim is still around, at UVA, econ. Perhaps he can explain it to us, mortals, and help us elevate econ. one notch up the ladder.

6.8.2020 4:56am - Here is another proof that a country condemned for existing can't be dogmatic. 39 years ago today, colleagues came to my office - indignant: How dare they do that? Unprovoked! etc. Yesterday I called my sister in Tel Aviv -- she and hers are alive! Optimistic. Worried about us!

6.8.2020 4:21am - @DeepMind interest in causal inference may signal a brighter future for data science. And @csylviav suggestion to start with #Bookofwhy, then proceed with Primer, has the potential of elevating many folks to Rung-2, then Rung-3. Happy sailing.

6.8.2020 2:26am - I think adopting the illusion that there IS an objective reality out there is extremely important, else we end up doing psychology or meta-psychology instead of science.

6.8.2020 2:14am - I have a strong hunch that Israel will take a lead in re-directing "data science" to its right course. After all, a country condemned for existing can't afford to be dogmatic.

6.8.2020 1:18am - Good point. We should change "data-assisted-science" to "data-intensive-science".

6.8.2020 1:00am - Agree. To be a "science", "data-science" should read "data-assisted-science". The way it reads now, "the science of data" has little to do with science, it has turned into "computer-assisted-statistics" or "statistics on stereoides". Yet donors love building those tall "Centers."

6.7.2020 9:50pm - (Replying to @eidlin and @DaveBrady72) My 2 cents. We must define things before we discuss how to estimate them. The hypothetical experiment of re-staffing police with color-blind cops is WELL DEFINED (removing an edge in the graph). Once we do that, we discuss what data/experiments are needed to assess its effect.

6.7.2020 9:37pm - (1/ ) Yes. You will be able to use "automated Machine Learning flows" to go from samples to distributions, then use causality to go from distributions to answer questions that cannot be answered by data alone, no matter how clever the learner. Awareness of the "no matter" is crucial
6.7.2020 9:37pm - (2/2) and it is THE mental block that keeps ML folks from seeing the necessity of causal modeling in tasks requiring more than data-fitting (Rung-1). I hope this road block disappears by the time you finish the book, and you'll then find yourself ahead of 90% of your ML peers.

6.7.2020 4:16pm - A word of caution. The statement below is about what data-science ought to be, not how it is practiced today. If you visit any of the newly erected "data science centers" you'l find that the taller the building the less likely you are to find a researcher who interrogates reality

6.7.2020 4:07pm - (Replying to @desai_pratik) My views are simple: We have so much philosophical soul searching to do, especially in the age of computers - the first model of human thoughts -- that debating what Hegel believed or did not believe seems a waste of time.

6.7.2020 3:47pm - (Replying to @luislamb) It just occurred to me. In computer science, we start with fundamental laws and go to the corollaries. In Epi, so it seems, folks go the other way; starting with phenomena such as collider bias, then proceed towards unifying principles. There must be wisdom in both approaches.

6.7.2020 3:32pm - (Replying to @resourcefulco) Kudos to the Bristol group for writing about collider bias even before COVID-19. Next conquest should be d-separation.

6.7.2020 2:37pm - One positive aspect of COVID-19 pandemic has been rising attention to the fundamentals of causal reasoning, in this case: collider bias. Yet rather than surrendering to bias, it should be coupled IMO with methods of RECOVERING from selection bias, eg

6.7.2020 2:12pm - (Replying to @AGI_programmer) I should have said "interrogating reality" or "interpreting data", as said in #Bookofwhy, or, more accurately, "using data to interrogate reality". Thanks for the pointer.

6.7.2020 2:01pm - For those who ask, I must add that my views on history of thought, hence on debating about ancient philosophers beliefs, is fairly whiggish, as summarized in this (unpublished) paper with Dana MacKenzie.

6.7.2020 12:21am - (Replying to @HenMazzig) You are right, sister Linda, "people of certain lineage" are saying explicitly that you ain't fit to lead any progressive movement, see, not because of your "lineage" but because you are a Zionophobic bigot, something you cannot bring yourself to deny.

6.6.2020 11:39pm - Titled: What's Sex Got to Do With Fair Machine Learning?, this paper will probably generate lively discussions. It's hard for me to see though how "constitutive diagrams" would tell us how the social meaning of a group would be changed.

6.6.2020 10:50pm - An interesting paper on how people judge causal responsibility of agents and events Note: I find the proposed "robustness-based" criterion hard to grasp, perhaps because it was not defined symbolically, as did causal necessity and sufficiency.

6.6.2020 10:11pm - (Replying to @EmraniMd and @PeterBeinart) Yet Linda Sarsour is not invited by Hillel Directors to speak on US campuses and sanctify their discomfort with Israel, as they did with @PeterBeinart on my campus in 2010. See "Jews of Discomfort"

6.6.2020 9:59pm - (Replying to @zevkalman) I was already 31, but the traumatic experience of anxiety, indecision and material fear of those few days have left such a lasting mark on our collective psyche, that it is still shaping our perception of what peace arrangement should prevent from happening.

6.6.2020 9:38pm - (Replying to @primedivider and @lexfridman) I second! Whoever is shocked by common sense deserves a transcript. I can use one too, though I am no longer shocked, just tolerating.

6.6.2020 9:10pm - I love philosophy, but what puts me to sleep is when philosophers start arguing about what another philosopher believed or did not believe.

6.6.2020 8:35pm - (Replying to @lmonasterio) I had a slight suspicion that you are referring to Richard von Mises whom I know from my philosophy reading. I can even see his book "Positivism" smiling at me from the bookshelf. But I did not realize he means so much to people in Brazil. Thanks for telling me.

6.6.2020 4:57pm - (Replying to @DaveBrady72 and @doc_thoughts) Such claims are not coming from counterfactualists but from Potential Outcome disciples (eg Holland 1986, Imbens and Rubin, 2015) who never snapped out of the outdated "treatment assignment" conception of causation. See

6.6.2020 4:10pm - As much as I dislike military parades and army uniforms, today, on D-Day, my head bows down at the sight (below) of these young men going into a battle they knew had to be won.

6.6.2020 7:06am - (Replying to @iGabrielX360) Half the world will never forgive Israel for choosing Life over Love. Which, as a student of human thought I perfectly understand; nothing could be more insulting than rejecting one's love. Luckily, the other half admires Israel for saying NO to postmortem love.

6.6.2020 6:11am - (1/3) Today marks the 53rd anniversary of the Six-Day War. A war that has left three deep scars in the collective psyche of all Israelis. First, the scar of ANXIETY on the days leading to the war: Is Nasser about to do what he has been promising for 19 years -- annihilation?"
6.6.2020 6:11am - (2/3) Second, the euphoric scar of SECURITY: We are alive and free to actually walk the paths we took so many times as children since we learned the Aleph-Beth. Thirdly, the scar of the Arab Summit (Khartoum, Aug 29, 1967) and its devastating "THREE-NO's" resolution - No peace,..
6.6.2020 6:11am - (3/3) No recognition, no negotiation with Israel -- which gave birth to the settlement ideology: "If we are destined to live by the sword to the end of days, we better do it from a position of strength, not at the mercy of guns or rockets." We are still carrying these three scars.

6.6.2020 5:11am - (Replying to @Claire_Voltaire) Well spoken, @Clare_Voltaire. The Chutzpah of some anti-Zionists may exceed their hypocrisy, see: And @PeterBeinart is master in ennobling this hypocrisy among "Jews of discomfort," as I observed here: Shabbat Shalom.

6.6.2020 4:06am - (Replying to @PeterBeinart) Excuse me, but on my campus it's the anti-Zionists who are waiving the "we aren't anti-Semitic" card to avoid debate of their moral deformity, as if Zionophobia is a Kosher type of bigotry, entitled to automatic free-speech pass, unlike Islamophobic or white supremacist speeches

6.6.2020 3:19am - (Replying to @DrIanKellar @SGSmith_87 and 2 others) I am glad to have been a catalyst for the clash between Matthew and his students/friends but the idea of not knowing "what he was talking about" strikes me strange. If he was talking causal inference in 2004 he must have known what he was talking about.

6.6.2020 2:51am - There are so many enlightened, learned and globally-minded colleagues to whom I would like to send this Tweet that I have decided to retweet it w/o a comment, just a humble mirror of their moral deformity.

6.5.2020 5:21am - What is badly amiss in our academic ivory towers is for colleagues of conscience to immediately speak out and publicly expose Ziono-phobic bigots for what they are.

6.5.2020 5:04am - Metascience: the science of doing science: The reason I think this collection will be of interest to readers is that the adventurous journey causality has taken in the past century is a microcosm for many of the issues discussed:

6.5.2020 4:15am - This is a nice introduction to graphical causal models for energy research folks, who study the effects of various intervention programs on household energy saving behavior. It provides a fairly comprehensive summary, and even touches on Simpson's paradox.

6.5.2020 12:05am - So true!. Ten years from now, historians will ask: How did scientific leaders of the time allow society to invest all its educational and financial resources in data-fitting technologies and so little on data-interpretation science?

6.4.2020 9:29pm - (Replying to @eliasbareinboim @PHuenermund and 13 others) Congratulations, Paul! And dont forget your comrades, still in the trenches.

6.4.2020 1:55pm - (Replying to @neuro_data @KordingLab and 2 others) Overslept - missed it - sorry. This post represents my current position on how healthcare leadership is missing on AI Feel free to re-post to participants.

6.4.2020 1:26pm - (Replying to @julianfdezme and @bradheath) A sample of 1 shows no correlation.

6.4.2020 1:24pm - (Replying to @tmorris_mrc and @stephensenn) Will have a look. Her examples are very clear.

6.4.2020 6:03am - (Replying to @HenningStrandin) I hate to consider myself "radical" though when it comes to philosophy, I dont mind. I ask different questions than Lewis and Hume: (1) How come people form consensus on counterfactuals much before causation? (2)How would a robot understand our natural language - counterfactuals?

6.4.2020 5:07am - (Replying to @stephensenn) In the simplest setting, we have n units, one treatment and its compliment (placebo), and we send k randomly chosen units to treatment and n-k to placebo. The calculus looks at the outcomes of all units and outputs (1) An estimate of ATE (2) Standard deviation. Am I getting it?

6.4.2020 4:41am - (Replying to @stephensenn) I think I am getting closer to understanding. Just some clarifications: (1) What is "block & treatment structure" (2) estimates of what? As Americans say "I am almost there".

6.4.2020 2:08am - (Replying to @stephensenn) I have tried several time to read your descriptions of Nelder's calculus but I am missing the basic vocabulary: eg. structure, block structure, design, balance, etc. Can you translate to non-experts like me? Perhaps in a form of a simple question that the calculus can answer?

6.4.2020 12:09am - (Replying to @desai_pratik) Try this long link: It takes me straight to the Microsoft lecture, Paris, 2013.

6.4.2020 12:04am - (Replying to @stephensenn) I am inclined to believe it but, so far, I have not been able to grasp what this calculus does, namely, what information it takes as input and what conclusions it delivers as output. I have been bought up to think in input-output terms, an incurable weakness which served me well.

6.3.2020 10:21pm - (Replying to @11kilobytes and @Kirsten3531) As strange as it may sound, I have not changed my mind.

6.3.2020 10:18pm - (Replying to @luislamb) Thanks. It is always invigorating to listen to a younger speaker, and a friendly one, too.

6.3.2020 3:24pm - Here is a better link to my Microsoft lecture It goes straight to the talk w/o commercials.

6.3.2020 4:52am - (Replying to @NandoDF @ProfBrianCox and @d_spiegel) This is precisely what I am worried about, that politically motivated actors will hijack the causal revolution, while our science leaders, both in stat and ML will be watching from a distance and continue the data-fitting gold-rush unquestioned.

6.3.2020 4:33am - In my Microsoft lecture I talked about statistics as the "calculus of shadows" and causality as a 3D calculus. I can think of several 4D's and 5D's, like object-based and agent-based calculi. But we first need to overturn th tyranny of statistical thinking

6.3.2020 4:14am - (Replying to @Bertrand_allen2 and @lmonasterio) I disagree with Hitchcock (author of, counterfactuals come first, causation second. The former is consensual, the latter debatable.

6.3.2020 3:36am - Homework solutions to the exercises here: will gladly be provided if you write to I just looked them over, the intellectual content compiled in this homework deserves a book of its own. Anyone volunteers as a co-author?

6.3.2020 3:21am - Glad you asked. I have been in this business for 3 decades and have not found anything clearer, more rigorous or friendlier than Chapter 4 of Primer. Here is a free link Enjoy, and don't let the mystics tell you counterfacuals are "hard".

6.2.2020 5:00am - (Replying to @19Naranjito82) I know nothing about antitrust, mergers, etc. So, I cant really be of much help, except to suggest: What do you know about those issues that a super statistician does not. That's the beginning, put it down on paper, first in words, then in DAG, the rest will follow.

6.1.2020 11:07pm - (Replying to @FarihaAnna) Debugging is where Bayesian networks hatched. Glad to see DAGs returning to debugging. But being unfamiliar with the systems being debugged, makes it difficult for me to generalize the proposed method.

6.1.2020 10:50pm - (Replying to @tytung2020) An intervention can be either external or internal, it means a "change" relative to the way things normally behave. An earthquake is external to my everyday life, but internal to geological dynamics. This paper may be helpful, an interpretation of do(x).

6.1.2020 8:53pm - (Replying to @wexlerwriting) Sanity demands an illusion of normalcy.

6.1.2020 4:15pm - An interesting paper on causal discovery using both observations and interventions The problem is to recover a causal graph while minimizing the cost of interventions.

6.1.2020 3:13pm - (Replying to @kavlak_batuhan and @dlmillimet) Speaking of agriculture, I've just received this paper on Mango tree growth: Hope it is of help.

6.1.2020 4:33am - My wife Ruth remembers that day, 6 yr old, Baghdad, 1941. Her mother forbade her to look out the window, she did anyhow and saw the bodies, in puddles of blood, as the stores across the street were looted, and the mob roared. The mob still roars. Not in Baghdad, in the USA!

6.1.2020 3:24am - Pleased to inform: The Supreme Court of Pakistan has deferred hearing our appeal of the Sindh Court decision that would free the four men convicted in the kidnapping and murder of our son, Daniel Pearl. We remain committed to pursuing our appeal and realizing justice for our son.

6.1.2020 12:36am - (Replying to @ZionessMovement) I was wondering what reception you had in the #BlackLivesMatter demonstration, and whether it was better than the one in the Women March. Did they allow you to carry Israeli flags?

5.31.2020 12:42pm - What do mainstream statisticians want to know about causal inference and The Book of Why? This interview with David Hand unveils part of the puzzle: And given that statistical thinking still dominates ML research, the answer may be of interest to many.

5.31.2020 11:03pm - (Replying to @PLetouze) Once you talk to computer scientists you can also compare assembly language to C++. And, to engineers: a wave form vs. its Fourier transform. I once wrote a note on it, let see,... here it is: I still enjoy reading it. Take a look.

5.31.2020 10:55pm - (Replying to @kavlak_batuhan and @dlmillimet) Nothing comes to mind, sorry. But if I had a problem in agriculture, I would try solving it the easy way: what do I know, what I want to know, what data can I gather. Done. The rest is sheer logic. I bet your solution will turn out better than what you can find in the literature.

5.31.2020 10:40pm - (Replying to @lqpeng @TimHarford and @FinancialTimes) Agree. @TimHarford should review @Bookofwhy for the innocent readers of @FinancialTimes

5.31.2020 10:32pm - (Replying to @mark_learns_sci) The trouble is that some grown ups who could find Israel on some maps are pretending to fall for these juvenile slogans and read them as saintly messages of peace.

5.31.2020 1:43pm - Some readers complain that economics was not mentioned in my tweet. Agree, econs have different kinds of questions, which we can see in @dlmillimet post: "Let's turn to the regression model so we can use some econometric". Plus, I wonder if they would interview an outsider.

5.31.2020 4:54am - (Replying to @yudapearl and @No_W_A_R) No country has ever been given sovereignty while calling for the obliteration of its neighbor. This is no paranoia but commonsense. Sovereignty entails the means to import arms, including missiles, and to accomplish what you have promised your children for 72 years.

5.31.2020 4:42am - (Replying to @No_W_A_R) I may be weak in Arabic but I know something about cause and effect. I know for fact that ALL my friends in Israel have been tuning their tallest antennas into Palestinian classrooms and media, yearning desperately for an inkling of acceptance - 72 years of nada. Thats THE CAUSE

5.31.2020 3:11am - (Replying to @No_W_A_R) Intentional laws, resolutions and declarations are always predicated on mutual acceptance and end of hostilities. As you know well, no Palestinian leader, cleric or educator has ever accepted Israel's existence (in Arabic!). To every Arabic-speaking person this means "no-Israel".

5.31.2020 1:20am - Can colleges launch data science programs fast enough? via @EdDiveK12 What does it take to tell them that data without causes ain't a science but statistics on steroids. I said so here: and am even more convinced now.

5.31.2020 12:48am - (1/3) Reread Keele etal's paper on "causal interpretations" and I agree with their view of how the question is mishandled in the Political Science literature. Readers seeking a quick answer, the solution is given in Theorem 5.3.1, p. 150 of Causality. I am
5.31.2020 12:48am - (2/3) surprised that Keele etal do not name the criterion, "single-door", which accurately and painlessly tells you when a parameter can be estimated by regression. It is also a question that economists often ask, can't answer, and refuse help. See why:
5.31.2020 12:48am - (3/3) @dlmillimet 's post also reveals why economists have such hard time answering their own questions. This says it all: "Let's turn to the regression model so we can use some econometric". They know that regression says nothing about causation, yet DAG-less habit intoxicates.

5.30.2020 11:35pm - (Replying to @No_W_A_R) I beg your pardon, but I am the last to connect anti-Israelism to anti-semitism. I am calling for treating the former as separate, more dangerous form of racism, whose aims and consequences you know very well, since they preceded the creation of Israel -- I was there.

5.30.2020 11:14pm - (Replying to @fields_unsown) Indeed. Anti-semitism is a disease of an unfortunate up-bringing. Anti-Israelism is a calculated genocide. The former is an emotional dislike, which everyone likes to abhor; the later, an eliminationist design which some take very very seriously, including some dear neighbors.

5.30.2020 10:46pm - Vandalism is never OK. Antisemitism is never OK. But anti-Israelism is the lowest.

5.30.2020 8:26pm - This post reinforces my conviction (Causality p. 135) that the question whether a parameter has a "causal interpretation" has no causal interpretation, through Keele etal's paper may help us decipher what authors are really asking by asking it.

5.30.2020 6:09am - Congratulations to Scott and his graduating class -- the FIRST ever high-school to master causal inference! These kids are heading to college and I don't envy their stat-101 professors; they would not let them get away with causality-blind cliches. They will demand substance.

5.29.2020 12:28am - I am not familiar with #AIOps, but I strongly resonate to what you say about #ML and #DL researchers. Climbing up the ladder of #Causality is not in their DNA. Even when they wish to do #CI, they just can't see the ladder. Not their fault; it ain't in their textbooks.

5.28.2020 3:17pm (Replying to @strangeloopdave and @mishtal) There were many stories of SS officers showing a nanosecond of humanity before returning to savagery. My aunt's point was that it was direct eye contacts which triggered those nanoseconds. American Jews often forget that self respect is what sustains the respect of others.

5.28.2020 2:35am - (Replying to @LebiedzMarcin) The new version is saved as PDF. See url: short url: If it works, please tell others.

5.28.2020 12:06am - (Replying to @mishtal) My aunt, who survived Auschwitz told me a lesson I will never forget. Inmates who fell to the SS officer's feet and begged for their lives were sent to their death, and those who looked him straight in the eyes were (often) spared.

5.27.2020 9:06pm - Current @AAUP 's leadership actually views Zionophobic racism as a cute form of academic "activism". It is members/donors outrage that will force them to resign and let adult educators advance the cause of higher education.

5.27.2020 4:16pm - For readers who had trouble opening the ppt slides of "The Silent History of Cause and Effect", here is a post-treatment version, surviving a interventional surgery. I hope it works.

5.27.2020 5:56am - (Replying to @jonasobleser) Try any of these: url: short url:

5.26.2020 11:46pm - Will try but, truthfully, I get impatient and frustrated listening to audience, moderators, even reporters not asking the questions I wish to ask. @benmartel97

5.26.2020 11:35pm - History-minded readers may also enjoy, which describes how statistics, lacking the necessary mathematical tools, managed to cope with increasing demands for policy evaluation from observational studies in the years 1955-1980. A tribute to William Cochran.

5.26.2020 3:57pm - (Replying to @wokebeartov) Sorry to admit ignorance but I've not heard about Friston's FEP. What is it (in Twitter language)? And I am not sure causality is a "new Bayesian idea" if you consider my observations in "Why I am only half Bayesian":

5.26.2020 1:35am - Economists tell us: If you have a good instrument, you dont need a model ( This new paper asks: What if you suspect selection bias? I don't know how a model-free economist would begin to derive the results obtained in this paper.

5.25.2020 10:22pm - (Replying to @yudapearl @scmbradley and 2 others) I was right! Gong and Meng totally miss the paradoxical element of Simpson's paradox. They attribute the surprise to incomplete specification of a probability function, forgetting that even with complete specification the surprise comes and goes depending on the background STORY.

5.25.2020 7:08pm - (Replying to @KambizzzKhan) "You are not expected to finish the job, but you ain't free to quit trying" (Mishna, Pirkei Avot 2:21)

5.25.2020 3:59pm - (Replying to @danijarh) The two options will forever be competing for our conception until we find a clash between the logic of causal modeling and empirical observations. For example, a violation of the inequality P(y_x) > P(x,y) [Causality, Eq.(9.33)]. So far, I have not seen a clash reported.

5.25.2020 2:19pm - (Replying to @MaartenvSmeden @Richie_Research and @matherion) It is not "easy", but it is the "shallowest" form of knowledge we can expect about mechanisms. As to COVID, see Why it is belated? Because we still do not have a "Causal Science Center" to prepare society for the next Tsunami of information confusion.

5.25.2020 2:10pm - (Replying to @scmbradley @fitelson and @olehjortland) I havn't seen it, thanks for the pointer. But considering that it was authored at Harvard-Stat, my guts tell me to expect the same consistent denial of the C-word as we have seen before (see, namely, round-about denial of the logic of human thought.

5.25.2020 4:29am - Per readers requests, I am happy to share the colored ppt slides of "The Silent History of Cause and Effect": Silent histories deserve colorful showcase. Enjoy.

5.25.2020 12:31am - Last week I had the privilege of celebrating a virtual Muslim-Jewish Interfaith Iftar, renewing my 2003-2010 dialogue with Islamic Studies Professor Akbar Ahmed. Here's a video of this communal celebration, with 25 inspiring interfaith activists:

5.24.2020 7:09pm - (Replying to @KristinaLerman @bgoncalves and @fitelson) I take it that if your data is found clean of Simpson's reversals you would be able to do 1) generalization, 2) explanation, 3) fair predictions with full confidence. My theory tells me that all 3 tasks are hopeless w/o causal models Where did I go wrong?

5.24.2020 3:46pm - (Replying to @bgoncalves @fitelson and @KristinaLerman) Before we spend the energy of identifying Simpson's reversal in very large datasets we need to sit down and ask "why"?, what would it tell us if we find such a benign pattern in the haystack. I don't see the answer in Lerman's paper.

5.24.2020 3:28pm - (Replying to @bgoncalves and @fitelson) I believe that if one digs deeply into the agendas of the philosophical and psychological approaches one would find them computational in disguise.

5.24.2020 3:15pm - (Replying to @fitelson) Purely probabilistic analyses, on the other hand, tend to reinforce the "probabilistic causality" language which some (admitted fewer and fewer) philosophers still cling to. I marvel the reasons for this clinging here I was once there, among the clingers.

5.24.2020 2:58pm - (Replying to @bgoncalves and @fitelson) If you really like debates, these two blog postings and will give you ample sports for the money. But as I always say: One toy example is worth 100 debates.

5.24.2020 2:47pm - (Replying to @lewbel) Glad you liked the slides. I thought you would scold me for asking whether Haavelmo's legacy will survive the two attempted hijackings.

5.24.2020 9:53am - (Replying to @roydanroy @guyvdb and 2 others) I would expect so too. But it did not happen in those centers that I've visited, and I now understand why.

5.24.2020 8:45am - I don't believe such corruption would last one day had members and donors of @AAUP known what is done in their name. I therefore retweet it to readers, in case some are still members in this hijacked organization, so called "American Association of University Professors".

5.24.2020 8:19am - (Replying to @AaronCarine @elderofziyon and 2 others) Those who tried to find consistent logic in Zionophobic rhetoric have given up long time ago. The one logic that I've found is: Smear, Malign and Lie. That's how they win the hearts of the ignorant and the gullible, for example, the dpt of Anthropology at my university, UCLA.

5.24.2020 3:46am - As a token appreciation to readers of #Bookofwhy, I am sharing the slides I used in a seminar titled: "The Silent History of Cause and Effect" given at UCLA History of Science Reading Group. Feel free to use any of these historical-philosophical snippets:

5.24.2020 3:15am - (Replying to @fitelson) Purely probabilistic analyses, on the other hand, tend to reinforce the "probabilistic causality" language which some (admitted fewer and fewer) philosophers still cling to. I marvel the reasons for this clinging here I was once there, among the clingers.

5.24.2020 2:57am - (Replying to @ArtisanalAnn @eliasbareinboim and 4 others) I wonder why? I never met Elon Musk, so I am wondering what dangers you are alerting us to.

5.24.2020 2:51am - (Replying to @PHuenermund @conjugateprior and @beyers_louw) What are those "coefficients for Z"? Is that how the typical econ/management/social science folks talk? W/o distinguising outgoing from incoming? w/o distinguishing structural from partial regression coefficients? Why exclude readers like me who is trying to study how they talk?

5.24.2020 2:45am - (Replying to @PHuenermund and @beyers_louw) First, why not clarify to readers what kind of coefficients you wish to exclude - I am still baffled. Second, are there more precise conditions than the single door criterion? Why not educate people when you got their attention; they will thank you forever.

5.24.2020 2:35am - These wise words of @RayDalio seem to be taken straight from the manifesto of the first "Causal Science Center". Hey, @RayDalio , would you be the Center 's first spokesman? Honorary President? Beneficiary? The algorithms you spoke about are ready to serve humanity.

5.23.2020 9:42pm - (Replying to @elprofe_ivan) I would start with a list of tasks that EVERY data analyst aspires to do and which require causal modeling. Seven such tasks are listed here: Formal, yet gentle introductions to these tasks are given in Primer: Good luck.

5.23.2020 9:31pm - Paradoxes that offend our intuition unveil the logic that governs our intuition, namely, the engine that is driving our minds. This blog describes how this logic varies with the causal story told. Highly recommended.

5.23.2020 9:12pm - An incredible production of "Jerusalem of Gold" ending the international celebration of Jerusalem Day. An on-line choir of beautiful angels on the background of the city that symbolizes peace, hope and the oneness of mankind:

5.23.2020 8:51pm - Abdulhadi's record of hate spreading activities is widely known. Therefore, her selection by @AAUP is a POLITICAL statement meant to encourage such activities in the future. The Tax Exempt status of @AAUP should be examined for compliance with its stated educational mission.

5.23.2020 8:33pm - Embroiled in choreographing Causal Science Centers I've almost forgotten that May 20 marks the 2-year anniversary of the publication of #Bookofwhy. Dana and I are grateful to all readers who contributed to our understanding that the book has made a dent on scientific methodology.

5.23.2020 6:51pm - (Replying to @eliasbareinboim) From my perspective, RL resides at Rung-1.5 of the causal hierarchy, laboring to cover Rung-2, but few RL researchers know about the Ladder, or where they are on it. Would you agree with this assessment?

5.23.2020 6:42pm - Readers wrote, and I agree: It is not the paradox that is stubborn but philosophers (and scientists). After getting your PhD and tenure viewing knowledge as a probabilistic engine it takes a bulldozer to make you accept reality: It is a CAUSAL engine that drives our minds.

5.23.2020 6:36pm - (Replying to @Benoit_Allen @tdietterich and @eliasbareinboim) Hat off. Though note that causal discovery is ONE component in the Causal Science scheme. What you do with what you discover is as important.

5.23.2020 6:30pm - (Replying to @tdietterich @neu_rips and @eliasbareinboim) This is what is happening right now when Data Centers justify their existence with a "causal" decoration. It will become a meaningful term when the scientific contributions of that Center are evaluated through a causal lens, defined by Causal Science Centers.

5.23.2020 5:33pm - (Replying to @elprofe_ivan) What about a data science (ie stat/ML) branch within a causal science center? After all, estimation is part of the Causal Engine, and Rung-1 inference is part of the Ladder of Causation. see #Bookofwhy and,

5.23.2020 5:08pm - (Replying to @richardtomsett @eliasbareinboim and 3 others) Nice beginning. Still, the time is ripe for a scholar who co-leads a special interest group in big data to deliver an introduction to ML and statistics at the University of Oxford "Causal Science Institute."

5.23.2020 4:57pm - (Replying to @tdietterich and @eliasbareinboim) That's not the point. All Centers brag: "We also do CI." Yet there is still not one Center for Causal Science where CI is declared THE focus of research and education and that generates a fraction of the number of data-centric PhD's that a typical DATA SCIENCE CENTER graduates.

5.23.2020 4:45pm - (Replying to @PACESconsulting) Yes, the fact Simpson's surprise disappears when you just change the causal story behind the data, not the data itself.

5.23.2020 11:50am - (Replying to @KateandPie) I perfectly see your point, and I see Ford's horse sh**g in every Data Science Center that I visit. That's the nature of a horse. Can the tiger change his skin? (Jeremiah 13)

5.23.2020 11:09am - I thought so too. So where are the Bill Gates, Jeff Bezos and Mark Zuckerbergs of today? Are visionary business leaders a thing of the past? I am not talking "causality research"; we have it. I am talking Causal Science Centers, to educate next generation Phd's for decision makin

5.23.2020 10:29am - (Replying to @ERMANigeria) Terrific question! I personally believe there should be a deep causal foundation to probability theory, but it's not well developed (IMHO). Funny thing, probabilists think such foundation is unnecessary, exception is Savage attempt to base statistics on decision-theoretic axioms.

5.23.2020 10:21am - (Replying to @anbarbazan) I have commented on Granger Causality here I have seen some sensible papers on CI in time series, but need to search for them.

5.23.2020 10:06am - In theory you are right but, in practice, these CENTERS either don't, or pay lip service to causal science. It's ALL in the name. Once it's named "CAUSAL SCIENCE CENTER" the director must live up to expectations, and "more of the same" turns inexcusable, even with a CI fig-leaf

5.23.2020 8:28am - Our well-intentioned funding agencies are partly responsible for this mis-balanced mis-investment. They all want & fund "explainable AI" "robust AI" "life-long learning" etc etc. forgetting that, once projects are run by traditionally-trained PI's they end up "more of the same."

5.23.2020 8:12am - (Replying to @yguangabroad) The proper set of variables for PS is one satisfying the back-door criterion. Some call it "backdoor admissible" or "sufficient set". It must be someplace in the slides.

5.23.2020 7:26am - It is for this reason that I feel uneasy seeing our society invest almost all its educational resources in "DATA SCIENCE CENTERS" run by good people mindcuffed by traditional statistical and ML paradigms. Not one "CAUSAL SCIENCE CENTER". A terrible long-term (mis)investment.

5.23.2020 6:42am - Readers wrote, and I agree: It is not the paradox that is stubborn but philosophers (and scientists). After getting your PhD and tenure viewing knowledge as a probabilistic engine it takes a bulldozer to make you accept reality: It is a CAUSAL engine that drives our minds.

5.23.2020 1:09am - Remember "Simpson'ts Paradox: The Riddle that Would not Die"? (See I've discovered another article, this time by a philosopher, attempting to do the impossible -- explain the paradox in probabilistic terms: A stubborn paradox!

5.22.2020 9:06pm - (Replying to @yudapearl @Jsevillamol and @causalinf) The "single door" criterion is described here, Theorem 2, together with many other tools applicable to linear systems. It is one of those basic tools that may lift economics education from its current mess.

5.22.2020 8:11pm - A lecturer who called me "white supremacist" was given an award for "leadership" that "transcends the division between scholarship and activism" by @AAUP , an organization claiming to "protect quality higher education." Read: @AAUPUC Dear Mr. Orwell, ....

5.22.2020 2:35pm - (Replying to @hangingnoodles) I like the metaphor: Statistics on steroids. May I borrow it?

5.22.2020 12:38pm - (Replying to @aquariusacquah) I like your figure 3, but I would reverse the arrows. CI is the umbrella that unifies stat., econ. and CS.

5.22.2020 12:26pm - Amazing! Yet with all this power and efficiency ML is still unable to tell us what to do in case we find a drug that's good for men, good for women and bad for a person, or how to handle any of these seven tasks:

5.22.2020 11:56am - (Replying to @Jsevillamol and @causalinf) You are solving a structure discovery problem. Identification assumes structure and seeks to estimate the parameters.

5.22.2020 7:19am - (Replying to @yudapearl @Jsevillamol and @causalinf) The simplicity of this problem entices me to add a couple of poetic versions that stone-age analysts couldn't handle: (1) What are the testable implications of your model? (2) What confounders can be added w/o spoiling identifiability? (3) What arrows can be added?

5.22.2020 3:15am - (Replying to @Jsevillamol) Fun problem, that @causalinf should include as homework: (1) Each parameter is identified using "single door". (2) Each mediated effect = sum of products along the mediated paths. In the stone age, economists solved such problems using matrix algebra. Today it's duck soup.

5.22.2020 2:31am - And for those who feel like singing "Jerusalem of Gold" with me (D minor please), lets go:
"And when I come today to sing to you and adorn your head with crowns, I am smaller than your youngest son, and the last behind your poets." Thanks.

5.22.2020 2:07am - (Replying to @Jsevillamol) Are all variables observable?

5.21.2020 9:23pm - My grandfather took me to Jerusalem when I was 3, I could not visit it again after 1948 (Arabs have funny definitions of "holy" "history" and "sharing"). I returned in 1967, my son was Bar Mitzva there, and every stone sings history to me. Happy Jerusalem Day - City of Magic.

5.21.2020 5:35am - (Replying to @Richie_Research @dingding_peng and @hardsci) The sex of an applicant to a given school can be a function of the education given to boys and girls in that community and of the kind of programs the school is offering.

5.20.2020 2:45pm - What makes Uruguay different than Venevzuela? Good judgement? Commonsense? Perhaps just good luck. I salute you, President @LuisLacallePou

5.20.2020 5:30am - (Replying to @mardelgiu @TimBartik and @causalinf) Well put; we all go for safety. But why WORRY about internal and external validity when mathematics has resolved these two issues? (eg

5.20.2020 4:53am - (Replying to @karlrohe and @omaclaren) Simpson's paradox is in the disputed territory between CI and Statistics. Discovered by the latter and resolved by the former. The part of CI that resolved Simpson's paradox is in AI; statisticians are still perplexed. See and the discussion around it.

5.20.2020 4:41am - (Replying to @dylAntoniazzi and @Richie_Research) On this educational chanel CI stands for Causal Inference. I did not realize the conflict with Confidence Intervals, I hope our statisticians will forgive the intrusion.

5.20.2020 2:11am - Anyone who believes that peace cannot be gotten by blindness to its obstacles should listen (below) to @EinatWilf describing her new book, "The War of Return", which bravely addresses the super giant elephant on the road to peace in the Middle East.

5.20.2020 1:36am - Roger Waters, the man who sang racism to music, finds himself banned from his band's website, lamenting his irrelevance, yet regretting none of his lyrics.

5.19.2020 11:57pm - This Workshop on “Explainable Logic-Based Knowledge Representation” deserves the attention of readers concerned with explainability of ML systems. Reading the abstract already puts you in a curious and humble mood.

5.19.2020 10:40pm - This paper is the first COVID-19 analysis I read that goes beyond data-fitting and tells society: Yes! AI can be trusted to extract meaning from data, and can do so rigorously and practically if properly directed and wisely supported.

5.19.2020 9:37pm - The enthusiastic reception of my slides and homework (800 likes) promises a bright future for CI in education. I went searching for the solutions to the homework. Eureka!! I've found them, and will make them available to any instructor who pledges not to spoil the fun for others.

5.19.2020 12:30am - (Replying to @causalinf) If my slides and homework would help, here they are:

5.19.2020 12:25am - Those who wish to teach a class on my Causality book would benefit from my slides and homework which I am happy to share: Merry teaching, it should be fun.

5.18.2020 12:33pm - (Replying to @TuanAIWS) Thanks for the advance warning, I hope I can tune in.

5.18.2020 12:25pm - (Replying to @mardelgiu and @TimBartik) Thanks, this kind of tasks happened to be my concern too. But look at me, I got sucked into theoretical work after realizing that they cannot be handled with conventional methods. How are non-theoretical economists coping with this realization? Are they using inadequate methods?

5.18.2020 12:11pm - (Replying to @jmorenocruz and @causalinf) Have you tried Primer? It was written to bridge between Causality and #Bookofwhy, and is supported by Daggity software and solution manual. And, beside, it's a truly good book.

5.18.2020 11:04pm - (Replying to @tdetonberry @MacroPru and @causalinf) Many of our readers will benefit from the extension, which will help their efforts to revitalize economics education.

5.18.2020 9:07pm - An Annotated Summary of #Bookofwhy. Milton Masur (MD) sent me this book summary: which I believe would be valuable for both new and veteran readers. It is more detailed than an typical book review and still captures the flow of the main ideas.

5.18.2020 3:23am - (Replying to @AvanteQuora @shriyamite and @TMoldwin) Agree, but this brings us back to the use of mathematical models. What makes an economist a better match-maker of models to contexts than a non-economist? And where does the former get the training and tools to be a good match-maker?

5.18.2020 3:17am - (Replying to @mardelgiu and @TimBartik) I was hoping to learn "what economists actually do" but the meat seems to be under pay-walls and/or high-talks, unless some economist can summarize it all in a typical example.

5.17.2020 3:07pm - (Replying to @TimBartik) Totally agree. And would probably agree more if given a description of what most economists do in fact do, what theoretical tools they use, where they get these tools from, and why amateurs (like me) would fail to accomplish what they do without those tools.

5.17.2020 3:58am - (Replying to @shriyamite and @nntaleb) This exactly why I object to calling statistics "data science", even when supported by "deep learning".

5.17.2020 3:17am - (Replying to @mardelgiu and @TimBartik) I was hoping to learn "what economists actually do" but the meat seems to be under pay-walls and/or high-talks, unless some economist can summarize it all in a typical example.

5.17.2020 12:14am - This brutal indictment of economics has nothing to do with our debate with dogmatic economists. It's in fact an indictment of all mathematical models. It's rich however with juicy quotes, historical anecdotes, and hard facts - worthy of our History Chanel

5.16.2020 11:37pm - Seconding our Pakistani attorney: It is a defining case for the Pakistani State.

5.16.2020 10:34pm - @RashidaTlaib would do well to read a page or two in @EinatWilf's new book "The War of Return", just to see how decent people manage to rise from tragedy, and others labor to perpetuate it.

5.16.2020 1:52pm - (Replying to @Jsevillamol) Thanks for sharing.

5.16.2020 3:36am - (Replying to @yudapearl @PWGTennant and 4 others) Sorry, there was a mix-up in the link. The correct link to my lecture on Epidemiology is this one: Please forgive.

5.16.2020 2:57am - (Replying to @PWGTennant @DavidEdmonds100 and 3 others) My outsider's perspective of progress in Epidemiology are described (with facts and dates) in this videoed lecture, which should also be of interest to readers of our History Chanel.

5.15.2020 11:01pm - (Replying to @fmg_twtr @kinggary and 5 others) Interesting! Why shouldn't conditioning on the PS be the same (asymptotically) as conditioning on those covartiates? Eq. (11.10) in states that they should be the same. Some colliders do NOT induce dependence among their parents, and PS is such a collider

5.15.2020 9:42pm - (Replying to @fmg_twtr @kinggary and 5 others) I am not sure this is the bias that @kinggary referred to, because he said: "even if you choose the right set of covariates". BTW, your graph model should define what is treatment and what is PS, etc.

5.15.2020 8:59pm - (Replying to @kinggary @LydiaManiatis and 4 others) I presume by "model dependency" you mean, dependency on the assumed distributional form of P(X| cov). This dependency is absent when we match directly on the strata of cov. Though other obstacles present themselves, eg. empty cells.

5.15.2020 8:45pm - (Replying to @theBjornErik @causalinf and @itaisher) Thanks for sharing this interesting approach to actual causation. Can you demonstrate the basic requirements for causation (realization, composition, information, integration, and exclusion) on a toy example like the Oxygen-match-fire story, as done in

5.15.2020 4:49pm - I think so too, despite being biased, and I do not see causal thinking dominate the conversation about #covid19 risk factors -- it should, especially when it comes to fusing data from multiple sources, as I argue here:

5.15.2020 1:25pm - (Replying to @FlorianWilhelm and @data_hpz) Sure; when bothered by unwanted ads customers imagine all kind of theories, eg. the seller is stuck with large inventory, or, the seller's profit margin are excessive, etc.

5.15.2020 1:15pm - This blog post aims to democratize debates about COVID-19 policies by demystifying the basic scientific questions that underlie the assignment priorities in health care.

5.15.2020 12:57pm - (Replying to @TuanAIWS) I wouldn't exactly call Ramsey an "thinker in AI", but he was definitely one of the clearest forerunner of subjective probabilities and the revival of Bayes statistics in the 20th century, which influenced the 1970-90 debate on how to represent uncertainty in AI systems.

5.15.2020 5:53am - More for our History Chanel, a new book on the life of Frank Ramsey, a mathematical genius (1903-1930) who, in 1926, defined subjective probability by the odds an individual would accept when betting on an outcome, thus anticipating VNM (1947) and Savage

5.15.2020 2:39am - @data_hpz . I believe the "Campaign Targeting" problem is the same as the "unit selection" problem, a counterfactual analysis of which is given in, combining both experimental and observational data. gives a dynamic visualization.

5.14.2020 7:01pm - Our Spanish speaking readers will be pleased to know that a Spanish translation of #Bookofwhy will be available next month and, moreover, the book was selected as "Libro del Semestre", see We will soon be "on the same page".

5.14.2020 6:18pm - Speaking of AI and Society, my attention was called (albeit belatedly) to this review of #Bookofwhy in AI&Society Journal: It is interesting for me to see how many of the book's issues are related indeed to society expectations of AI.

5.14.2020 11:39am - It turns out that the Guided Tour of AI can be accessed free, through this link, which should give our History Chanel readers direct access to the chapter on causation.

5.14.2020 10:49am - (Replying to @RaulMachadoG) Our works on IV have always assumed non-linearity, as in the instrumental inequality or non-compliance bounds Linearity was assumed in #Bookofwhy for ease of exposition only. As to citations, I cite all original contributors.

5.14.2020 12:54am - Friday, 4 pm, May 14, 1948, Israel declared independence. Wild dancing in the hearts, not in the streets; a day earlier Gush Ezion surrenders to the Arab Legion and 127 defenders executed; a day later we (12 yr olds) found ourselves bombed by Egyptian war planes. What a miracle!

5.13.2020 10:50pm - (Replying to @yudapearl @kinggary and 5 others) Infinite sample is not a matter of wishful thinking, but one of understanding where bias comes from. Asymptotic analysis, fortunately, allows us to separate two such sources: finite sample and wrong set of covariates. The former diminishes with bigger data, the latter does not.

5.13.2020 10:14pm - (Replying to @kinggary @LydiaManiatis and 4 others) The PS weaknesses that @kinggary unveils are orthogonal to those I explicate in The former are weakness of estimation, the latter of identification (infinite sample). Curious: what is "fully blocked randomized experiment" that you compare PS to?

5.13.2020 4:49am - Time for our History Channel. This new book is an encyclopedic bibliography of almost every paper published in AI (skipping pros and cons comments). The chapter on causality would be of special interest to our readers, esp. historians and bibliographers.

5.12.2020 1:39pm - (Replying to @JakeVigdor @causalinf and 3 others) Now I understand where the notion of "pay to publish" comes from. Unheard of in my corner. And let's not forget the "cite to publish" industry. That's how some journals become "high impact": The more you cite the guru, the higher your chances of a favorable review.

5.12.2020 5:43am - (Replying to @KetchupEconomi1 @sterndavidi and @causalinf) There is no "pay to publish". How come so many economists miss my tweets that the fee is OPTIONAL. I can't grasp the perception of "incentive to accept more papers". Are reviewers/editors paid in economics? Not so in JCI. Where do you get this perception from?

5.12.2020 1:19am - (Replying to @lmonasterio) This video made me wish to study Portugese. I hope you liked #Bookofwhy

5.11.2020 10:53pm - Another reason for submitting to JCI: Your idea-paper will be reviewed by idea-friendly scholars, and you won't spend 2-years fighting with causality-blind reviewers, compromising your causal vocabulary to pacify their dogmatic minds. That was the purpose for starting JCI (2013).

5.11.2020 8:56pm - (Replying to @petersuber) No press release, just rewording of the publication policy, after weeks of negotiation with the editors who insisted on making the fee optional, for all the reasons that @causalinf has mentioned. They should reword it, to make the optionality as clear as the word @DeGruyterOA

5.11.2020 6:12pm - Recommended Machine Learning e-lecture from Sydney Australia, describing causal inference as a framework for understanding ML, not the other way around.

5.11.2020 5:07pm - (Replying to @LydiaManiatis @CMichaelGibson and 4 others) I am happy to answer, if I only understood what the question is. You refer to "this technique" and I do not know what "this technique" is. Please describe the idea in a few words, so I can pass it through my mind's filter.

5.11.2020 1:31pm - (Replying to @LydiaManiatis @CMichaelGibson and 4 others) Can you describe "this technique" in conceptual terms? What does it do?

5.11.2020 6:38am - (Replying to @LydiaManiatis @CMichaelGibson and 4 others) The bias reducing power of PS matching is equal precisely to the bias reducing power of adjusting for the covariates that PS deploys. It is proven here

5.11.2020 2:01am - I am retweeting my last contribution in a long debate on the new JCI open access policy (no submission charge, and voluntary publication fee), in which the value of JCI publications to rigid tenure committees was questioned. I encourage all readers to submit idea papers to JCI.

5.11.2020 12:21am - (Replying to @causalinf and @dedubyadubya) Allow me to summarize my contribution to this discussion on the policy and quality of JCI by giving your followers a glimpse at the kind of research questions JCI is entertaining: Econ students will find it a friendly and inspiring amplifier of ideas.

5.10.2020 10:48pm - (Replying to @causalinf @hyperplanes and @dade_us) Correct.

5.10.2020 10:27pm - (Replying to @causalinf @hyperplanes and @dade_us) There are no submission charges, and article processing charges are voluntary. Those who can afford it, pay, and those who can't, just need to say so. And, again, considering the unique culture and ideas nourished by JCI, its a real bargain, unmatched by any contender.

5.10.2020 10:03pm - (Replying to @causalinf @hyperplanes and @dade_us) No, it's like Robin Hood social justice: Take from the guy loaded with fat grants and give to the poor un-tenured professor who has a small grant and big ideas.

5.10.2020 9:56pm - (Replying to @causalinf and @BitsyPerlman) JCI is a high quality peer-reviewed journal. While it cannot guarantee you tenure, it can assure you top reviewers who know CI and appreciate ideas. It can also guarantee you readers who seek CI ideas, knowing they cannot get them in journals run by tenure-minded authors/editors.

5.10.2020 7:16pm - (Replying to @StevePerfocus @eval_station and @BetterEval) As an engineer, I was spoiled into thinking input-output. What is the input to your analysis and is the output, assuming that it goes smoothly?

5.10.2020 6:57pm - (Replying to @causalinf) Scott, There is no submission charge, and the publishing fee is strictly voluntary. So what's all the commotion about predatory journals and tenure committees? Eager to understand.

5.10.2020 2:02pm - (Replying to @prisonrodeo) The submission fee is VOLUNTARY. We need to convince the publisher to highlight this option in capital letters.

5.10.2020 1:55pm - (Replying to @mukdal and @PeterGardenfors) I hope that paper of yours was followed up by many more. Has the belief-revision community been impacted at all by the causal revolution?

5.10.2020 4:39am - Readers will be pleased to note that submission fees for the Journal of Causal Inference will be on a voluntary basis only, meaning that authors will declare how much they are ready to pay, depending on their budget restriction. This, I believe, should achieve truly Free Access.

5.10.2020 3:14am - (Replying to @birkenkrahe) We have invited a new Associate Editor from Germany, who will start in January 2021. Good point.

5.9.2020 6:55pm - Readers of our educational channel would welcome this new paper: It argues that "the incapability of causal reasoning is the reason of DNN’s vulnerability to (adversarial) data manipulations." Remedy: Make adversarial factors explicit.

5.9.2020 6:37pm - A Review of Using Text to Remove Confounding from Causal Estimates: In other words, using Text as a noisy proxy for unmeasured confounders, as in A survey of on-going works and pending problems.

5.9.2020 6:18pm - Peter Gardenfors is a philosopher of mind with whom I interfaced in the hay days of "belief revision" (1996): This paper shows Peter embracing causal reasoning from his broad and interesting perspectives. Welcome @PeterGardenfors

5.9.2020 4:51pm - Users of SEM's are in the habit of estimating the overall fit between data and the hypothesized model. I've never understood why one would do that, when local fitting tests are available (d-separation). This paper seems to combine local and global tests.

5.9.2020 4:12pm - (Replying to @zevkalman) The Journal of Causal Inference is a peer-reviewed journal and, in this respect, is different from arXiv. Thanks for the comment on #Bookofwhy which leads me to ask: How are "network graphs" used in data analysis?

5.9.2020 4:36am - (Replying to @quantadan) I dont believe recognition of one person or another will rescue economics education from its stagnation. Perhaps the Nobel committee should suspend the Prize until such time when 50% of econ PhDs understand indentifiability and testability of (simple) nonparametric econ models.

5.9.2020 3:48am - Announcing new publication policy for the Journal of Causal Inference. The Journal has switched to Open Access. Articles are now published with unrestricted access for everyone, thus boosting visibility and impact. Authors still retain copyrights and there are no embargo periods.

5.9.2020 3:05am - (Replying to @halbfinger) David. "A new target: saving life" has no other interpretation but a defamatory contrast that you probably did not intend to make. You owe us an apology, see why:

5.9.2020 1:32am - (Replying to @birkenkrahe) Let us know if any snug.

5.8.2020 11:03PM - The NYT owes an apology to me and my generation, who were saved by IDF from a promised 'monumental massacre" (Azam Pasha, Nov. 1947), for insinuating, even jokingly, that IDF has any other mission but that of saving lives. Awaiting a public apology. Judea Pearl @yudapearl

5.8.2020 5:49am - (Replying to @sbuhai) Of course there is more to Econ than causal inference. But neglecting causal inference would be judged "20 years behind" by economists recalling the econ aims at policy making. And please, not "your" graphical model again; take ANY model that can do what graphical models offer.

5.8.2020 5:37am - Speaking about "explainable AI", this paper shows that, even in classification tasks, and even after agreeing on a Bayesian Network classifier, answering "why" is not a trivial matter.

5.8.2020 5:28am - (Replying to @sbuhai) Had the top journal in my field not encourage content that I consider important (ie, the content of Paul's slides) I would not consider it an insult to state so, but an invitation to alert the editors to their responsibilities. Can you name a paper using graphical models in ECMA?

5.8.2020 5:02am - (Replying to @sbuhai) Please read carefully. There is nothing in my tweet resembling *all*. It addresses "econ textbooks" and "econometrica" which are 20 years behind. And if forward-looking economists do not shake their establishment, they will remain behind. So why the anger when someone tries?

5.8.2020 4:05am - Recommended for economists and other readers who are asking: Where can we get an on-line course in modern causal inference? Not exactly a "course", but pretty close, and you won't learn it from econ textbooks or econometrika, now lagging 20 years behind the times.

5.6.2020 1:57am - I blush when people call me a "philosopher"; perhaps because to us, engineers, philosophy is just "poetry". Now that I have been called a "poet", I am more forgiving towards philosophers. Thank you, @amalkhan , that tree branch is still peeking at Danny's room, in expectation.

5.5.2020 2:24am - (Replying to @ilan_sinelnikov and @SSI_Movement) Congratulations!!

5.4.2020 10:22pm - #TheodorHerzl. And I almost missed his birthday! Here is his picture from a balcony in Basel, a picture that ignited hearts, moved masses, made deserts bloom, created a state, and saved my family from two infernos. What a balcony!!!

5.4.2020 4:43pm - (Replying to @zeemo_n and @AndrsMontealegr) The more I think about it the less certain I am about the boundaries between science, philosophy and artificial intelligence. Especially if the science you are doing is not conventional and the philosophy you are doing is driven by the puzzle: "So, how do people do it?"

5.4.2020 5:10am - Philip Dawid has written a comprehensive overview of a his approach to CI The do-operator is simulated by a decision variable in a Bayesian Network. The paper illuminates what can be done without counterfactuals, a topic of my paper

5.3.2020 8:41pm - (Replying to @EinatWilf) A tested way to uncover Palestinians' agenda to undo Israel is to give their intellectuals a stage and let them rant freely about Zionism and its quest for coexistence. The truth will come out immediately, on the first page, in perfect English, and in no ambiguous terms.

5.3.2020 5:27am - World leaders: Release imprisoned journalists worldwide - Sign the Petition! via @Change

5.3.2020 5:22am - (Replying to @eyad_nawar and @eliasbareinboim) Determinism disappears if the action is disjunctive, as in "paint the wall either green or red",see From the painter viewpoint the action is deterministic (one choice) but from the policy maker, it is stochastic, not knowing what color will be chosen.

5.3.2020 3:12am - There is a bit more to it than a language/calculus, it is a tractable algorithm that tells us, not merely if a query is estimable, but also how to estimate it and, if it's not estimable, what kind of additional measurements or experiments will turn it estimable. It was undersold.

5.3.2020 1:01am - (Replying to @ExogenyKarl and @eliasbareinboim) The paper you sited is one of my favorites, but I think you missed the punch-line: "economics students should now be able to solve the eight toy problems I posed in Pearl, 2013 (see Appendix A, Section A.2)." Do you think it has happened?

5.2.2020 10:30pm - Readers who have been shunning the do-operator for being "too deterministic" can now embrace its stochastic cousin, "soft intervention", formalized in a recent paper by Correa and @eliasbareinboim :

5.2.2020 5:32pm - (Replying to @agpatriota @guilhermejd1 and 4 others) It sure can be captured. E(Y|do(X=1)) = E(Y|do(X=0)) and E(Y|do(X=1),Z=0) =/= E(Y|do(X=0), Z=0) See Causality page 35-36

5.2.2020 4:57pm - (Replying to @RanaBil10485167 @AsraNomani and @UmarCheema1) Justice to Daniel Pearl will usher justice to all innocent victims of brutality and extremism.

5.2.2020 4:50pm - (Replying to @Taks87) We hope that justice to Daniel will usher justice to all innocent victims of extremism

5.2.2020 4:41pm - The parents of murdered Wall Street Journal reporter Daniel Pearl asked Pakistan’s Supreme Court to prevent the men convicted of his abduction and murder from going free via @WSJ

5.2.2020 5:30am - We wish to thank The Pearl Project @AsraNomani for leading the effort to realize justice for Danny, the Committee to Protect Journalists @pressfreedom for their spirited support of this effort, & the Wall Street Journal, for contributing half the legal fees needed for our appeal.

5.2.2020 4:52am - (Replying to @rameshfilms) Remember you very well, and the documentary too, with Christiane Amanpour @camanpour . Glad you are going to screen it again soon; the world needs a reminder.

5.2.2020 3:54am - Today, my wife and I filed an appeal to the Supreme Court of Pakistan to overturn the "acquittal" of our son's murderers. We ask all people of conscience to support our efforts to protect journalists from imitators of Omar Sheikh and his ilk. For details

5.2.2020 1:33am - (Replying to @ExogenyKarl and @RayDalio) Thanks, I'll try to refresh:

5.2.2020 12:44am - (Replying to @mribeirodantas @tdietterich and @StanfordHAI) I do not think the good people at Microsoft and Google have the institutional clout to overturn the deeply entrenched thinking of the data-fitting culture. Recall, it is quite traumatic to grasp the idea that "interpretation" means going beyond the data and that it is necessary.

5.2.2020 12:22am - Could @RayDalio be the visionary executive who will save ML from itself by establishing a "Center of Data-Interpretation" at the University of XYZ, to counter the "data fitting" hype? see If Microsoft and Amazon won't, why not Wall Street? Dead serious!!

5.2.2020 12:03am - (Replying to @matt_vowels) Strangely, I had a similar experience, and so did many readers. Doubly strangely, the authors of the right book write: "we have not found this approach to aid drawing of causal inferences." You can't aid those who refuse your aid, but you can aim to understand of the refusal.

5.1.2020 7:53pm - (Replying to @kerstingAIML @mitbrainandcog and 4 others) Which "we" are we disagreeing on?

5.1.2020 7:33pm - (Replying to @bjh_ip) Aggree, and I am working on it.

4.30.2020 5:55am - A great article by David Suissa which, although does not target Dianne Lob explicitly, provides an in-depth analysis of her mentality, and the mentality of other "Jews of discomfort", of whom I wrote here:

4.29.2020 11:12pm - (Replying to @KyleCranmer @tdietterich and @StanfordHAI) Simulators are causal, and every causal model (completely specified) can be used as a simulator. But I can't see how DL can act as surrogates for these models, since the inputs to DL are data, not the ropes behind the data, and the inverse mapping data-->ropes is not unique.

4.29.2020 1:00am - A fairly harsh indictment of AI: Why hasn't AI had more impact? I have asked my colleagues at @StanfordHAI the same question:, but I am not sure they took notice. People blame the noisy data; shouldn't AI outsmart the noise-makers?

4.29.2020 12:15am - (1/ ) (Replying to @mribeirodantas @tdietterich and @StanfordHAI) It will happen when an enlightened executive of Microsoft or Amazon, etc., will realize that data-fitting in addictive, and building another "data-science center" will only worsen the addiction, and by the time industry will need to switch to "data-interpretive science"
4.29.2020 12:29am - (2/ ) (Replying to @yudapearl @mribeirodantas and 2 others) there will be no one to hire for the job. Nadda. The enlightened executive will then say: Let's establish a "Data-Interpretation Center" at the University of xyz, and endow it with 0.01% of the resources now pouring into data-fitting centers. Get ready, it will happen soon.

4.29.2020 12:10am - And the eyes of the world turn again to the Supreme Court of Pakistan and ask: What message will it send to their sons and daughters?

4.28.2020 6:39pm - To readers who keep on asking: what's the secret of looking so good at 72, and of feeling so happy despite 72 years of siege, the answer is simple: Its plain 1978 years of waiting, that's all it takes, just 1978 years.

4.28.2020 2:23pm - (Replying to @YFeyman @raj_mehta and 4 others) ignorability is "as if randomized, given Z" and exclusion is "going only through X". Two independent restrictions that even seasoned IV experts can't conceptualize w/o graphs, see why:

4.28.2020 2:07pm - (Replying to @YFeyman @raj_mehta and 3 others) Honestly, this is the first time I hear the adjectives "clear" and "tractable" attached to the PO framework. It makes me extremely curious: can you perhaps show us how you judge the plausibility of a simple ignorability assumption, like @stuartbuck1 :

4.28.2020 1:47pm - (Replying to @raj_mehta @YFeyman and 3 others) Good point. I could imagine conversations like those we are witnessing on Twitter going on w/o the language of DAGs to facilitate them. Even devout antaginsts of DAGs are using expressions such as "there are many paths from X to Y" "going only through W" etc. It speaks a lot.

4.28.2020 3:46am - Another celebration starts today at sundown - Israel's Independence Day. Join me in celebrating the most inspirational miracle of the 20th century: A tribe of beggars and peddlers lifting themselves from the margin of history to become a world center of science, art and business.

4.28.2020 4:24am - (Replying to @richarddorset) Many oppressed peoples, Yazidis, Asyrians, Kurds, even Tibetians are inspired by the Israeli experiment and studying the secret of its creation. The secret: Jews were not a "nomadic tribe" after all, but a dormant nation, driven daily by the dream of returning to their homeland.

4.28.2020 4:46am - (Replying to @yudapearl and @richarddorset) For a good book on the creation of Israel, I'd recommend Benny Morris "1948", though it may not meet your criterion of "new state based on ancient claims". For my grandfather it was "old state based on daily dreams". He prayed 3 times a day "return us in sovereignty to our land."

4.28.2020 3:16am - (Replying to @maximananyev @analisereal and 6 others) I think we can generalize your observation to read: applied econ use graphical metaphors quite often, albeit shyly and tacitly (eg see textbook descriptions of IV), and the language of diagrams can improve their practice by encouraging them to make those intuitions explicit.

4.28.2020 1:05am - Congratulations to @PHuenermund who will be joining the Editorial Board of the Journal of Causal Inference (JCI). Paul's addition is an open invitation to all economists to embrace JCI as an effective publication for disseminating ideas across disciplines.

4.27.2020 9:25pm - (Replying to @amt_shrma @autoregress and 8 others) The assignment need not be "random". It can be by student's birthdate, or by results of an eyesight test, as long as it is not correlated with U and e_y.

4.27.2020 2:08am - Update on Israel Memorial Day. The sirens will sound today (Monday) at 8:00 pm Israel time (1 pm EST,10 am PST). Join me in freezing (1 min) for Dani Balkore, who went to fight five invading armies in May of 1948, and made it possible for my generation to live a semi-normal life.

4.26.2020 9:22pm - (Replying to @patrickkloesel and @PHuenermund) Based on my many conversations, the percentage is around 95%. And, more importantly, it is not that they do not understand DAGs, they are committed to not understanding them, expressed by: "we have not found them useful in our type of applications".

4.26.2020 4:40pm - (1/ ) (Replying to @stuartbuck1 @PHuenermund and 2 others) I confess to not knowing everything about "how economists think and reason about ignorability" because my knowledge comes primarily from published articles, which may not reveal the entire process through which authors "think". I would therefore be grateful to you if you could
4.26.2020 4:52pm - (2/ ) (Replying to @yudapearl @stuartbuck1 and 3 others) help me, and many other readers, fill in the key principles of that process, by sharing with us how YOU think about ignorability. I capitalized YOU because, economists may vary in their intepretations of the term. So, if you are willing, let us start with the simplest
4.26.2020 5:00pm - (3/ ) (Replying to @yudapearl @stuartbuck1 and 3 others) 3/ "conditional ignorability" statement, Y(X) || X | Z, where X is a treatment, Y the outcome, and Z a set of covariates that are candidates for making the statement true. Can you walk us through the process by which you would judge whether the statement holds in a given problem
4.26.2020 5:07pm - (4/4) (Replying to @yudapearl @stuartbuck1 and 3 others) or scenario, or a model, or a story, or whatever helps you discern whether the statement holds true or not. No need to be pedantic on issues such as "strong ignorability" vs. "weak ignorability", or how Z was chosen. Just examining the problem and deciding true of false.

4.26.2020 6:52am - Good catch. This link should work I hope:

4.26.2020 6:27am - I am in receipt of a comprehensive paper on the role of causality in vision Although I find the section on "causal Bayes Nets" a bit incoherent, there is a lot I can learn from it, especially the interplay with naive physics and object-oriented reasoning.

4.26.2020 3:41am - Caught me by surprise! But we must add that, with the exception of a few dragonic islands, the dragons were eventually overcome by angels of commonsense.

4.26.2020 5:52am - (Replying to @PHuenermund) Not at all. It was in perfect timing; just a few days after the angry dragons have issued a non-ignorable ultimatum: dragonize or perish.

4.26.2020 3:47am - 100th anniversary !!! And I almost missed it !! To readers who have not heard of the San Remo Resolution, or the Balfour Declaration that triggered it, I wrote an oped on the latter which would be illuminating to read on this historic day.

4.26.2020 2:28am - ECOL VS. ECON While economists are moving from models to field experiments, ecologists seem to be going the other way. This paper:'s_guide_to_developing_explanatory_statistical_models_using_causal_analysis_principles/links/5e8e72c9a6fdcca78901f151/Scientists-guide-to-developing-explanatory-statistical-models-using-causal-analysis-principles.pdf seeks the advice of causal models to "guide scientists in their quest to develop explanatory hypotheses for evaluation."

4.25.2020 4:09pm - At sundown today, siren sounds freeze Israel into a 2-minute silence in remembrance of its fallen - In May 1948, our next-door neighbor, Dani Balkore (19), kissed his family, smiled to us, and returned in a coffin two weeks later. I freeze for you Dani.

4.25.2020 2:46pm - (Replying to @LennyVds) I would not phrase it exactly as you did. I would say: The power of SCM is that you can compute the likelihood of EVERY counterfactual sentence. For example, Y would be less likely had X not happened, given that X actually did happen.

4.25.2020 2:27pm - (Replying to @aulderic @Liv_Boeree and @robertwiblin) Agree. The birth-weight paradox should not be classified as an example of "Simpson's paradox". The surprise in the two paradoxes is of different nature. Someone should fix Wiki, and add "not" to: "it can NOT be explained statistically".

4.25.2020 6:13am - Israel Yom Hazikaron National Online Memorial Ceremony 8:30EDT via @israeliamerican

4.25.2020 1:15am - (1/3) Here is a golden opportunity to learn something of lasting value during our COVID-19 hibernation. There have been dozens of articles, dialogues, discussions, even tantrum debates, laboring to compare DAGs and PO frameworks, mostly written by distant commentators, who could
4.25.2020 1:15am - (2/3) not expose first-hand what it takes to solve a problem from beginning to end using one framework or another. This link will do it for you, in just a few paragraphs, using a simple chain X-->Y-->Z. Once you read these paragraphs you will never
4.25.2020 1:15am - (3/3) be intimidated by an invitation to compare the two frameworks. You might even become curious as to why some people would prefer one framework over another.

4.24.2020 10:21pm - (Replying to @maximananyev @Jabaluck and 7 others) I am not a bit surprised. "front-door" is one of many models involving 3 variables and one latent confounder, so i am sure it came up naturally in many applications. The fact that it permits nonparametric identification, came as a shock to statisticians (narrated in #Bookofwhy)

4.24.2020 2:29pm - (Replying to @HannesMalmberg1 @Jabaluck and @maximananyev) You are beginning to see why I called it "undersold". But the gains are not merely educational, they are methodological and computational. Tasks that are intractable in PO (consistency, redundancy, testability, etc) suddenly become as easy as writing down an SEM. Sadly undersold.

4.24.2020 2:17pm - (Replying to @HannesMalmberg1 @Jabaluck and @maximananyev) When I see the word "design", I see a potential for an exciting PhD project: Let's take the mental model that gives the "designer" a feeling of "design", formalize it, and show the designer how to design things better. This was done in several key areas, and science progresses.

4.24.2020 2:07pm - (Replying to @autoregress @PACESconsulting and 4 others) Thanks for adding a new acronym to my arsenal. The missing-data literature refers to your CMAR as MAR. But there, "missing" means unobserved variabls in some individuals, not missing individuals.

4.24.2020 1:56pm - (Replying to @autoregress @PACESconsulting and 4 others) What is CMAR? Is it Missing Completely at Random MCAR, as used here If so, my answer is probably YES. Most non-parametric problems in economics now have solutions. But it was undersold. Eco. was not fortunate to have a leader saying: Look guys, Its Great!

4.24.2020 1:29pm - (1/ ) (Replying to @The_RickMc @PHuenermund and 3 others) This is an important point. What make us judge a treatment as "as-if random" when we have no coins nor lotteries? It is our subjective inability to conceive of a confounding mechanism. Take the price of beans in China and the traffic in LA. None is random.
4.24.2020 1:40pm - (2/ ) (Replying to @yudapearl @The_RickMc and 4 others) Each varies systematically in its own dynamics. Yet, unless you are a smart alec and argue that some future-broker in LA may panic and call his clients to rush to their cars, the two variables are judged to be unconfounded. Randomization is sufficient, but not necessary.

4.24.2020 1:20pm - (Replying to @PACESconsulting @gelbach and 4 others) My first sentence says that for decades Heckman's selection-bias correction was tied to strong parametric assumptions. Nothing wrong with it. But if we want to snap out of such assumptions we need the logic of non-parametric models. That logic was a "novelty", highly undersold.

4.24.2020 11:00am - (Replying to @PACESconsulting @gelbach and 4 others) New results after decades of stagnation do demand new tools. Specifically, the idea that selection bias can be reduced or eliminated without making parametric assumptions demands tools to handle non parametric models. Why the resistance to novelty?

4.24.2020 10:49am - Measurement errors are represented as proxy variables, as shown here . The do-calculus just operates on the graph after these proxy nodes are added. In most cases identification requires auxiliary machinery, as shown here

4.24.2020 2:24am - For the many who inquired on the latest in the movement against the "acquittal" of our son's murderer, this article offers a fairly comprehensive report. Thanking all readers for their help and empathy.

4.24.2020 2:21am - To our readers in Israel, The Israeli Association for Artificial Intelligence has announced a Doctoral Dissertation Award. Details are here:

4.24.2020 1:37am - (Replying to @teemu_roos and @UusitaloLaura) You are so right, skipped my mind and my search engine. It is indeed in Which, BTW, re-reading it 32 years later, is a damn good paper!. I don't recall if it was folklore. The firing squad was folklore (mentioned in I. J. Good)

4.24.2020 1:25am - From the day I was told that RCT's require "careful design", I suspected that RCT's experts must be using a mental causal model. Here is a paper that supports that suspicion: and shows experts what they can gain by making that mental model explicit.

4.24.2020 12:32am - (Replying to @Jabaluck @PHuenermund and 2 others) Luckily, I have not met a serious economist who thinks I am a charlatan. They know what you know, that my "guilt" is only in calling to make assumptions explicit in a language that is most meaningful to researchers (SCM) but became unfashionable in 1980, yet producing new results

4.23.2020 10:51pm - (Replying to @yudapearl @PHuenermund and 3 others) And there is a good reason why justification is avoided. Justification boils down to saying: "I cannot think of any factor that would affect both X and Y." It is tabooed on 2 counts: (1) It's dreadfully subjective, (2) It invokes"factors affecting," almost like graphical criteria

4.23.2020 10:43pm - (Replying to @PHuenermund @Jabaluck and 2 others) Moreover, students (and their advisors) reading the final papers get the impression that as soon as you say: "we assumed ignorability" and cite revered authors who also did so, you are a scientist. The ratio of such papers to justifying paper is 100:1 (based on mental counting).

4.23.2020 10:28pm - It is always a thought provoking exercise to imagine how the general public would react to issues that we have been debating on Twitter with such intensity. One thing is clear, causality is gaining stature as an independent species of scientific inquiry.

4.23.2020 9:49pm - (Replying to @maximananyev and @Jabaluck) You are almost right. Applied econ. will challenge every assumption that is explicit, but would rarely challenge those that hide under fashionable technical titles such as "ignorable", which hardly anyone knows what it means. It's the old peril of being honest and transparent.

4.23.2020 9:26pm - (Replying to @Jabaluck @autoregress and 3 others) What's the point? Creative people can always find a link between the price of beans in China and traffic in LA. If they believe the link is strong, put an arrow and continue. If its weak enough to ignore, remove the arrow and continue. W&C did the latter and failed to continue.

4.23.2020 9:10pm - (Replying to @Jabaluck @autoregress and 3 others) The only economist I met who thought "carefully about DAGs" was the late Hal White. If you know more, they were not here: when I asked them to answer a few elementary questions, like which parameter can be identified by OLS, or which model is testable.

4.23.2020 8:53pm - (Replying to @Jabaluck @autoregress and 3 others) There are many ways to justify absence of an arrow. For me, it is enough to say XWY are medical factors and Z's behavior is a bureaucratic one. Done. If we must cheer Angrist for his insight, fine, but we still need to decide if the query is identified.This is where WC fumbled.

4.23.2020 7:47pm - Great initiative, and a very informative window into the enterprise of "Evidence Based", which I never understood. Speaking of evidence from population to individual, the idea of combining experimental &observational data should illuminate the discussion:

4.23.2020 4:07pm - (Replying to @autoregress @gelbach and 3 others) If treatment Z is administered by a clerk who follows some protocol on how to respond to symptom W, then this clerk is shielded from the confounder U, which affects patients, not clerks. I'm sure you have clerks in economics?

4.23.2020 3:57pm - (Replying to @Jabaluck @autoregress and 5 others) But to assure consistency among a set of dependence-independence assertions, we must assume that they emanate from some, possibly unknown distribution.

4.23.2020 2:35pm - (Replying to @Jabaluck @autoregress and 5 others) This is the beauty of DAGs. You do not commit to a specific structural model. You commit only to the class that is compatible with your knowledge, namely, the source of variation of each variable. This is many time more meaningful than "which PO is ignorable conditional on whom"

4.23.2020 2:21pm - I surely read it, and commented on it here:

4.23.2020 2:12pm - (Replying to @Jabaluck @autoregress and 5 others) I have a friend who "understands" the connection between independence and probabilities but he would never put the definition X||Y iff P(X,Y)=P(X)P(Y) on paper, or in words. His papers are full of X||Y and X||Y|Z sentences but no mention of P. His disciples claim he "understands"

4.23.2020 1:25pm - (Replying to @autoregress @guilhermejd1 and 5 others) If "Yup" means "Yes" then we disagree on the 2nd "Yup". Imbens-Rubin-Angrist not only do not mention this connection in their books, they resist the connection like a plague. But I welcome your first "Yup", and hope that we can agree on its logical implications

4.23.2020 1:09pm - (Replying to @autoregress @guilhermejd1 and 5 others) I did not say "ARE" causal models. I said "emanate from" a structural model, or, "can all be derived from" a structural model. Do you buy that? Do Imbens-Rubin-Angrist-Peschke buy that? This is the 1st Fundamental Law of causal inference.

4.23.2020 12:48pm - (Replying to @guilhermejd1 @autoregress and 5 others) "Embrace"? You must be kidding. I guess the idea that potential outcomes are intrinsic properties of SCM has not been sunk in yet. Every SCM model assigns a probability to every conceivable potential outcome. Potential outcomes are rooted in SCM, and trees w/o roots wither.

4.23.2020 4:53am - (1/ ) I am retweeting this post, knowing that hundreds among our readers are intimidated by colleagues on the issue of "real-world" example. As if going to Somalia and conducting field experiments elevates you to a new level of understanding. The post demonstrates a "real-world"
4.23.2020 4:53am - (2/ ) problem that was mishandled by prominent investigators precisely because they dismissed the tools of causal inference. In my follow-up Tweet: I ask: How many "real world" problems are mishandled by those who talk "real world" to dismiss CI tools?

4.23.2020 1:48am - (Replying to @yudapearl @autoregress and 4 others) Now, Wermuth and Cox are not DAG fanatics. They posed this problem b/c it came up in their investigation of sequential treatments. They wrote 3 articles on this problem, which involves only 5 variables. What does it tell us about economists and their "real world" arguments?

4.23.2020 1:29am - (Replying to @yudapearl @autoregress and 4 others) More on "Real-world" herrings. I was asked repeatedly to give one "Real-world" example which do-calculus can solve, and which "real-world" experts could not. Here is one,, called "indirect confounding" by Wermuth and Cox, also discussed in #Bookofwhy p.241

4.23.2020 12:59am - (Replying to @autoregress @gelbach and 3 others) Every setting that you would be willing to call "real-world" can be solved by these problems, or through the general theorems that these problems illustrate, or labeled "unsolvable" by the completeness theorems that follow (in later papers)."Real-world" boasting is a red herring.

4.23.2020 12:45am - (Replying to @SylvainCF @pa_chevalier and 17 others) I am embarrassed to be on the same list as the great Hume. But, given what I say about him and about counterfactuals on page 266-9 of #Bookofwhy, perhaps he would accommodate my presence on the list.

4.23.2020 12:27am - (Replying to @Jabaluck) The trouble comes when you assume what you need BECAUSE you need it, not because you know it. And this indeed is the common case when PO folks assume conditional ignorability, save for RCT's, and some IV's settings, when natural lotteries are available.

4.22.2020 11:43pm - (Replying to @autoregress @analisereal and 3 others) How can you say that Pearl ignore monotonicity? See my discussion on monotonicity with Imbens Not only can SCM articulate monotonicity when it is plausible, but we can also test it when it is in doubt (See Causality). We do not shun tools, we use them.

4.22.2020 11:02pm - I like your crisp distinction between stating what one "needs" vs. stating what one "knows". It really summarizes the entire debate between the CI and PO frameworks. In CI we use structure to encode what we know, PO to encode what we wish to know, and logic to connect the two.

4.22.2020 10:47pm - (Replying to @analisereal @autoregress and 3 others) I have seen some. Imbens and Rubin (2015) call these identifying assumptions "The Science."

4.22.2020 10:35pm - (Replying to @autoregress @gelbach and 2 others) Puzzled. How could economists possibly have fought this or a similar war, if they did not have a 3-rung causal hierarchy in mind, nor complete tools for managing each rung. Can you point me to any similar war? Was "moving on" a resolution or "give up, lets do what we do best???"

4.22.2020 9:08pm - (Replying to @gelbach @guilhermejd1 and 3 others) Can you use Heckman's correction to solve any of the siimple problems solved here: ???. As I stated before. CONCERN - yes. Results - ??? And I dont blame you, new results demand new tools, which your students will have to acquire by independent reading.

4.22.2020 8:56pm - (Replying to @jeffreywatumull) No, this kind of assumptions have no testable implications and are needed as input to causal calculus.

4.22.2020 8:43pm - (Replying to @maximananyev @autoregress and 5 others) Can you summarize how @ecnomeager represents disparities between two populations.

4.22.2020 8:38pm - (Replying to @guilhermejd1 @autoregress and 3 others) Agree with you completely. Economists have indeed been "concerned" with these questions, as were most scientists since Campbell etal. But when you ask them to solve a tiny problem, eg you see how far behind they have allowed their "concerns" to take them.

4.22.2020 8:14pm - (Replying to @Jabaluck @gelbach and 2 others) Can we put aside me and my failures to engage. Let's return to economics "embracing" machine learning, and examine what precisely has economics "embraced" in the past 3 decades.

4.22.2020 8:09pm - (Replying to @autoregress @Jabaluck and 2 others) See my reply to Imbens who made similar points.

4.22.2020 8:06pm - (Replying to @Jabaluck @gelbach and 2 others) It is more than a question "easier". It is one of "intractability" which in practical terms means "impossible". And let's leave DAGs aside. I am insisting only on transparency, and would welcome a transparent alternative to graphical models.

4.22.2020 8:00pm - (Replying to @Jabaluck @gelbach and 2 others) I seriously doubt that you will that that flexibility. Your students will, but not those who grew up in the insular green house of 1990's econometrics.

4.22.2020 7:51pm - (Replying to @Jabaluck @gelbach and 2 others) Forget discovery. I want economists to do what they have been advertising to be doing since Haavelmo, before being hijacked by statisticians who believed that conditional independencies among counterfactuals is the "science" that researchers carry in their minds while modeling.

4.22.2020 7:41pm - (Replying to @Jabaluck @gelbach and 2 others) Now you are generalizing. I never insist on "my work", this is your hang up. I insist only on starting with cognitively meaningful and defensible encoding of what we know, combining it with data, and formally deriving what we wish to know. Has any of your model papers done that?

4.22.2020 7:13pm - Thanks for this positive and informative reviews. I found a short link here This is, if I am not mistaken, the first review of #Bookofwhy in a flagship scholarly journal. I hope to see more soon, so that readers of BOW will begin to feel mainstream.

4.22.2020 6:52pm - (Replying to @Jabaluck @gelbach and 2 others) Can we get to the causal part?. These papers are still in the data-fitting paradigm, taking "conditional ignorability" as a God given gift.

4.22.2020 6:37pm - Thanks for the feedback. And, now that you are at it, please read some of the antagonistic critics, as many as you can, to truly appreciate the stifling paradigms that #Bookofwhy has attempted to shift.

4.22.2020 6:28pm - (Replying to @Jabaluck @gelbach and 2 others) Tell me more on how economics has embraced machine learning, in substance please, not in hype. And please focus on that part of machine learning which is not curve-fitting, namely computer-aided statistics. FYI, CI folks consider CI to be part of ML, albeit above Rung-1

4.22.2020 6:22pm - (Replying to @Jabaluck @gelbach and 2 others) Beg to differ. The resistance generated was not to overselling novelty but to my insistence that new tools are needed. Challenge: show me one oversold novelty in, or in:

4.22.2020 12:47pm - (Replying to @gelbach @Jabaluck and 2 others) The stagnation we see in certain (unnamed) disciplines calls for courageous and rebellious spirit, not for professional humility (or timidity), which has led to the insular stagnation.

4.22.2020 12:36pm - (Replying to @jeffreywatumull) Yet, unless we suspect that subjects response was affected (positively or adversely) by the very idea of being selected for treatment, it is reasonable to take the un-selected subjects as control group.

4.22.2020 12:30pm - Kudos to Sacks et al for creating the CRAN program, which should be used together with their Biometrika paper - A comprehensive account on computing tight bounds for causal effects.

4.22.2020 12:28am - (Replying to @RemyLevin @HeerJeet and 2 others) I am not familiar with a lot of cross-disciplinary critiques, except the ones I have articulated here and here #Bookofwhy. Speaking of these critiques, I see concrete examples, formal proofs and powerful tools - not one shred of dishonesty or X-measuring!

4.21.2020 11:13pm - (Replying to @RemyLevin @HeerJeet and 2 others) Nice and rosy. Still, you would agree, I hope, that some fields, at some points in time, could benefit from some jolt of self reflection. And the jolt may not always come from inside, given the self-perpetuating structure of our academic culture.

4.21.2020 11:00pm - (Replying to @Jabaluck @autoregress and 2 others) Sorry, I did not realize that. So, back to waiting for a prominent economist to take a sober look at the education and culture dominating the field.

4.21.2020 7:11pm - Responding to critical yet constructive reviewers, we have revised our "generalizing experimental results.." paper:, and we are happy to report new discoveries concerning bounds and new Bayesian estimation of real data on Vitamin A supplement.

4.21.2020 4:21am - (Replying to @RWerpachowski and @Spinozasrose) I did not think you said it in a mean way. My point still is that, in order for the memory of the Holocaust to be long lasting we must tie it with revival and rebirth. And I do not think the directors of our 60 Museums understand it; you can hardly see a glimpse of Israel in any.

4.20.2020 4:18pm - (Replying to @RWerpachowski) My grandparents failed to escape from the train, and I know that they wanted to be remembered not by the train, but through their grandchildren lifting themselves from the ashes and carving a future of hope and dignity for themselves, their people, and their descendants.

4.20.2020 3:52pm - This picture, not the cattle trains nor the gates of Auschwitz, should decorate the entrance hall to each one of the 60 Holocaust Museums in the US.

4.20.2020 3:27pm - (Replying to @HeerJeet @guacamolebio and @beyerstein) I've never dreamed to hear this echoed in the middle of the "credibility revolution." #Bookofwhy

4.20.2020 4:21am - (Replying to @RWerpachowski and @Spinozasrose) I did not think you said it in a mean way. My point still is that, in order for the memory of the Holocaust to be long lasting we must tie it with revival and rebirth. And I do not think the directors of our 60 Museums understand it; you can hardly see a glimpse of Israel in any.

4.20.2020 10:12pm -

4.20.2020 4:21am - (Replying to @RWerpachowski and @Spinozasrose) I did not think you said it in a mean way. My point still is that, in order for the memory of the Holocaust to be long lasting we must tie it with revival and rebirth. And I do not think the directors of our 60 Museums understand it; you can hardly see a glimpse of Israel in any.

4.20.2020 10:12pm - (Replying to @VUspenskiy and @andrewheiss) Actually, the sequences of do-calculus rules have been systematized into an algorithm that gives us the result without going through the rules: But when I look at a diagram, I first ask if it contains one of the recognizable patterns stored in my mind.

4.20.2020 10:01pm - Naive questions are everyone's questions. Ans. The backdoor and frontdoor are logical consequences of do-calculus. So why do we decorate them with names? Because they are easily recognizable in the DAG, so we store them explicitly in our arsenal of tools, so skip re-deriving them

4.20.2020 4:53pm - (Replying to @EinatWilf) Does anyone understand why the American Jewish Committee would not acknowledge the dreams and ideas that inspired the Warsaw Ghetto Uprising?

4.20.2020 4:18pm - (Replying to @RWerpachowski) My grandparents failed to escape from the train, and I know that they wanted to be remembered not by the train, but through their grandchildren lifting themselves from the ashes and carving a future of hope and dignity for themselves, their people, and their descendants.

4.20.2020 3:52pm - This picture, not the cattle trains nor the gates of Auschwitz, should decorate the entrance hall to each one of the 60 Holocaust Museums in the US.

4.20.2020 2:25pm - (Replying to @danilobzdok @shakir_za and 3 others) Agree. It's an excellent Sunday read. But the message of this paper was "a reminder of how easy it is to fall into a web of paradoxical conclusions when relying solely on intuition, OR solely on statistics". We must snap out of statistics to get it right.

4.19.2020 2:15pm - Conditioning just gives us a more specific class of individuals, with all the problems of sparse data; it is still not the individual itself. This paradox highlights the difference, and the role of counterfactual reasoning in getting to the individual.

4.19.2020 12:41am - (Replying to @Physical_Prep) Not surprised that the #sportscience literature suffers from similar myths. ACE is merely one evidence for what we wish to know about an individual. See eg our recent blog on which individual is "in greater need" for a hospital bed.

4.19.2020 12:32am - (Replying to @omaclaren and @unsorsodicorda) I am genuinely trying to learn from the vast literature of DE. Can you give me the simplest example you have, time invariant, where the task of inferring effects can benefit from the teachings of DE? I've just given you one from which qualitative DE folks can benefit.

4.19.2020 12:12am - (Replying to @omaclaren and @unsorsodicorda) This sounds exactly like CI. In which case can DE folks get the causal effect of X on Y, given data taken from Pr(X,Y,Z) if the input is three unknown functions, x=f1(z,u1) y=f2(z,x,u2) z=f3(u2) and u1,u2,u3 are independent yet arbitrarily distributed random variables? How?

4.18.2020 11:57pm - My current interest in "personalized medicine" compelled me to read this new paper I must conclude though that, despite 3 decades of CI, "personalized nutritionists" still have to accept that there can be no "personalized" inference w/o counterfactuals.

4.18.2020 11:33pm - (Replying to @unsorsodicorda and @omaclaren) Can you summarize some of this literature in input-output language, as we folks do with CI? What must the input be like + what kind of data is taken and what kind of answers we get from the analysis?

4.18.2020 11:19pm - (Replying to @seema_econ) Yes, there sure is "a little virus going around"!

4.18.2020 7:14pm - (Replying to @omaclaren) I would be very happy to enlist the work done in CI as a modest contribution to the vast umbrella of DE, assuming the latter is open-minded. For example, the adjustment formula can easily be viewed as an exercise in qualitative DE, so is the Instrumental Inequality etc. etc. etc.

4.18.2020 4:09pm - Readers suggest that this tutorial video may well serve as an introduction to an online course in CI. Unfortunately, the videographer was sloppy on the slides, so I am posting the slides here: Feel free to use them in any class.

4.18.2020 3:30pm - I am hearing great passion for ODE, so I ask myself: "Why not use ODE for causal inference?" One answer: We don't have the functions that make up ODE. All we have are some qualitative properties, eg. who are the arguments of each function, and we still venture to estimate effects

4.18.2020 4:20am - Another exciting paper arriving at my desk reads: Causal Relational Learning: which promises to revolutionize causal inference the same way first-order predicate logic has transformed Boolean logic.

4.18.2020 3:36am - I am always happy to see "completeness" results in CI, but these results are especially gratifying, because graphical modeling of missing data problems ( have totally been ignored by MD practitioners. I hope things will change soon.

4.18.2020 3:04am - On the basis of the first 2 accessible sections, this is the best on line course on the market. Highly recommended. Especially for economists!!!

4.18.2020 1:17am - Here is a fun question that came up in conversation with my grandson. If it is bouncing off the moving piston that speeds up molecules as gas is compressed, how come slow-moving and fast-moving pistons result in the same temperature increase? A question for AI reasoning systems.

4.18.2020 1:04am - (Replying to @attilacsordas) What about "music breaks the silence". For every thing that is created the absence of that thing is destroyed.

4.18.2020 1:00am - (Replying to @shaul_ido) Trying to find one

4.18.2020 12:56am - (Replying to @mribeirodantas) I love it too. And this paper describes a machine that generate such data on demand.

4.18.2020 12:49am - (Replying to @firstmn59) Good for Rung 1 and Rung 2 (interventions), weak on Rung 3 (counterfactuals and SCM), so I would supplement it with chapter 4 of Primer:

4.18.2020 12:46am - (Replying to @juli_schuess and @Jean_DeCarli) Highly recommended.

4.18.2020 12:40am - Your question is a good one, and it reminds me to remind readers of this link which would provide them with a searchable file with all my past tweets (now numbered 46K). Please search for "cyclic" or "feedback".

4.17.2020 5:29am - I am retweeting your query to all our readers, I hope some can point you to a good on-line course based on Primer I lost track.

4.17.2020 3:19am - The more I listen to it, the more I feel I can speak Mandarin. Perhaps because cause and effect are universal or because (no, this is impossible!!) I can really understand Mandarin.

4.16.2020 10:13pm - Amazing photo. I believe it is the first public demonstration complying with COVID-19 "social distance" rules. Israelis protested yesterday government corruption and the erosion of democratic institutions.

4.16.2020 5:46am - Here is how Richard Feynman explains why molecules speed up when gas is compressed. About 5-6 minutes into this video he explains how a gas molecule, fairly stupid one, knows that it must increase it kinetic energy to make up for the work done on the gas.

4.16.2020 5:12am - Replying to @omaclaren) The choice of the word "governing" may be misleading. He meant the law is never violated, which does not imply that it is sufficient for, or that it cognitively "explains" all phenomena.

4.16.2020 4:42am - This is incredible! Straight in the Journal of Statistics Education?! If it's true that "The Book of Why (Pearl and Mackenzie, 2018) increased the motivation to teach concepts that Pearl calls the causal revolution" then writing the book was worth every drop of ink (key stroke.)

4.16.2020 4:18am - Conservation laws do not satisfy our quest for causal explanation. Does a molecule know that it needs to conserve energy? So what makes it speed up in a Venturi pipe? Or when gas is compressed? Feynman always preferred putting himself "in the shoes" of those molecules. Me too.

4.15.2020 7:05pm - (Replying to @weissiam) A theory that explains everything, from quantum mechanics to cosmology, humbles the wise to remain silent ("Yidom" Amos, 5). I must remain silent till I learn how to use this theory to resolve Simpson's paradox, or to spot imperfect experiments. Will be back if wised.

4.15.2020 3:25am - Very pleased to find a powerful completeness result in the theory of transportability: This time the heterogeneous data sets can have missing components and the targets can be group-specific causal effect.

4.15.2020 1:05am - (Replying to @rodakker) Thanks for the link. Again, the abstract raises my objection. The link is MORE than just interaction, its confounding too. Traditional mediation analysis attempts to cast a causal problem in statistical language - an impossibility, as shown here:

4.14.2020 10:40pm - A pay-wall prevents me from reading this whole paper But the abstract raises a huge red flag: "In psychology, the causal process between 2 variables can be studied with statistical mediation analysis." IMHO, there is no such thing as "Stat. Med. Analysis."

4.14.2020 9:04pm - Happy birthday, Tel Aviv I've never seen your beaches so sad Your sons and daughters alone, yet playful and glad, Your true face, unmasked, smiling openly, wide Chosen by tomorrow, a city in white

4.14.2020 4:37am - Malki and Danny, my two fallen souls, Will the darkness of the night remember Two shooting stars went by? Shining their most brilliant lights, their noblest, In their last dance through the sky?

4.14.2020 12:10am - (Replying to @aminsaadou) Of course, provided the editor of Econometrica commits to publishing it. But why do we need a paper? Readers and students will benefit more from seeing economic solutions to the questions I posed to Heckman here: A toy problem is worth 100 papers.

4.13.2020 7:43pm - Econonists are trained to find flaws in models, however plausible, so much so that they have lost the capacity to ask: "Suppose the model was correct, what do I do with it?" As a result, they lost the skill to repair the flaws they labor to discover -- others (epi?) are doing it.

4.13.2020 5:59pm - (Replying to @stephensenn @pash22 and 2 others) Good try, but causal calculus was never proposed to represent "what we mean by causal effects" as did RCTs. "Causal chains" (eg DAGs) and "counterfactuals" were proposed as the building blocks of causal thinking. And they are!

4.13.2020 5:56am - (Replying to @thosjleeper) Oh, thanks, I missed the beginning of the chatter; got there only when the question was put for a vote: who is better? Eager to see the initial trigger.

4.13.2020 3:16am - Funny, we asked this question six years ago: and received 38 juicy comments. The summary then describes sympathetically what it feels like being an economist, denied the guidance of graphs, and pretend you don't need any.

4.13.2020 1:28am - Anyone concerned with inferring causes from effects as well as predicting individual behavior from group data has probably stumbled on the questions raised in this blog entry, and would surely benefit from the answers provided by Carlos Cinelli.

4.12.2020 2:11pm - We are grateful to Congressmen Adam Schiff and Steve Chabot for taking a strong, bipartisan stand on behalf of justice and press freedom . @justicefordanny

4.12.2020 2:32am - "Walking among alligators" means seducing scientific leaders to undergo a paradigm shift under hypnosis, thinking they knew it all along. No one will forgive he/she who finds your wallet and proves that the time you spent under the lamppost - an entire career - was a waste.

4.12.2020 1:55am - (Replying to @phyecon1) I was not aware of this incident. My Pakistani friends always depict Jinnah as a champion of modernity and moderation.

4.11.2020 6:12pm - (Replying to @quantadan) I think I owe this affliction to my Kibbutz, in upper Galilee, for sending me (1954) to study choir conducting in Haifa, see, and to my son Danny, who was the copy-editor of my first book, and kept on complaining about my writing: "It doesn't sing, Dad!"

4.11.2020 5:06pm - (Replying to @vthorrf) Brilliant! I wish he lived in the digital age. He would have become an AI pioneer; he wouldn't have allowed philosophy to linger in philosophy departments.

4.11.2020 2:30pm - (Replying to @vthorrf) Is it Hume, or Locke?

4.11.2020 4:20am - (Replying to @ClaudeAGarcia) At the risk of provoking more anger I would just say that, understandably, the naked is angry at the sunlight, though we are all naked under our pajamas, and we all know it.

4.11.2020 12:31am - (Replying to @yakir_bella) If you can find a translator, I would waive my royalties and donate a free copy to every college library. Why? Because Israel has limited resources, and can't afford to see its science education taken over by the data-fitting style of machine learning - it's addictive & blinding.

4.11.2020 12:16am - If we look carefully at his hand movement, we see that the little one is testing the "mediation formula"; how the spiral wheel turns and stops with and without the linkage in between. He is surely climbing rung-3, toward creative scientific thinking.

4.10.2020 3:56pm - (Replying to @11kilobytes) Two reasons for anger: (1) "We have been doing it already in 19xx" (2) "There is more to it than #Bookofwhy"

4.10.2020 3:16pm - I find this tweet impossible to disagree with, though I am suffering from an illusion that you won't find a scientist who hasn't read #Bookofwhy, barring those who are angered by it.

4.10.2020 2:29am - (Replying to @Michael_D_Moor) Hillarious! No wonder economists refuse to solve "toy problems" -- everyone would be able to see how their wheels turn.

4.9.2020 8:16pm - (Replying to @tylerlu) And the chemist will say: Why study connectionism if we can simulate the chemistry of proteins and have machines solve the human brain mystery.

4.9.2020 7:37pm - I have discovered a spec of truth in Bajwa's fable: The order to keep Shaikh in custody on the grounds of ‘public safety’ was taken INDEED under pressure from external agencies -- I was one of those agencies when I reminded the world who Shaikh is and what he did to my son Danny.

4.9.2020 6:27pm - This is my favorite photo of Danny . The article echos our hopes for justice, moderation and press freedom.

4.9.2020 3:42am - I was about to send warm wishes to all readers celebrating Passover, when this came in. Obama is a much better orator, so I'll ride his words and second his wishes: Happy Passover !!!

4.9.2020 3:31am - Only those who have travelled this road in normal times would appreciate what it takes to bring it to this state of Covid-19 paralysis. It is the road connecting Haifa and Tel-Aviv on Passover eve.

4.8.2020 7:35pm - You go with Sarsour and Omar, Senator, not me. You have brought shame and immeasurable harm to your (former) people. Tonight, as Jews say: "Keivan Sh'Hotsi et Atsmo Min Haklal" ("estranged himself to the community") we will be thinking of you, among other painful disappointments.

4.8.2020 1:48pm - Anyone who wishes to obtain the solution manual for the beautiful examples in should write to and indicate that it will be used for self-study and will not undermine instructors who assign these questions as class homework #Bookofwhy

4.8.2020 2:51am - (Replying to @y2silence) Best compliment an author can dream of. #Bookofwhy

4.8.2020 1:09am - (1/ ) (Replying to @bariweiss) My memorable Seder was on April 23, 1948, 3 weeks before Israel was created and attacked. We, kids, heard all the threats of "monumental genocide" and "rivers of blood" that our neighbors broadcast from Cairo and Beirut, and were curious to see if our "adults" would dare
4.8.2020 1:16am - (2/ ) (Replying to @yudapearl and @bariweiss) celebrate Passover under such conditions. My grandfather just read the Haggada, as usual, got to the story of the five Rabbis who defied Roman's rule and chanted the exodus in Bnai-Brak (my home town), stopped, smiled, looked us in the eye and said: "So, what's new?"

4.7.2020 11:12pm - (Replying to @fuzzydunlop123 @omaclaren and 7 others) Thanks for posting. I was not aware of this insightful paper on the connection between the two models.

4.7.2020 10:45pm - For the life of me, why should hypocrisy and stupidity evoke more twitter chat than being a Zionophobic racist? Who said inconsistency is a greater perversion than immorality.

4.7.2020 7:03pm - Another juicy proverb, just provoked by a discussion with a "missing data" expert: "When it comes to perplexity, nothing tops the science of inertia, except perhaps the inertia of science."

4.7.2020 5:36pm - Any reason you did not like the ending?: "Data science is not a mirror through which data look at themselves from different angles under different makeups"

4.7.2020 4:02am - (Replying to @Razorwindsg) We do not have an organized movement per se. But we have a commitment to education (on-line) and a strong conviction that commonsense will prevail. I would start with spreading the intellectual content of CI in whatever circles you dance; most circles would benefit from it.

4.7.2020 3:22am - A more effective way to write Pakistani Prime Minister Imran Khan in support of justice for Daniel Pearl would be to use these three email addresses: Cc: Thanks, and may justice prevail.

4.7.2020 2:16am - (Replying to @nickchk @Jacobb_Douglas and @Fhanksalot) Discussants on this thread might enjoy a glimpse at the general condition for Z to be a valid IV. The condition goes: 1) There is an unblocked path between Z and X, and 2) Every unblocked path between Z and Y contains an arrow into X. This condition confirms of course the cases

4.6.2020 3:08pm - Responding to readers comments, we have improved our discussion of "Which Patients are in Greater Need?..with reflections on COVID-19" Same with the related post, that argues for "data-interpreting technology"

4.6.2020 3:56am - (Replying to @maliniw90th) Your father is lucky. I wish my daughter would ask me to explain Fourier transform or integration by parts. Fun topics.

4.6.2020 1:55am - Speaking of DL and causal inference (CI) and keeping with our commitment to on-line education, I am retweeting here a video-ed lecture on the subject It's more than a year old, but covers many of the questions raised here, squarely and transparently.

4.5.2020 1:36am - (Replying to @desai_pratik) I haven't kept ups with Soar type of projects. From what I remember this was a "production system" and we, expert systems folks, moved from rules to causal models. But the cog. science people still use Causal Bayesian Networks as a model of cognition and "understanding".

4.5.2020 9:05pm - Agree. Yoshua Benjio is one of the few DL leaders who understands the role of causality in AI, the rest are either avoiding the issue or paying curve-fitting lip-service to it. We, AI and society, will be paying dearly for this neglect, as I note here:

4.5.2020 8:44pm - (Replying to @VincentAB @causalinf and 2 others) I agree of course with your conclusions. What I do not understand is your self-doubt. There is only one definition of IV and one way of verifying it. Only economists are confused today about IV's because, having vowed to shun graphs, they are still debating what "exogeneity" is.

4.5.2020 8:29pm - (Replying to @adepstein1 @RLong_Bailey and @Keir_Starmer) And I hope UK Jews get the courage to talk honestly to @Keir_Starmer and tell him what his litmus test is: How he talks to Labor members about Israel and Zionism and lasting peace in ME, not how he flatters UK Jews.

4.5.2020 3:16am - Readers who asked if they can help, please write to @ImranKhanPTI , the prime minister of Pakistan, and express your hopes to see his government appeal the acquittal ruling, thus reaffirming Pakistan's commitment to universal values of justice and the sanctity of human life.

4.5.2020 2:55am - (Replying to @RonKenett and @mario_angst_sci) I tend to side with @mario_angst_sci on this issue. There is a profound difference between a tool that makes you ask: "How come no one told me?" and one that leaves you blank even after being told. The former lets you do things you always wanted to do and couldn't!

4.4.2020 11:14pm - (Replying to @DocGTLBrown and @mario_angst_sci) Extremely pleased and, if anyone was busy preparing a "guide for students seeking enlightenment" I would recommend adding your course to the White List. One needed ingredient: Add CI to the title, to show that your department is not bowing to pressures or taboos.

4.4.2020 5:54pm - My thoughts on Imben's ideology are here: Indeed, the econometric leadership will have to answer some tough questions to the court of history. Econ. students will be summoned as witnesses, and econ. textbooks as material evidence. Wish I live to see it!

4.4.2020 3:28pm - Likewise, I've been advising schools that advertize for students, postdocs and faculty to state explicitly their interest in CI education and research, else they would not get the students and faculty they are looking for. "Times they are achangin" (Dylan 1964).

4.4.2020 3:14pm - This past week, the Stanford's HAI Institute has organized a virtual conference on "AI and COVID-19," a video of which is now available: I have asked the organizers to share the following note with the participants:

4.4.2020 2:41pm - And how on earth is it that prominent leaders and educators in respectable scientific disciplines do not ask themselves the same question to lift their fields from the margins of the causal revolution? Historians of 21st century science will ask this question.

4.4.2020 7:21am - To all readers who shared hopeful thoughts with us, we are grateful and happy to inform you that the government of Sindh has ordered that Daniel's murder suspects will be kept in detention for another 90 days, pending an appeal. Thanks for being with us.

4.2.2020 8:13pm - A new blog-page, carrying a humble contribution of causal inference to the fight against Covid-19 was just posted here: It asks: "Which Patients are in Greater Need?" and displays the options in vivid colors.

4.2.2020 5:27am - It is a mockery of justice. Anyone with a minimal sense of right and wrong now expects Faiz Shah, prosecutor general of Sindh to do his duty and appeal this reprehensible decision to the Supreme Court of Pakistan.

4.2.2020 3:21am - (Replying to @Abel_TorresM @dileeplearning and 3 others) Agree on first, disagree on second. Structural Causal Models exhibits true understanding of a domain (defined by the variables in the model) in that it answers all causal and counterfactual questions about that domain, as shown in #Bookofwhy chapter 1, "The Mini-Turing Test."

4.1.2020 4:19pm - (Replying to @Dr_Cuspy @tdietterich and @MarcioMinicz) I would hate to use the word "autonomously" in a definition, b/c it is hard to tell if a program does things by choice, agency, or by being programmed to do it. "Predicting consequences" is easier, we can just ask the program to predict, and test the answer against outcomes.

4.1.2020 3:57pm - (Replying to @HundredthIdiot and @deaneckles) Really? I was told many readers filter their tweets by subject matter, and will get only Tweets containing those hash tags. If it ain't true, I'll be more than happy to gain 10 precious characters not using #Bookofwhy (I just did!!).

4.1.2020 3:47pm - (1/ ) In re-reading Holland (1986) I noted that he was the first to define CI as a missing data problem, through his "Fundamental Problem of Causal Inference", which so many PO folks love to quote. Today, in contrast, we are classifying "missing data" as a causal inference problem
4.1.2020 3:47pm - (2/ ) (see for example Wow! How "the times they are a changin" (Bob Dylan, 1964). #Bookofwhy

4.1.2020 3:20pm - For readers who are re-raising @deaneckles question, I am retweeting my reply (below) and add that I also praise DAGs for reminding us that it is not the end of the world, and guiding us to augment our knowledge to get identification: IV, front-door, do-cal. etc. #Bookofwhy

4.1.2020 1:37am - Glad we have a causal model for COVID19 testing, so that we can accumulate facts and judgement into one arena, and talk one language. #Bookofwhy

4.1.2020 1:23am - Thanks for sweet nostalgia. I often play with the idea of writing: "Statistics and CI - 30 yrs later", but not sure JASA would publish it. Holland's paper has shaped the mindset of most living statisticians working on CI, some irreversibly.#Bookofwhy

3.31.2020 8:43pm - Missing data is when Chinese journalists disappear, eg and statisticians later fix with multiple imputation. #Bookofwhy

3.31.2020 7:15pm - For coordinating, comparing and making sense of pandemic data sources, corrupted by local idiosyncrasies we need causal meta-analysis. The theory is available: and is awaiting an elite force of CI PhD's to teach & train practitioners. @EpiEllie #Bookofwhy

3.31.2020 4:56pm - This incredible scene comes straight from the twilight zone, where "The wolf lives with the lamb, the leopard lies down with the goat, the calf and the lion and the yearling together; and little child leads them." (Isaia 11:6). No child in sight, unfortunately.

3.30.20 11:28pm - The warm reception of my conversations with N Jewell tempts me to post another ancient video: It took place 10 years ago, on the foundations of causal reasoning; a topic more alive now than ever, especially while fighting for explanations #Bookofwhy

3.30.20 8:04pm - (Replying to @MEMRIReports and @HananyaNaftali) I know a few US Universities who would vie for his scholarship in the department of Mid East Studies.

3.30.20 3:25pm - The crucial methodological part of the debate rests with fusing coherently all the data coming from so many diverse and noisy sources. We have a calculus+algorithms for doing it, awaiting experts trained in CI and listen; see #Bookofwhy

3.30.20 3:07pm - I'm flattered, but don't stop here. Continue to Primer, and join the common sense revolution - every rebel counts. #Bookofwhy

3.30.20 12:09am - The chloroquine debate should be settled by domain experts. As a methodologist, I can only state that algorithms have been developed that can look at data and estimate the probability that a drug is harmful to a given individual. I assume the experts are reading #Bookofwhy

3.29.20 6:00pm - Written especially for revolutionary-minded people like you. Who said epi is not poetry? #Bookofwhy

3.29.20 3:40am - Another interesting paper: which provides an insightful analysis of cross-world assumptions in mediation analysis. See #Bookofwhy chap. 9, and for nice plots of the mediation formula in linear, logistic, and probit models #Bookofwhy

3.29.20 3:08am - Retweeting an interesting paper by a Microsoft team on one of the most demanding task performed by humans: system debugging It combines causal and counterfactual logic, as opposed standard debugging softwares which are statistically driven. #Bookofway.

3.29.20 2:17am - (Replying to @mgaldino) This is only part of the answer. Let's not forget that formalization serves both: automation and coherence. Applied researchers, following their informal intuition might find themselves producing contradictory recommendations (as in Simpson's paradox) #Bookofwhy

3.28.20 9:17pm - (Replying to @awmercer) Thanks for your honest and informative answers. Recall however that for us, AI-ers, "makes the most sense" is just the beginning of a new challenge: How do we represent (on a machine) the knowledge that tells you that one framework "makes more sense" than another. #Bookofwhy

3.28.20 6:08pm - (1/ ) (Replying to @awmercer) This is the first time I hear of structural equations in the service of the survey field, which makes me super curious: Do these equations enjoy different notation than regressions? Different treatment? Surely, the model-free approach misses something; it misses an opportunity
3.28.20 6:16pm - (2/ ) (Replying to @yudapearl and @awmercer) to benefit from valuable information coming from the model, when such is available to the analyst. What is the simplest survey example in which we can see the tradeoffs between the two approaches? A toy example is worth tons of scholarly words. #Bookofwhy

3.28.20 4:08pm - Another thought evoked by your informative description of survey methods. Since these methods rely on the "no-confounding" assumption, how did survey experts articulate this assumption mathematically? Or they just carried it in the head? What's in the tech literature? #Bookofwhy

3.28.20 3:44pm - (Replying to @shirokuriwaki and @deaneckles) This paper is not very helpful because, following the PO tradition, it assumes away the hard part of the problem. Quoting: "The conventional solution to this problem is to assume ignorable treatment assignment and overlap (Rosenbaum and Rubin 1983)" #Bookofwhy

3.28.20 3:41am - (1/ ) I'm glad, Sean, that our brief exchange has resulted in your great clarification of the issues, from which I have learned a lot. Two thoughts come immediately to mind: (1) It is a blessing that we can enjoy a division of labor between CI and statistics, the former generates
3.28.20 3:41am - (2/ ) causal estimands, the latter estimate them. Note though that the former is not totally oblivious to the type of data available. Different types of data will result in different estimands. eg.,experimental vs. observational, corrupted by missingness or by proxies or by
3.28.20 3:41am - (3/ ) differential selection etc. (2) I don't buy the mystification of "collecting adequate data". I am in the business of automating a scientist, so, if there is human judgement involved in the data collection task, we do not stop here and surrender it to humans. We see it as an
3.28.20 3:41am - (4/ ) invitation to ask: what knowledge allows you to decide that some data are "more adequate" than others. We then model that knowledge and automate the process. I strongly suspect that behind this piece of knowledge you'll find a causal model. I am willing to bet!! #Bookofwhy

3.28.20 3:29pm - Truly appreciate your substantive reply. So, generalization across population is the common theme here. In we solved such problems using graphs. Have we missed something by not listening to the survey literature and to the tools developed there?#Bookofwhy

3.28.20 3:19pm - (Replying to @awmercer and @deaneckles) Great. Now suppose we have good understanding of confounders. What formal language is available to us to articulate our understanding, so that we can process it coherently when the number of confounders exceeds our mental capacity? Is Meng's language sufficient? #Bookofwhy

3.28.20 3:11pm - It is a great paper, and I retweeted it in Dec. 2020 with a warm blessing. However, I am still in a learning mood: Survey methods have been around for 2 centuries. What tools have been developed that we, CI folks, can use to speed up our research agenda? #bookofwhy

3.28.20 2:46pm - (Replying to @agostbiro and @deaneckles) No one doubt it. The entire success of DL rests on this surprise. But we, CI students, have certain problems on our plate for which we cannot find (yet) salvation in understanding correlation alone. And we are begging for help. #Bookofwhy

3.28.20 1:39pm - I am retweeting with the hope that perhaps some other readers beside @deaneckles can help us on this question, which has come up again and again. In a way, ML/DL folks are operating as survey samplers, so they too could help us: What can we learn from your experience? #Bookofwhy

3.28.20 12:45pm - There is another point to this pseudo-critique. Suppose you launch a very smart discovery algorithm and get a causal structure, what then? If you do not teach your students how to leverage it to answer causal questions we are back to where model-free analysts are today.#Bookofwhy

3.28.20 12:29pm - (Replying to @ewerlopes and @Muenchner_Junge) Thanks for posting this course. I have seen many advertised under the catchy title "causal inference" which are doing ONLY estimation, leaving the assumptions to divine intervention. This course has Primer in its reading list - a faithful sign of enlightenment. #Bookofwhy

3.28.20 12:11pm - (Replying to @AngeloDalli) There could be some other type of knowledge guiding statisticians in what they call "design". But the fact that they treat it as divine wisdom prevents us, AI researchers, from automating it. We will, eventually. #Bookofwhy

3.27.20 8:32pm - (1/ ) I beg to differ. For readers interested in my unbiased opinion, this review is ill-informed, ill-motivated and misleading. Bent on showing that "there is more to it than #Bookofwhy", sweating hard and unsuccessfully to find any such "more to it", the reviewers neglect to tell
3.27.20 8:32pm - (2/ ) readers what they would miss not reading the book. The quote: "causal inquiry cannot be reduced to a mathematical exercise nor automatized." is truly indicative of the occult "more to it" agenda of the review, and its failure to appreciate how much has been automated already.
3.27.20 8:32pm - (3/ ) Another unbiased opinion I have regarding "the particular value of randomized experiments". I challenge anyone to show me a clearer/deeper exposition of RCTs and their "particular value" than that given in #Bookofwhy ch. 4, The Skillful Interrogation of Nature, Why RCTs Work?

3.27.20 7:53pm - (Replying to @blake_camp_1 @kchonyc and 2 others) I am late on the philosophy of SSL, so I commend you for giving us a glimpse of what it does. Can you elaborate on how fillings blanks can get you a causal effect without some causal assumptions encoded somewhere in the system? #Bookofwhy

3.27.20 7:42pm - (Replying to @omaclaren and @analisereal) In which case these undergrads should be able to prove that back-door condition is sufficient for identifiability in a simpler, static system, with no derivatives. They should also be able to run DAISY and decide yes/no on any of the models described in #Bookofwhy. Eager!

3.27.20 7:27pm - (Replying to @omaclaren) I would qualify it a bit. ‘Modernity’ is the sum total of things that have been accomplished in the past 3 decades regardless of any particular formalism. This is how #Bookofwhy defines the "causal revolution", and it lists those things explicitly, task after task.

3.27.20 7:16pm - (Replying to @analisereal and @omaclaren) I would add that structural models are causal entities and should be annotated by arrows Y<--f(X,e), not equality signs. Once the distinction is made explicitly, I embrace the paper as "relevant", else, "suspect of generating confusion" as in #Bookofwhy

3.27.20 2:51pm - (1/ ) (Replying to @seanjtaylor) "Useful" is a rich concept. Statistics papers, formal logic, cognitive science, even history books and Greek mythology have been useful to my work. But when it comes to causal questions, a paper that does not articulate the question in mathematically tells me a lot about what
3.27.20 3:19pm - (2/ ) (Replying to @yudapearl and @seanjtaylor) to expect from reading it. It tells me that either (1) the paper deals with stat. estimation after assuming away all causal considerations, or (2) it has nothing to do with causation despite the catchy title, or (3) the author is confused - like many regression analysts -
3.27.20 3:19pm - (3/ ) (Replying to @yudapearl and @seanjtaylor) writing regression and thinking causation. It does not make the paper useless, but it warns you that it ain't going to be easy to excavate for the relevant material if the has any. So, if your time is valuable, you already saved 1000 hours, all for want of notation. #Bookofwhy

3.27.20 1:49pm - (Replying to @blake_camp_1 @tyrell_turing and @ylecun) No need to toy. Just look at the input information. If it contains a causal element (or interventional data) then your toying may succeed, if not, you can't create it from data alone. It's like the conservation of energy (Helmholtz, 1847.) #Bookofwhy

3.27.20 1:42pm - (1/ ) You almost got it correctly. You can read 1000 papers, for 2000 reasons, but if you want to know ahead of time whether it deals with causal questions formally, then it is not a matter of "preference" or "taste", it is a matter of distinguishing causal from statistical
3.27.20 1:42pm - (2/ ) quantities and, then, the use of do(x), Y(x), or DAGs is a good indicator whether a distinction was attempted. I've saved literally thousands of hours by this filter and I would be curious to know of any causally-relevant paper that I've missed. Notation reveals! #Bookofwhy

3.27.20 1:56am - (Replying to @tobytfriend @davidpapineau and 3 others) Good catch, thanks. There are indeed 3 notational indicators to causally-minded articles: do(x), Y_x, or good old fashion DAGs. Thanks. #Bookofwhy

3.27.20 11:44pm - (Replying to @ewerlopes) It is very kind of you to trust me for giving such a course. I'll do some checking around and see if it can be realized given everyone's constraints. #Bookofwhy

3.27.20 1:45am - Responding to interests in the foundations of causal inference, and lacking resources/army/time to marshal a full Zoom course, I am posting a video interview with N Jewell: Introduction to Causality, Part I,, Part II,, #Bookofwhy

3.26.20 7:33pm - (Replying to @btolgao @tvladeck and 3 others) I've been in this playground since 1990, see The PC algorithm (Verma's IC) may discover beautiful structures and we know its limitations. Jona's-type algorithms rest on non-scrutinizable assumptions (eg. additive noise), may turn out promising.#Bookofwhy

3.26.20 4:16pm - (Replying to @btolgao @tvladeck and 3 others) When the DAG is not known, it means the knowledge is not available; we need to acquire that knowledge by experimentation guided by analogies and metaphors, like the Greeks did. Not like the Babylonians. #Bookofwhy

3.26.20 4:09pm - (Replying to @Jabaluck @VincentAB and 2 others) No, I have not seen those "other approaches" arrive at those insights. Let's take the simplest of all: backdoor. Can you illuminate us on how a DAG-less "approach" would arrive there. But, please, let's use input-output, no ignorability acrobatics. #Bookofwhy

3.26.20 3:58pm - (Replying to @deaneckles @tvladeck and 2 others) Yes, we often do not have the knowledge required to answer our questions - I praise DAGs for telling us. We often have 3 equations with 4 unknowns - I praise algebra for telling me that the solution is not unique. Who else would tell us? Statistics? PO? #Bookofwhy

3.26.20 3:50pm - (Replying to @peterlorentzen @ladder2babel and 3 others) ?????????????????

3.26.20 2:35pm - (Replying to @DanielHGill @davidpapineau and 3 others) No, orthogonal is you have left from an orthodentist who failed dentistry school. Same as what's left from big-data lacking a causal model. #Bookofwhy

3.26.20 2:14pm - (Replying to @davidpapineau @totteh and 2 others) The nice thing about modern thinking is that you can tell orthogonality w/o even reading the paper. Just look at the equations, if all they have is probabilistic expressions, no do(x) and no counterfactual subscripts - forget it, its traditional statistics, no causation#Bookofwhy

3.26.20 1:46pm - (Replying to @totteh @davidpapineau and 2 others) Beg to disagree. Meng's paper is orthogonal to the issue of randomization. It is written in the best tradition of statistical orthodoxy - avoid the C-word, and avoid causal thinking. #Bookofwhy

3.26.20 1:21pm - (1/ ) (Replying to @deaneckles @rabois and @Jabaluck) You have assumed correctly. The #Bookofwhy is all about the insufficiency of "big data" to get us over from association (rung 1) to causation (rung 2 and 3), no matter how skillful we are in processing the data. Large n only gets us from samples to distributions. Another push
3.26.20 1:32pm - (2/ ) (Replying to @yudapearl @deaneckles and 2 others) is needed to get us from distributions to causal conclusions. The push may come from interventional studies, or from causal models. Note that RCT's must invoke interventions -- randomization alone is as useless as large n.

3.26.20 1:17am - (Replying to @Jabaluck @VincentAB and 2 others) The best way to address such complaints is to ask: Do you acknowledge the existence of some inference tasks which DAGs perform well, and which the alternatives find intractable, that is, close to impossible? #Bookofwhy

3.25.20 9:22pm - (Replying to @peterlorentzen @rabois and 2 others) Sorry, I haven't been around to catch what the dispute is about, or what @rabois said that triggered @jabaluck anger. I am not sure #Bookofwhy can help but I have found it a good start for any discussion about cause and effect.

3.25.20 1:27pm - I'll be back!!! I promise!!!

3.25.20 1:03pm - This is the first Certificat I have seen in "Causal Data Science"! Congratulations!!! I am anxious to see the first PhD Diploma in Causal Data Science, perhaps 4 years from now. #Bookofwhy

3.25.20 6:34am - (1/ ) I used to think that a journal editor can help refresh a field by inviting "survey articles" from outsiders. I no longer think so, b/c the questions that controvertial papers evoke need be answered by direct interaction with the author. My recommendation: Publish a dialogue
3.25.20 6:34am - (2/2) (a technical one) between the outsider and the top leader in the field that seeks fresh air. Anecdotally, I have just finished such a dialogue (for a prominent journal) and the editor who appointed the interviewer rejected the outcome. Lesson: that's what's needed #Bookofwhy.

3.25.20 5:34am - (1/ ) Feldman's observations are insightful and thought provoking. I think we, scientists, should learn from politicians (don't choke!). They knew that power corrupts and instituted 3 branches of Gvt to check on each other. We do not have a "checking" branch to tell any of our
3.25.20 5:34am - (2/ ) self-sustaining "fields": Hey we are X years behind the time! We are waiting for outsiders to do it, which is too slow a process. Here is a counterfactual thought: Vice Presidents in charge of "Fresh Air", that is, innovation, openness, cross-field communication etc.#Bookofwhy

3.24.20 10:44pm - (1/ ) To assess where quantitative psychologists stand on the road up Mt. Causation, I examined the citations in the "taboo paper" and noticed a dearth of causally sound papers in psych.. [And I thought economists have the hardest time with modernity.] While
3.24.20 10:44pm - (2/ ) the survey paper by Foster is pretty good (for 2010), not much else from a field hopelessly sunk in SEM culture. I knew that Psychometrika and JSEM can't shake stagnation, but I thought the rest of the field is more progressive. Cognitive Psychology,
3.24.20 10:44pm - (3/ ) for example, is a model of progressive thinking. It would be helpful to hear from readers active in this field. Am I off in this gloomy assessment of quantitative psychology? #Bookofwhy

3.24.20 6:36pm - I have an even shorter link and note that the authors have not yet used the heavy guns against the supremacy of RCT's: Selection bias (as discussed here: ) #Bookofwhy

3.24.20 6:04pm - I hope this link works:

3.24.20 5:56pm - The Taboo Against Explicit Causal Inference in Nonexperimental Psychology is undergoing a harsh examination here file:///C:/Users/Judea/Downloads/Grosz%20et%20al.%20(2020,%20preprint)%20The%20Taboo%20Against%20Explicit%20Causal%20Inference%20in%20Nonexperimental%20Psychology.pdf

3.23.20 10:36pm - (Replying to @AdanZBecerra1 @mayfer and @_MiguelHernan) Of course. Likewise, doing real world dpi is "doubly-super-hared," if not impossible, without SCM. I have not heard about "other causal frameworks" that distinguish "cause of death" from "observed and death" and estimate the former, not the latter. #Bookofwhy

3.23.20 10:24pm - (Replying to @AdanZBecerra1 @mayfer and @_MiguelHernan) In the case of CV, it is reasonable to assume: "CV caused death" = "CV verified & death." Counterfactual logic is needed when we doubt this identity. Note that "verified" is still different from "tested positive" which is behind the statistics we are getting.#Bookofwhy

3.23.20 4:27pm - (Replying to @djnavarro) My personal apology to you, Danielle. Sorry.

3.23.20 3:02pm - (Replying to @AshbyHarper) Had I been in agreement with the ideology of the writer, I would not have apologized. But I object to everything he stands for, so I apologize for having created an unintended false impression. Sorry!

3.23.20 5:26am - (Replying to @Satoohn and @ChrisAdamsEcon) The weather predictor is observing two different needles, goes through some mental process and makes a prediction. Someone asks "what is the effect of Needle-1 on the prediction, given data from the past 100 days. This is where collinearity hurts. #Bookofwhy

3.23.20 3:38am - (Replying to @athos_damiani and @_rchaves_) I give publicity to many articles that invite humor, criticism and, God forbid, even ridicule.

3.23.20 3:22am - (Replying to @athos_damiani and @_rchaves_) Sorry, the title was 2+2=4. The paper may have jumped to 2+2=5, but the title is innocent. BTW, Who is God?

3.23.20 2:31am - (Replying to @DVBartram) Reconsidered. Retracting. Havn't foreseen the side effects or the implications. Asking for forgiveness. Which, btw, is a counterfactual notion demanding formal explication. But I'd settle for ordinary, pre-scientific forgiveness.

3.23.20 1:17am - (Replying to @_rchaves_) What's wtf? What's transphobic? What have I done wrong this time? Can't we enjoy a title saying 2+2=4 w/o offending a mathematician (or an economist)?

3.22.20 12:48pm - I heed to your proposal and offer our Crash Course to all data-intensive professions: I mentioned economists only because they coined the name "bad control" and have been most obstinate in refusing the science of its remedies. #Bookofwhy

3.22.20 12:25am - I see that you singled out "psychologists" as potential beneficiaries of our "Crash Course on Good and Bad Controls". What circles in psychology deserve this honor? #Bookofwhy

3.21.20 2:45pm - (Replying to @hangingnoodles and @aminsaadou) Causal modeling tools beat and escape the Lars quote better than unaided "thinking", which some (eg., Keynes) see as a superior tool to rescue science and policy makers from the complexities of social systems. #Bookofwhy

3.21.20 2:31pm - (Replying to @MehboobFarah) The distinction between "good and bad controls" is not used for "constructing models" but for "utilizing models" once constructed and deemed plausible. #Bookofwhy

3.21.20 12:55pm - (Replying to @aminsaadou and @hangingnoodles) The causation tools available nowadays are EASY for everyone (with the exception of some outdated old-guards), see, and they also make clear when old tools are adequate. #Bookofwhy

3.21.20 12:36pm - For the many students of economics who express frustration with the way a problem known as “bad control” is evaded, if not mishandled in econometrics, our "Crash Course" in now accessible as Technical Report: Enjoy and teach your professor. #Bookofwhy

3.21.20 5:10am - Proud to congratulate my colleague-in-law, professor Furstenberg, on winning the Abel Prize in Mathematics: Wishing El-Al Airline a quick recovery, so he can fly to Norway to receive the Prize in person. #Bookofwhy

3.21.20 1:28am - "Causal Transfer" and "Imitation Learning" is what called my attention to this new paper:, plus the fact that it approaches the problem scientifically, as opposed to data-fittingly. A prerequisite to data-science. #Bookofwhy

3.21.20 12:26am - To clarify, my comment below relates to @LarsPSyll statement "Econometric modelling should never be a substitute for thinking." which some take as a license to substitute modelling with "thinking". Modeling is a necessary amplification of the solid part of our thinking #Bookofwhy

3.20.20 7:31am - (Replying to @hangingnoodles) I am willing to accept the unmanageable nature of social systems. What I find hard to accept is that an unaided and fallible machine called "commonsense economics" is better equipped to manage those unmanageable systems.

3.20.20 1:38am - (Replying to @jules__shen) Impossible! How can a skill that governs so much of our every day thinking and emotions be "very hard"? Impossible! Indeed, the 3-step process of computing counterfactuals is easy. #Bookofwhy

3.19.20 7:15pm - (Replying to @tdietterich and @zacharylipton) Not familiar with the specifics, but would like to mention that society paid dearly for brilliant perpetual motion machines, proposed, pursued, funded, argued in courts etc. many decades after Helmholtz (1847) paper on conservation of energy. Seek balance & principles! #Bookofwhy

3.18.20 8:21am - (Replying to @100trillionUSD @pierre_rochard and 5 others) No. I said: IF cointegration analysis aimes at distinguishing spurious and non-spurious relationships, then it must be based on causal assumptions and, thus far, I have not seen the literature on cointegration explicate such assumptions, nor use causal vocabulary.#Bookofwhy

3.17.20 7:34pm - (Replying to @pierre_rochard @ercwl and 2 others) But what IS this remarkable relationship? Can you show it in just 3-4 variables and 3 time slots? I would like to add it to my list of "remarkable" relationships of correlated variables. If it is truly remarkable, it should be demonstrable in 3-4 variables. Curious #Bookofwhy

3.17.20 7:26pm - (Replying to @JohnMarkoff2 @ercwl and @100trillionUSD) If cointegration is a statistical property of time series then @btconometrics is right, it does not imply causation. See Whether it "smells" causation depends on the assumptions made by the smeller; they may be plausible, but we need to see them #Bookofwhy

3.17.20 7:14pm - (Replying to @pierre_rochard @dopamine_uptake and 5 others) I am yet to find a signal that a certain "correlation is not spurious" which is not based on causal assumptions. So, has any of its advocates been able to explicate those causal assumptions? I know that Granger was not able, see Who was? #Bookofwhy

3.17.20 4:03pm - (1/ ) Recent exchange with theory-averse readers compels me to retweet a slogan: "I would rather follow Eratosthenes and measure the radius of a legendary turtle than fit curves in a Babylonian observatory." It refers to Greek science in Toulmin's book "Forecast and Understanding"
3.17.20 4:03pm - (2/ ) (see, last page) and his observation that it takes a theory to design a meaningful experiment. Moreover, we are fortunate to live in an era where the "understanding" part of Toulmin's title can be emulated on a computer and tested against its rivals.
3.17.20 4:03pm - (3/ ) I am speaking of course about SCM which, to the best of knowledge, is the first system to deserve the title "understanding", due to its ability to pass the mini-Turing test on all three levels of human understanding (#Bookofwhy Ch. 1): Seeing, Acting and Imagining.
3.17.20 4:03pm - (4/ ) Its opponents do not usually offer computational alternatives, preferring instead to speak in generalities about the superiority of data over "false narratives." Still, by the time they are ready, we should be able to run the first empirical test of the "Scientific Method."

3.17.20 3:16pm - (Replying to @Saravanan_CU) The main disparities that I would include are disparities among gv't policies. Still, transportability should be used whenever we generalize across populations.Unfortunately the theory is only one decade old and analysts in power take 3 decades to shift paradigms. #Bookofwhy

3.17.20 9:40am - (Replying to @sean_a_mcclure and @HarryDCrane) The opaque culture you prefer reminds me of Babylonian astronomy in Toulmin's "Forecast and Understanding" (see, last page) It may get us to Babylon, but not to Athens; I prefer the footsteps of Eratosthenes and measure the turtle's radius #Bookofwhy

3.17.20 12:57am - (Replying to @raymondshpeley) I like it too. And am going to keep it. I truly feel for those who judge #Bookofwhy by how it confirms what they knew already, instead of rejoicing the tools it provides for doing today what they wanted and could not do yesterday. Boy! how much fun they are missing!

3.16.20 9:16pm - On the contrary! I tell students: "formality amplifies common sense." Look at common sense logic before and after Boole. Look at reasoning with uncertainty before and after Bernoulli + LaPlace. #Bookofwhy

3.16.20 8:49pm - Commiserating with BDS cronies - blazing beacons of deceit. They smile while lying and smile when caught. Do they really know the difference?

3.16.20 6:29pm - (Replying to @DKedmey and @HarryDCrane) One important caveat. If you know nothing about your past behavior except that you are like anyone else in a given subpopulation S, then the personal question: "Would I recover if I take this drug" can be answered from population-level data: find P(recover|do(drug),S). #Bookofwhy

3.16.20 6:00pm - (1/ ) Back to students of CI and other sapiens fascinated by logic, counterfactuals, and everyday language. Here is a commonsensical, everyday notion called "NEED", as in "lets first test people in greatest NEED". We take it for granted that, if the need arises, analysts would
3.16.20 6:00pm - (2/ ) be able to define what "being in NEED" is, and how to quantify the degree of NEED. It now downed on me that this simple notion is intrinsically counterfactual, hence only those familiar with PNS would be able to agree on how to quantify it, given data and commonsense knowledge
3.16.20 6:00pm - (3/ ) I have nothing exciting to say to those who get angry at #Bookofwhy. Knowing everything they are not fascinated by anything, and miss so much of the fun of discovery.

3.16.20 5:09pm - (Replying to @MalikMashail and @McDonalds) I tried not to get too personal. Oh, almost forgot my logo: #Bookofwhy, sorry.

3.16.20 4:54pm - (Replying to @DKedmey) Well put! I could not express it more clearly. In addition, this difference translates into (i) individual level vs. (ii) population level inferences. To guess whether "Joe would have reacted differently" I need to undo what I know about Joe. #Bookofwhy [no off @HarryDCrane etal]

3.16.20 4:32pm - (Replying to @HarryDCrane) FYI, and for all those offended by my tweet: please note: The tweet was addressed to "students of CI" who have been estimating PNS for e-commerce and drug approval. Identifying "those in greatest need" is another application we should keep in mind.

3.16.20 4:13pm - I end my tweets with #Bookofwhy because, so I am told, many readers filter their tweets by this tag, and I do not want them to be left out of the discussion. It is not meant to suggest that the book "solves every problem"; an epsilon of commonsense would rule out such suggestions

3.16.20 3:40pm - (Replying to @DavidPoe223) Right on the nail. It is now 30 years since I started asking regression analysts: "What puzzle a regression is supposed to solve?" Some say:"Given information X, we want to predict Y". Fine. But some say "We want to understand, to decompose, to make sense". Dont get it.#Bookofwhy

3.16.20 3:25pm - To students of CI, Coronavirus is a reminder that "people in greatest NEED" is a counterfactual notion; the degree of "need" is the probability of "helped if treated & doomed if not", dubbed PNS in Causality p.286, and tightly bounded in #Bookofwhy

3.16.20 2:57pm - (Replying to @DavidPoe223) In Simpson's paradox the puzzle is simple: We cant conceive of a drug that is good for men, good for women, and bad for a person. What is the puzzle that "multilevel and fix effects" models are trying to resolve? #Bookofwhy

3.16.20 7:02am - (Replying to @DavidPoe223) I am still curious: What is being sought? What is to be estimated? What is it that does not make sense before you start writing down equations? #Bookofwhy

3.16.20 5:59am - More from Michael Levitt on why the spread of Coronavirus is slowing #Bookofwhy

3.16.20 4:03am - (Replying to @JRubinBlogger) And we can safely assume she won't be Linda Sarsour. If I were Biden, I would make this point loud and punchy.

3.16.20 3:54am - (Replying to @rodakker @tudorcodes and @imleslahdin) I hope what I mean by "causal" coincides with yours, causality is a human language. As to Jaynes, he was sold on probabilities, so I do not believe he can define "causal arrows" better than other Bayesians, who are semi-alien to causation: #Bookofwhy

3.16.20 12:10am - (Replying to @rodakker @tudorcodes and @imleslahdin) This happened to be a Causal Network, even though we are only interested in probabilities. For a non-causal network of the same physics, consider AND--->D1---->D2 plus AND--->D2. It contains non-causal arrows which are necessary to get the probabilities right. #Bookofwhy

3.15.20 3:14pm - (Replying to @tudorcodes and @imleslahdin) All data are generated by some causal processes, hence causal diagrams can always be used to capture the statistical features of the data, provided we allow for hidden variables (& selection B). To save the labor of dealing with hiddens, you may use undirected graphs. #Bookofwhy

3.14.20 10:43pm - Of all the statistical analyses of the C-Vir outbreak, the most optimistic evaluation was given by (Nobel) Michael Levitt, in an interview to China Daily: His technical paper, tables, curves and figures can be viewed here #Bookofwhy

3.14.20 6:04pm - (Replying to @svonava @Joshyblogz1 and @lexfridman) This wikipedia article on counterfactual thinking was written before the algorithmization of counterfactuals. Someone ought to tell the authors to reexamine those theories because much of what they describe can now be done by machines, eg., or #Bookofwhy

3.14.20 1:58pm - Why soap? And why it matters? Statistical analysis of the Coronavirus outbreak is illuminating, but the most useful article I've read in the past two weeks is this: And it has a touch of #Bookofwhy in it.

3.14.20 11:37am - Causal diagrams capture a small but important chunk of human understanding. To their credit, they give us a friendly and formal object from which we can derive what Keynes probably meant by "the broad reasons why past experience was what it was". #Bookofwhy

3.14.20 10:48am - (Replying to @mgaldino and @mendel_random) Barring a few exceptions, of course.

3.13.20 12:59pm - (Replying to @ganesh__s and @eliasbareinboim) Sure! Even static physical equations (eg Hooke's Law) convey counterfactual information. But you have to be explicit how to simulate "undoing", e.g., "what if event E did not occur?" #Bookofwhy

3.13.20 11:39pm - (1/ ) On curve fitting vs. science, @eliasbareinboim showed me a beautiful quote from John Locke (1960, Essay VI): "By the color, figure, taste, and smell and other sensible qualities, ...we cannot tell what effects they will produce; nor when we see those effects can we so much as
3.13.20 11:39pm - (2/ ) (Replying to @yudapearl) guess, much less know, their manner of production." Locke understood in 1690 that the key to human understanding, knowing "their manner of production," cannot be inferred from data. #Bookofwhy

3.12.20 1:47pm - The March issue of the Philosophical Transaction of the Royal Society celebrates 50 years to Price Equation, which describes how genes change over time. I was pleased to find two papers on causal interpretations of the equation.

3.12.20 1:28pm - (Replying to @ewerlopes) Yes. I purchased a new kindle version and, aside from their usual sloppiness in handling equations, they have implemented almost all the errata. #Bookofwhy

3.12.20 5:36am - (Replying to @Stebbing_Heuer @regev_nir and @BernieSanders) My general philosophy is that every experience, however subjective, is a piece of information that should be integrated with other sources to forge theories. I don't know anyone among my colleagues who is not disgusted at the way @BernieSanders is selling his soul to angry voters

3.12.20 5:21am - (Replying to @NNrehman) Sure. But please consult the Errata list here:

3.12.20 4:46am - The day we were all been waiting for has arrived!!! Our publisher informs us that the new, revised and corrected printing of Primer is now out at all US outlets. To make sure your copy is "revised", check that the "about the authors" page cites #Bookofwhy

3.12.20 4:39am - (Replying to @Stebbing_Heuer @regev_nir and @BernieSanders) My personal, first-hand acquaintance with my compatriots, the American people, and their deeply ingrained respect for leaders who do not sell their heritage for a few angry votes.

3.12.20 4:31am - (Replying to @jasndoc and @Farzad_MD) A highly motivated reason to examine how modern causal analysis can be harnessed to assist a task which appears purely associational but involves fusing together diverse sources of information from diverse populations. #Bookofwhy

3.12.20 1:34am - (1/ ) (Replying to @regev_nir and @BernieSanders) "Think not with thyself that thou shalt escape in the king's house, more than all the Jews. For if thou keep thy silence at this time, relief will arise to the Jews from another place, but thou and thy father's house will perish". Mordechai would'nt use such harsh words, unless
3.12.20 1:53am - (2/ ) (Replying to @yudapearl @regev_nir and @BernieSanders) he sensed reluctance on Esther's part to plead with the King despite timing considerations ("I wan't called for 30 days"). I think Michigan voters felt the same about Bernie's selling Israel, the culmination of his people's history, for a few angry votes, promised by Rashida.

3.10.20 11:26pm - (Replying to @AnneHerzberg14 @richard_landes and @stan_state) B. Franklin's words keep ringing in my ears: "He who sleeps with dogs will rise up with fleas." And the Mishna says: "Leaders, watch your words! Lest your followers misinterpret them!" Poor Bernie, will he ever understand the dangers of bad company?

3.10.20 10:57pm - (Replying to @AnneHerzberg14 @richard_landes and @stan_state) It is not funding that the faculty of @stan_state should worry about, but their reputation. As much as I am telling myself "dont blame the grass for one bad weed", it does reflect on their mentality, what they tolerate as "scholarship" and what they feed students. STANISLAUS

3.10.20 10:32pm - (Replying to @regev_nir and @BernieSanders) Esther did need this. She was about to brush it off with: "The timing is not right, etc etc." Mordechai reminded her: "Perhaps history elevated you to this position for a purpose". Bernie had no one to remind him. I tried, but I doubt he can hear the bells of history. Poor Bernie

3.9.20 1:21am - Poor Bernie, he thought by abandoning Israel he would win the hearts of his Zionophobic followers. Alas, they now reject him as "rich Jew": He forgot what Mordechai told Queen Esther: "You wont save your skin by abandoning your people". @BernieSanders

3.9.20 10:11pm - (1/ ) Columbia President Lee Bollinger has issued a statement condemning antisemitism on campus. Kudos! But he evidently has not read my advice To stop BDS hostilities it ain't enough to decry anti-semitism; you've got to expose what
3.9.20 10:11pm - (2/ ) BDS stands for: the elimination of Israel, and the expulsion of Israel-inspired students (and faculty) from campus life. And you've got to express how you, personally, feel about what they stand for, as President Martha Pollack of Cornell did: @Columbia

3.9.20 2:03pm - (Replying to @charleendadams) Never abandon babies, with or w/0 waters. It's good for science to bemoan what we need: triangulation, external validity, selection bias, missing data, etc etc. Bemoaning loud enough and long enough may get someone attention, lest it becomes a habit that avert us from doing it.

3.9.20 5:28am - (Replying to @kurtosis0 and @BernieSanders) Repeating: "Abandoning Israel" means supporting, endorsing, and giving platform to those who seek Israel's destruction. You can abandon Israel being an American or a Hottentot, all you need is moral frailty and Zionophobic voter base

3.9.20 5:10am - (Replying to @jmourabarbosa and @BernieSanders) Abandoning Israel means supporting, endorsing and giving platform to those who seek Israel's destruction. Simple! Simpler even than causality.

3.9.20 3:34am - The word "triangulation" has been used by many to express dissatisfaction with established methodologies and romantic aspirations for something new. This paper however, recognizes formal ways of triangulating information: #Bookofwhy

3.9.20 3:06am - (Replying to @Stebbing_Heuer @kurtosis0 and @BernieSanders) You got it perfect. Being Jewish is not genetic, nor religious, nor being a victim of the holocaust. It is identifying yourself with the history and destiny of a collective calling itself Jewish. Anyone can acquire it. E.g., Large segment of former Soviet citizens now in Israel.

3.9.20 1:32am - (Replying to @kurtosis0 and @BernieSanders) And Queen Esther was Persian. Truman and Kennedy were American too, but would not admire Zionophobic racists like Linda Sarsour and Ilhan Omar.

3.8.20 6:27pm - (Replying to @GeoffWinestock) I did not realize you have a question. I agree that the British could have done many things (eg. allow movement on the Danube to the Black Sea) but this does not absolve the Mufti from the crime of strangling European Jewry by the anti-British anti-Jewish riots of 1936-39,

3.8.20 4:59pm - Today is international women's day, coinciding with the holiday of Purim, the day that Queen Esther said: I must do it, whatever comes. I salute the Israeli female soldiers who served with me in the army (1953) and kept my hopes high until this very day:

3.8.20 4:33pm - It is the first time that I hear this music and I see the charming smile of Al Husseini, the man who prevented my grandfather from escaping Hitler, the father of BDS (1936 boycott of Jewish products) and the creator of Zionophobia. May history remember him for his deeds.

3.7.20 2:01pm - (Replying to @RokoMijicUK and @BonbonFork) No qualms about "big jumps", but when I characterize a tool, I say what the tool delivers, not how it operates. #Bookofwhy

3.7.20 9:03am - (Replying to @ChrisAdamsEcon) McFadden's random utility model is one model in which we can compute certain POs. RCT is another. Speaking in general, however, given a structural model M, which PO's can be computed, and how? #Bookofwhy

3.7.20 8:04am - (Replying to @ChrisAdamsEcon) That is why I prefer "first person" comments, rather than hearsay. Can you state: "To me, Phil's views resolved several questions that I had: (1), (2)...." Or: "If you havn't read Phil's slides you might be tempted to believe (1),(2).. Not so!" #Bookofwhy

3.7.20 7:44am - (Replying to @FriendlySceptic) Speaking of intellectual honesty, the word "equating" is out of place. I said: When you endorse Zionophobic racists like Linda Sarsour and Ilhan Omar, you should expect other racists to smell blood and yell: "Free season!" Havn't heard Sanders' defenders defend these endorsements

3.6.20 9:41am - (Replying to @TheMongoose4) Never mind what right-wing Jews are saying. Listen to what left-wing Jews are saying: When you endorse Zionophobic racists like Linda Sarsour and Ilhan Omar, you should expect other racists to smell blood and yell: " free season!". Signed: a disappointed Bernie sympathizer.

3.6.20 9:12am - (Replying to @roieki) Chas V'Chalila!! Never "BaMachane", and never "Nazism". But when you endorse Zionophobic racists like Linda Sarsour and Ilhan Omar, you should expect other racists to smell blood and yell: " free season!"

3.6.20 8:38am - Schoking? Yes, Surprised? No. I believe it was Benjamin Franklin who said: "He that lieth down with dogs shall rise up with fleas". When you open your house to haters you can't screen them for table manners.

3.6.20 6:18am - (Replying to @ChrisAdamsEcon) We discussed Haile's slides here, on Twitter, and I did not get the impression that economists agree with him. Correct me if I am wrong. Also, help us find where economists define PO's from structural equations. The closer I got was Heckman: #Bookofwhy

3.5.20 11:51am - Econometrics has two camps: Structural and quasi-experimental. The former say: PO comes from structure, but don't tell us how. The latter resist structure, and pretend PO resides already in one's mind. Both camps needs a shake, I tried here, & #Bookofwhy

3.5.20 11:44pm - (Replying to @JDHaltigan) Thanks for educating me on this new measure. I'm still trying to figure out when I would feel that strange need to hold Z constant for just X. BTW, the expression "hold Z constant" is misleading, it should be "selecting only cases for which Z=z". #Bookofwhy

3.5.20 11:27am - (Replying to @jiaming_mao and @causalinf) Symmetric? No. Each can be separately derived from a completely specified SCM (a set of structural equations). It is so nicely described in Primer, and it is available for free:, Why mystify? #Bookofwhy

3.4.20 11:34am - (Replying to @MrTeshnizi) Good question. I think we are born with a template for gathering and using causal knowledge, which DL agents do not. If they do, they should be able to pass the mini-Turing test of #Bookofwhy chapter 1.

3.4.20 4:49am - I'm absolutely delighted to be elected Honorary Fellow of the Royal Statistical Society, an institute that has been rather lukewarm towards causality in the past. I see it as an invitation to all of us, causal enthusiasts, to meet the challenge of becoming mainstream. #Bookofwhy

3.4.20 2:36am - Better in the sense of having lower asymptotic variance.

3.4.20 12:30am - Explanations of what? Examples of what? Sorry, I am a foreigner, please speak slowly to me.

3.4.20 12:04am - A reader just resurrected this yr-old tweet, and reminded me that my question is still unanswered: Which estimator is better? The ML estimator of the product or the product of the ML estimators?

3.3.20 11:53pm - My heart goes to MN voters. Who will protect their children from the Zionophobic hatred spewed by @BettyMcCollum04 ? Who will protect their reputation from the thought they might share her mentality? My heart goes to Minnesota voters.

3.3.20 9:29am - (Replying to @mariabloec @jonasobleser and 5 others) I have a conciliatory ending for this debate: The notion of "Well-definedness" is ten times more ambiguous than the causes it tries to disambiguate. #Bookofwhy

3.3.20 8:46am - (Replying to @mariabloec @jonasobleser and 5 others) I cannot think of a well-defined interventions for earthquakes and, still, their causal effects on property damage is not vague. I discuss it here and here "Well definedness" suppresses causal thoughts. #Bookofwhy

3.3.20 8:36am - I love the ending of this old post: "Once tried - always used". Thanks for re-posting. It applies to so many tools of causal inference. #Bookofwhy

3.3.20 6:22am - An arrow between A-B would be a different story, which we could try later as the hard part of the test. #Bookofwhy

3.3.20 12:22am - It is an honor to be in the company of these stellar authors, Thanks

3.2.20 11:23pm - My advice to readers who are asking me to help them recruit students and faculty in ML: State explicitly that you seek candidates in the CI area. This would signal to candidates that your department is serious about changing the causality-free tradition of ML education.#Bookofwhy

3.2.20 12:37pm - (Replying to @BlueManifold and @dmonett) Indeed, now that I re-read it, they do emphasize the importance of causal and counterfactual reasoning. I did not pay attention to it in 1969, being deep into heuristic search. It is only when I saw causality as the force behind Bayes networks that I asked: what is it? #Bookofwhy

3.2.20 7:46am - (Replying to @mktopuz) It's easy to check using the mini-Turing test on the firing squad (#Bookofwhy chapter 1). "Would the prisoner still be dead had soldier-1 refrained from shooting?"

3.2.20 7:23am - The next step is to ask: what kept the 1912 guy from recognizing the importance and/or generality of his/her observation, and turn it into a "theory of 1912"? Language has lots to do with it. #Bookofwhy

3.2.20 7:14am - (Replying to @BlueManifold and @dmonett) Computer science is the only discipline I know, in which people understand that languages are here to serve a purpose, to be extended and replaced, not to be religiously wedded to, till death do us part. #Bookofwhy

3.1.20 11:41pm - For curious readers, the intrasitivity of correlations, treated as an "issue" in @nntaleb paper "fooled by correlation," is an invariant feature of colliders. In X-->Z<--Y we have Cov(XZ)>0, Cov(ZY)>0 yet Cov(XY)=0 regardless of what numbers we spray on the arrows. However,

3.1.20 11:41pm - DAGs enjoy a weaker form of transitivity: If X and Y are each dependent on Z, then they must be dependent on each other, either marginally or conditional on Z. See "Probabilistic Reasoning.." (1988, p.129). It is not an immutable property of probabilistic dependence #Bookofwhy

3.1.20 3:29pm - "English" is great to describe a phenomenon and show how it is reflected in data. We need a mathematical language to fully understand a phenomena, so as to meet the criteria proposed here vis a vis Simpson's paradox. (English description=1899) #Bookofwhy

3.1.20 3:13pm - I like the way you put it: "formalize the intuition that you take for granted". But I dont foresee "a big blunder" for those who resign to taking certain intuitions for granted. The only problem might be repeating analyses on intuitions that come from the same source. #Bookofwhy

3.1.20 2:36pm - This is precisely the problem of answering counterfactual queries from group's Data, see for a gentle introduction and a tool kit. The "threats" paper you cite should help us motivate modern "remedies", thanks. #Bookofwhy

3.1.20 2:12pm - (Replying to @nntaleb) Your paper shows how certain expectations (eg additivity, transitivity) are often violated. I ask: whence those expectations come from? The logic of causal intuition provides insights into this question. BTW, MI is also intransitive; is it bad? good? surprising? Why? #Bookofwhy.

3.1.20 8:07am - (Replying to @AngeloDalli @AdanZBecerra1 and 2 others) It depends on the granularity of our variables. If I measure the movement of my coin-flipping finger in mm, the outcome is causally independent of that movement, but if I measure it in Angstroms, it's a different story. #Bookofwhy

3.1.20 6:16am - It is fascinating to watch how traditional probabilists handle puzzles that emerge from causal reasoning (in this case the collider structure) lacking a language to capture causal relations. How do we tell them about the causal language w/o offending them? @nntaleb #Bookofwhy

2.29.20 2:03pm - Indeed our conception of cause and effect has only ONE point of contact with do-calculus - the starting point. Do-calculus insists on starting with human conception of reality, faithful to the way it is stored in the mind. The rest is mechanical, like algebra. #Bookofwhy

2.29.20 1:30pm - (Replying to @tmorris_mrc and @y2silence) We can't allow any reader to be left with "low points". The quote is about human conception of cause effect relationships. While external tools for answering certain causal questions may be absolutely necessary, this does not make them part of the conception. In the case of RCT,
2.29.20 1:41pm - we can read text after text of scientific explanations of biological phenomena - what leads to what, what prevents what etc etc - with not a single mention of RCT. See eg the Appendix to RCT is a tool for interrogating, not explaining nature. #Bookofwhy

2.29.20 1:16pm - Why does it matter if Zionophobia IS antisemitism or just another form of racism. The question is how genocidal it is in its aim and how deceptive it is in its advocacy.

2.29.20 5:41am - Our conception of cause and effect is almost always non-deterministic (eg,. careless driving causes accidents), so what? It still can be reasoned about ,formally and informally, without RCT's. #Bookofwhy

2.29.20 5:02am - Got a neat brain teaser from a reader of #Bookofwhy regarding Game 2, page 159: Controlling for D would constrain the mediator E, so why is it legitimate? It's good for a homework, for it ain't trivial. The answer is in the Appendix of Enjoy.

2.29.20 4:18am - Thanks for the compliment. All we need to do now is to convince the establishment.

2.29.20 4:09am - Agree. The computational approaches rangesfrom "simple statistical models like linear regression, to... [more statistical models] like convolutional neural networks". Hoping the 2nd edition snaps out of Rung-1 to methods that interpret data. @DShaywitz #Bookofwhy

2.29.20 3:42am - Zionophobe is one who does not believe in the right of the Jewish people to a homeland, in the context of "2 states for 2 peoples, equally legitimate and equally indigenous". Examples are: Palestinian leadership, educators, clerics, BDS activists and their European supporters.

2.29.20 3:32am - I spread messages of peace and equity, summarized as: "2 states for 2 peoples, equally legitimate and EQUALLY INDIGENOUS." I am risking my academic status to spread this message, because others don't, and many of my like-minded colleagues are prevented from openly doing so.

2.28.20 7:18pm - (Replying to @jos_b_mahoney) No. Apology. I do not know this work.

2.28.20 6:28am - I have always thought that "object-property" relationships (ge, birds fly) require new semantics, not captured by causation. This paper proves me wrong: Time to preserve old beliefs and time to revise them (said King Solomn). #Bookofwhy

2.28.20 5:42am - I would not call a statistical equality "constraint". It is more a "happening", or "coincidence" in the observed data than a constraint imposed by the model's assumptions. #Bookofwhy.

2.28.20 5:37am - Of course! If the outcome of a random coin makes me win a dollar, then it is certainly causal. You are probably asking whether the dependence of the coin outcome on the way I flip it is causal. Classical mechanics says: Yes, despite the absence of prediction/control #Bookofwhy

2.28.20 3:00am - Sure! I would lump all such factors under "metabolism" and add it as a common cause of W_I and W_F, on top of the arrow W_I ---> W_F. #Bookofwhy

2.28.20 12:37am - A Zionophobe is he who says that Israelis, as biological objects "have the right to live in peace" , but will never say that Israel, as a sovereign state of the Jewish people, has the right to live in peace.

2.28.20 12:25am - (1/ ) On Dec. 9 1666, the Rabbis of Constantinople excommunicated Shabbetai Zvi, a self-proclaimed messiah who led a sect of X-Jews to speak against the greater Jewish community. In Feb. 24 2020, 347 Rabbis rebuked Bernie Sanders for speaking against the greater Jewish community .
2.28.20 12:25am - (2/2) If I were a Rabbi, I would move to pronounce Bernie Sanders a X-Jew. This would not diminish his standing as a Messiah, but it would revoke his credentials to speak as a Jew.

2.28.20 12:08am - Three days ago, 347 Jewish Rabbis rebuked Bernie Sanders for maligning AIPAC and Israel. I have not heard any of them say: "Hey, Bernie you either are, or act like a Zionophobe, and this does not reflect well on your moral character!"

2.27.20 2:49pm - In Lord's (1967) paradox, one constraint in his own plot is, cov(S, WI)=cov(S, WF), where WI & WF are two repeated weights. To embrace it in his DAG, @yudapearl had to claim that i) S->WI (a) equals the sum of ii) S->WI->WF (ac) & iii) S->WF (b): a=ac+b. Not super satisfying 2 me

2.27.20 9:13am - I do not usually engage in futuristic speculations, but this interview with Oren Etzioni on "How to know if artificial intelligence is about to destroy civilization" strikes me as a serious thought provoking session. #Bookofwhy

2.26.20 5:00pm - Sure. The best place is this personal story from the Jewish Journal I also have some photos that go with it: Note the neighborhood bully, it's me, no messing around.

2.26.20 9:32am - Who is a proud Jew? S/he who calls Bernie Sanders a ZIONOPHOBE, in public. As far as I know, no one has done it yet! May I have the honor? Proudly so!

2.26.20 8:19am - For readers who are asking about software tools to implement do-calculus, data-fusion and other goodies mentioned in @Bookofwhy, here is

2.26.20 7:42am - No Jew is a proud Jew who fails to call @SenSander a Zionophobe, in public. And, as far as I know, no one did. May I have the honor? Proudly so!

2.26.20 7:35am - Sharing a rare and breathtaking view of the Dead Sea coastline, which I, as a boy, knew only as a sub-burned yellow desert.

2.26.20 7:09am - (Replying to @yudapearl @AngeloDalli and 3 others) Halpern's complete axiomatization of counterfactuals, and Spirtes proof that d-separation holds in linear feedback systems. But we lack an algorithmic method, like do-calculus to tell us, in the non-parametric case, when a query is identifiable and when it is not. #Bookofwhy

2.26.20 7:04am - (Replying to @AngeloDalli @ylecun and 2 others) Depends on what we mean by "solution". We have a satisfactory semantics, given by the 3-step process, once the model is fully specified. This is the logical basis for next step: identification. It isn't as easy as in recursive models, but some nice results do exist, among them

2.26.20 12:59am - Readers of #Bookofwhy who are in the healthcare practice or research will find this comprehensive survey of Bayesian Networks to be most informative: It covers predictive, diagnostic as well interventional applications.

2.25.20 9:27pm - Don't under-rate yourself. If you read #Bookofwhy you already know more than the millions who have not read it, and that's quite a lot.

2.25.20 9:22pm - (Replying to @AngeloDalli @ylecun and 2 others) There is a paper by Richardson and Lauritzen which gives cyclic interpretation to chain graphs. #Bookofwhy

2.25.20 9:18pm - I had to write them in this order because, lacking courage and audacity, I could not face the public till I was sure I was right - a personality weakness. But now that I am sure, I do not buy the wimpy "multiple approaches" myth - another personality weakness. #Bookofwhy

2.25.20 8:59pm - (Replying to @yudapearl @mribeirodantas and @diomavro) on the "linear microscope", and, which clearly distinguishes the two tasks and shows that the partial regression BETA(yx|z) is independent of the value z of the variable Z. This is no longer true in non-Gaussian models #Bookofwhy

2.25.20 8:49pm - Bias in predictive tasks occurs in finite-sample estimation, but not asymptotically, as in causal tasks. Unfortunately, most texts on regression analysis confuse the two, as do texts in Economics and Social Science. I must therefore overcome modesty and refer you to my papers

2.25.20 8:31pm - (Replying to @mribeirodantas and @diomavro) Bias due to bad-control only hurts us in causal tasks, where our target is a causal effect. In purely predictive tasks there is no such notion as "bias", and traditional wisdom says: the more information you have the better the prediction. True in linear Gaussian model #Bookofwhy

2.25.20 12:39pm - In my blog comment, in case you could not find it, I provided three relevant links: 1. 2. "Bayesianism and Causality, or, Why I am Only a Half-Bayesian" -- 3. Causality, section 11.1.1 #Bookofwhy

2.25.20 10:36am - O.A. Arah has called my attention to a lengthy discussion on Gelman's blog: concerning a paper on "causal inference with Bayes Rule" I have added a comment to the former, and will examine the latter. #Bookofwhy

2.25.20 8:35am - The channel you are now using, by following @yudapearl , is an educational channel, dedicated to the scientification, algorithmization and democratization of causal inference. #Bookofwhy

2.25.20 3:23am - (Replying to @yudapearl) On this occasion, let me also recount the availability of two resources. (1) A searchable file of all our 4,263 Tweets., and (2) A listing of all UCLA-CSL papers and technical reports, dating back to 1965. Happy sailing #Bookofwhy

2.25.20 3:23am - My faithful Twitter counter says that, today, our educational channel has garnered the attention of 30K followers. This encourages me to continue our journey towards the scientification, algorithmization and democratization of causal inference. Thanks for the encouragement.

2.25.20 3:07am - (Replying to @AngeloDalli @ylecun and 2 others) Do you mean cyclic causal paths? (ie. feedback loops)? If so, please examine causality p. 215-217. There is a robust solution, though the identification phase is trickier. #Bookofwhy

2.24.20 5:48pm - (Replying to @jim_savage_ @steventberry and 5 others) For structural economists, an agreement on "non-parametric" should come easy, it is a fancy surrogate to "structural", namely, properties that emerge from the bare structure of the model (excluding for example shape restrictions, or distributional assumptions) #Bookofwhy

2.24.20 5:20pm - (Replying to @steventberry @jim_savage_ and 5 others) The "honor" of structural models is currently being restored through (1) clear definitions (which precede identification) and (2) tools (eg graphical) for liberating models from their non-structural burdens and unleashing truly non-parametric identification properties #Bookofwhy

2.24.20 4:52pm - (Replying to @steventberry @jim_savage_ and 5 others) As a student of philosophy I see it as a challenge to track down sources of confusions in various disciplines and, although econ. is still an enigma to me, I can see interesting patterns in econometric thought, going from inadequate notation to inadequate definitions. #Bookofwhy

2.24.20 1:26pm - (Replying to @steventberry @jim_savage_ and 5 others) It feels great to be in agreement: "The answer is: structural." Perhaps you also share with me another conjecture, that much of the confusion in Econ emanates from not having a crisp, undisputed definition of Y_x, in terms of a (fully parametrized) structural model. #Bookofwhy

2.24.20 12:17pm - (Replying to @jim_savage_ @steventberry and 5 others) Thanks for re-posting. The word "counterfactual" appears dozens of times in these slides. Perhaps they can clarify how traditional econ defined Y_x(u) or, even more ambitiously, which of "structural vs. reduced" form is more convenient for evaluating E[Y_x|X=x', Y=y']? #Bookofwhy

2.24.20 12:16am - I am still not clear if there is anything new in p. 97 that econ students can use, or is it all 3rd-week freshman class. I was ready to offer a big apology for wrongly believing in the former, but I am not getting a definitive affirmation of the freshman model. Help! #Bookofwhy

2.23.20 5:58pm - (Replying to @steventberry) I am counting the steps needed AFTER identification. How do we compute the expected salary of Joe who now has skill level Z=1 had he had one more year of college (see Fig. 4.3 in I do not know how to skip the 3-steps of p.97, please teach me.#Bookofwhy

2.23.20 3:21pm - (Replying to @steventberry and @PHuenermund) Sorry this tweet got out of thread

2.23.20 3:18pm - Yes, I recall a lively discussion with Heckman on the definition of "effects": But counterfactuals like E(Y_x|x',y')are a bit more complicated, needing a 3-steps computation, which I have not seen in the econ. lit. Perhaps you do it in one step? #Bookofwhy

2.23.20 1:25pm - I am happy to hear that counterfactual questions are now solved in freshmen econ. classes, and it was only my bad luck that I could not find anyone who could solve it in 1998. I am still curious, though, what definition of the counterfactual Y_x is used ? #Bookofwhy

2.23.20 5:51am - I hate to disagree, but I must note my disagreement with the analogy. Econometrics starts with a (structural) model that every economist can comprehend. PO starts with ignorability assumptions that almost NO epidemiologist can comprehend. See 3 bullets

2.23.20 12:16am - I am still not clear if there is anything new in p. 97 that econ students can us e, or is it all 3rd-week freshman class. I was ready to offer a big apology for wrongly believing in the former, but I am not getting a definitive affirmation of the freshman model. Help! #Bookofwhy

2.22.20 6:19pm - We are in receipt of a course on "Superforcasting",, part IV of which says: "Superforcasting requires countefactualizing". This means that an algorithm equipped with a causal model, should out-predict one based on data alone. True, but how? #Bookofwhy

2.22.20 1:32pm - Commenting on my "Two new weapons.." a reader claimed: "Protestantism led the way in Zionism, not Jews". Reply: Not so; my grandpas prayed 3 times a day "Restore us in sovereignty to our land" (Hamazon)

2.22.20 11:41am - (Replying to @pierre_rochard @dopamine_uptake and 5 others) I am eager to agree, but I do not know what "s2f shills" is. Culturally deprived.

2.22.20 11:27am - (Replying to @dopamine_uptake @TheStalwart and 5 others) If by "useful" you mean "causative", then the question need be rephrased to read: What causal assumptions are needed to make Granger causality "causative"? I recall (vaguely)

2.22.20 11:18am - (Replying to @pierre_rochard @dopamine_uptake and 5 others) Another mathematical result says that there is no way to distinguish spurious from non spurious correlations without causal assumptions. See These ought to have been part of econ. education, but see what happened: #Bookofwhy

2.22.20 9:41am - (Replying to @TheStalwart @dopamine_uptake and 5 others) It is not exactly a mathematical dispute as it is a mathematical fact that one cannot define causal relations in the language of probabilities, nor can one claim causal relations without causal assumptions. The latter are missing from Granger's causality. #Bookofwhy

2.22.20 7:30am - The shameful complicity of the journalistic fraternity will always embitter my memories. Like those "journalists" who cooked up the story of Mohamed Al Dura, whose picture was used by Danny's murderers.

2.22.20 7:04am - Well put. "facelift to Sharansky's 3 Ds"; so badly needed. The 3 D's served their purpose, but they also reinforced the outdated fixation that anti-Zionism is evil only in as much as it contributes to antisemitism, which has spawned and legitimized Zionophobes like Linda Sarsour.

2.21.20 9:07am - "How much the system IMPROVES the likelihood of a good outcome" is precisely the counterfactual term that PNS captures. #Bookofwhy

2.21.20 5:21am - And let's not forget the Passover Hagada: Whoever did not say the following three things in the past week, has not fulfilled his obligations to students on campus: Zionophobe, Zionophobic and Zionophobia. Words count, and silence corrupts.

2.21.20 5:00am - Cornell's President, Martha Pollack, should serve as a role model to all academic leaders, especially to Berkeley's Chancellor, Carol Christ, who said last week: "Jewish students have the right to feel dismay and concern" instead of: "I, personally, feel dismayed and concerned."

2.21.20 3:28am - (1/ ) Legality aside, I would assess how likely I am to encounter a complication never seen by the AI surgeon, which a human surgeon can handle through her knowledge of anatomy. It actually happened to me, when a famous surgeon quoted us 90% success record but, when asked
2.21.20 3:28am - (Replying to @yudapearl) (2/ ) how many cases like THAT have you seen? He said: "None. I have never seen a lung as bad a lung as this." We cancelled the surgery, and my wife is still alive. Remarkably, counterfactual logic permits us to bound the probability that a given patient will benefit from a given
2.21.20 3:28am - (3/ ) treatment, once we have a combination of both experimental and observational data. See tight bounds on PNS (prob. of necessary and sufficient causation)

2.20.20 3:01pm - (Replying to @omaclaren @AdanZBecerra1 and 2 others) Who would care about identifiability if is were not a necessary condition for estimability? And who would care about "a particular class of models" if it were not a faithful representation of your very own understanding of the world? "Estimability" reminds us or that. #Bookofwhy

2.20.20 12:56pm - I define an expression to be "estimable" if it is made up of sums and products of conditional probabilities, and the underlying distributions are non-pathological. #Bookofwhy

2.20.20 12:44pm - It was an honor to speak to Alums for Campus Fairness (not 'algorithmic fairness') and to share with them my two defensive weapons. I hope these weapons will be used effectively by all colleagues and students who are subjected to Zionophobic harassments on their campuses.

2.20.20 10:36am - I can forgive them all, including the last one, is he only did not sell his Jewish soul to Linda Sarsour and other Zionophobes.

2.14.20 6:40am - Today we will say Kaddish in memory of our son, Daniel If you happen to be in Encino, CA, join us at 5pm at Valley Beth Shalom. Else, please join us spiritually. After the Kaddish, please sing with us: Tsadik Katamar Yifrach (Like an Immortal Palm Tree)

2.20.20 6:25am - The gives indeed a good overview of causal inference, albeit from a fairness perspective, especially the CI chapter I'm glad they took Berkeley's sex-discrimination debate as a test case, and built new ideas around it. #Bookofwhy

2.20.20 3:15am - (Replying to @ChengSoonOng @mrityunjay_99 and @suzatweet) Beg to differ. The test is fruitful, necessary, available and has led to only few debates, mostly among traditional statisticians, who could not reconcile to the fact that the most powerful language they knew, prob. theory, cannot do what they always wanted to do (CI). #Bookofwhy

2.20.20 3:08am - (Replying to @ChengSoonOng @mrityunjay_99 and @suzatweet) The test is as formal and crisp as mathematics Herself. See nfGPQ#page=5. Any inference defined in terms of probability distributions of observable variables is Statistical, not causal. #Bookofwhy

2.20.20 2:51am - Bergstein's article raises all the important issues. I would only wish it was titled "what AI can and will do" instead of "what AI cannot do", which reminds me of the old speculations of Penrose, Searle and Dreyfus. #Bookofwhy

2.20.20 2:25am - The key point is "have the architecture" which is an innate "template" to store and interrogate what we learn through experience. What we cannot do having this template we surely cannot do having to acquire it from scratch. #Bookofwhy

2.20.20 2:14am - Glad you asked me, the least biased arbiter. The @Bookofwhy will equip you with elementary tools to solve elementary problems that even the most sophisticated t heories in statistics, logic and NN cannot solve. It's "low-level" on the one hand and super "high-level" on the other.

2.20.20 1:58am - (Replying to @ChengSoonOng @mrityunjay_99 and @suzatweet) What I meant is a chapter on the mathematical framework (called CSM) that unifies all these seemingly different "approaches" or "styles". (BTW Granger causality is not causality). #Bookofwhy

2.19.20 4:47pm - Statisticians are not to be undervalued. Rather, they need to be awakened to the age of causation and internalize the fact that the language of statistics need be extended to adapt to the new age. #Bookofwhy

2.19.20 7:07am - With the number of enemies I have on Twitter, I am not sure my retweeting would serve you well. Watch your steps carefully and ...(almost forgotten).. take a look at the Sequential Backdoor for more pleasant "time varying" journeys in the future.#Bookofwhy

2.19.20 5:47am - If do-calculus is a special case of X, I would be thrilled to see how X derives the front-door formula. It reminds me how, not too long ago, epi folks considere d Rung 1 and 2 and 3 as the same rung, see #Bookofwhy page 152. Should epi students wait another 25 years?

2.19.20 4:38am - Replying with my retweet to keep the thread unbroken, and to add an iota: I am willing to help an interested author put together a comprehensive textbook on MLL that includes a CI chapter. Will do so anonymously and delightfully. #Bookofwhy

2.19.20 4:18am - Where is MML heading? It seems to be heading towards another period of Rung-1 existence, awaiting textbook authors to discover that a unifying mathematics of causality does exist, within which all seemingly different "approaches" have converged. A golden opportunity! #Bookofwhy 6

2.18.20 12:24pm - When I see political scientists learning from epidemiologists, I see signs of progress, especially when they're using graphical models, like here: What I do not understand is why they aren't using do-calculus, to get identification with no tears #Bookofwhy

2.18.20 10:52am - Israel celebrates the disappearance of an island in the Sea of Galilee: On a personal note, this island was completely covered (a latent variable) during my army-service days, 1953-1955.

2.18.20 2:30am - My point is that Jews were not actively "criticizing" revealed truths. "Debates" were forced upon them. What provoked hostility was not the persistence of their beliefs as much as the persistence of their biological existence; Wrong beliefs had to show up as abnormal existence.

2.18.20 2:06am - This new paper, "Conditional Path analysis in singly-connected path diagrams" provides interesting factorization results to the "Linear Microscope" of #Bookofwhy

2.18.20 12:21am - There is some truth to it, as told in #Bookofwhy p.244, but not exactly as Clark describes it. What I learned from Spirtes (aug. 1991) was arrow-cutting for intervention, but the causal interpretation of DAGs came al ready in 1987, eg and others, 88-90.

2.17.20 1:35pm - The protest was legally tolerable yet morally despicable. Berkeley is waiting no w for the Chancellor's personal condemnation of such Zionophobic disruptions on moral, not legal grounds.`

2.17.20 12:18pm - Debra is right in NOT equating the two. By saying: "anti-Zionism IS ANTISEMITISM " one implies that the latter is more despicable than the former and, indirectly, bestows some leniency onto an eliminationist ideology, which is more dangerous than its emotional rival.

2.17.20 7:25am - (Replying to @ghost_of_roger @deborahlipstadt and @IzaTabaro) The only way to cure this misconception is to use the term Zionophobia, which makes people jump to their feet and ask: What do you mean? Tell me more! The gullible don't understand that being anti-Zionist means being ugly and immoral.

2.17.20 7:12am - anti-Zionism is crisply defined, unlike anti-Semitism which invites infinite discussions and disputations. Perhaps because the latter fell into ill-repute after the holocaust, for a short period of time.

2.17.20 7:06am - Really? I am truly delighted. It is such a fitting metaphor, that I thought would help statisticians understand why causal notions exist only in the 3rd dimension, and what they need to do to capture it. Still waiting. #Bookofwhy

2.17.20 6:59am - You sound like a reasonable person, but you speak with Zionophobic slogans like "apartheid policies". I refrain from using such terms for fear of being identified as gullible and misinformed, who buys into BDS propaganda.

2.17.20 4:35am - Logic does dictate that, if you read my elaboration on why the logic of Zionopho bes makes most Jews suspects of criminal activity, unless proven "innocent," lik e those fringe X-Jews (the new Maranos), who think distancing themselves from Is rael would protect them from harassment.

2.17.20 4:22am - Good point. We should really make clear it's not "probabilistic models" but "mod els of reality", as opposed to models of data. I hate to say "causal models", b/ c ML folks would say "It must be awfully complicated". "Reality" on the other ha nd is supposed to be simple. #Bookofwhy

2.17.20 4:03am - Emil Coman just sent me a new toy, a Stata module to do-the-DAGs: Some of our readers might find it useful to have a '1 stop shop' to do things th ey couldn't do before. #Bookofwhy

2.17.20 1:04am - I heard people say "he was a spy", or "he had no business investigating Pakistan " This is the first time I hear someone saying it was "propaganda". The capacity of certain cultures to deny reality exceeds our capacity to be surprised.

2.17.20 12:31am - The reason I said "insurmountable obstacle" is that it cannot be solved by model -free ML methods. The examples in show that the same shi ft in probability may require two different repairs, depending on which structur e caused the shift.#Bookofwhy

2.16.20 9:17pm - Mike, you made my day! Finally, a Jewish leader with a spine! Who recognizes tha t we must change the conversation from whether anti-Zionism is or is not antisem itism to whether it is genocidal or just disgustingly ugly.

2.16.20 8:03pm - (Replying to @MisterSquirrel2 @LomaahhMore and 4 others) What's the point? That distancing ourselves from Israel should protect us? The y (Zionophobes) disrespect Jews who trade their values for social acceptance, so mething they wouldn't do. They will never stop hating us just because "hate" is socially unacceptable.

2.16.20 7:52pm - Dear @BDSreport . I just tweeted this Q. Why don't you use the word "Zionophobia"? and turn it into the ugliest word o n campus? When we use the word "antisemitism" we bestow some legitimacy onto ant i-Zionism, legitimacy that it does not deserve.

2.16.20 7:29pm - In my entire life I havn't heard anyone calling for the elimination of Iran or C hina from the surface of the earth, or calling whoever opposes the elimination " criminal" or subhuman. Linda Sarsour just said: Any one who humanizes Israelis i s on the side of the "oppressor".

2.16.20 7:13pm - This harassment will continue as long as we play their game that, unlike antisem itism, anti-Zionism has forgivable qualities - it doesn't. The harassment will s top when we raise counter-signs saying: "Its worse!" or "Zionophobia kills," per haps even in lectures on Islamophobia.

2.16.20 11:12am - (Replying to @Prof_Livengood @deborahlipstadt and @IzaTabaro) I am not inventing those terms, I am just quoting my Zionophobic colleagues: " Yes, every Israeli who does not support the dismantling of Israel and turning it into a Palestinian-majority state is complicit with the crime of the century." I read what they write.

2.16.20 10:57am - Both intention and material support are a statistical fact. Recall however that the crime is in Israel's very existence, so are the intention and the support -- both are aimed to help Israel survive.

2.16.20 10:09am - I would say "structure" - everything then follows.

2.16.20 9:23am - I don't think we can afford to wait for them to realize that they are committing suicide, for they will drag all of us into the abyss. It's time to close ranks, and save American Jewry from itself -- Israel and Zionism have the inspirationa l power to do it.

2.15.20 10:38pm - (1/ ) The book you mentioned oX.rEbXEZHHE is indeed a penetrating and unparalleled window to the mystery of Jewish identit y. It was created at an emotional moment of history, when one was able to convin ce 150 Jews, from all walks of life, to take time off their busy
2.15.20 10:41pm - (2/ ) (Replying to @yudapearl @kisjdmls and @nprscottsimon) schedules and address a question they would normally push aside: "When you say " I am Jewish" what do you really mean?" The result was an authentic panoramic vie w of how Jews see themselves in the 21st century. We were gratified to know that , 17 years after publication, a book
2.15.20 10:42pm - (3/3) inspired by Danny's last words is still informing and empowering curious minds; quite a few schools are giving it as a gift to their graduating youngsters.

2.15.20 2:42pm - (Replying to @Ostrov_A) Your tweet has made my day, thanks. And may that, from now on, the inspirational story of Israel will be told through its miracles, not through sacrifice.

2.14.20 2:56pm - Sharing a poem I wrote on the tenth anniversary of Daniel's death: The Lion's De n anniversary-of-his-death/

2.14.20 6:40am - Today we will say Kaddish in memory of our son, Daniel If you happen to be in Encino, CA, join us at 5pm at Valley Beth Shalom. Else, please join us spiritually. After the Kaddish, please sing with us: Tsadik Katamar Yifrach (Like an Immortal Palm Tree)

2.14.20 5:22am - I am surprised that, with all their wisdom and resources, the authors do not consider combining the two, as advocated here: Can anyone inform them that the tools of combination are available. #Bookofwhy

2.14.20 4:07am - For ML folks, "domain adaptation" connotes an insurmountable obstacle. For CI folks it is a causal graphs problem embraced under "transportability" theory. This paper views the problem as Bayes inference on graphical models. #Bookofwhy

2.13.20 2:20pm - Thanks for posting. I was not aware of this paper, which is the most comprehensive I've seen on instrumental inequalities and their extensions. Highly recommended to those who wish to test the assumptions behind instrumental variables #Bookofwhy

2.13.20 12:51pm - It's great to see the principles of causal inference propagate through multiple eyes, multiple lenses and multiple wavelengths. At the end, I believe we will have a super bright sun beam that even blind heathens would notice. #Bookofwhy

2.13.20 6:31am - From my recollection, the Instrumental Inequality is tight for binary treatments, see, and we also have tight bounds for monotonicity , see Thus, if we find violation, its good to know if its structural or functional #Bookofwhy

2.12.20 8:59pm - The insight of Yossi Klein Halevy and his careful choice of words have always swept my admiration. This piece in particular deserves careful attention, it is half prophesy and half calling.

2.12.20 6:40am - A new paper has reached my desk "Revisiting Sequential Attributable fractions" . "Attributable fractions" is an old epidemiological concept that has been treated badly when statistics ruled epi. Interesting to see it in causal dressing. #Bookofwhy

2.12.20 7:23pm - Epi was "ruled by statistics" when it allowed causal concepts such as "attributable fractions" and even "confounding" to be defined in statistical terms, and no one complained. This was "ruling" by language and thought. #Bookofwhy

2.12.20 1:10pm - Sample complexity has been both an enigma and Achilles heel of causal, non-parametric estimation. This paper establishes sample and time complexity bounds on non-parametric identifiability, thus bringing CI closer to the general ML fold. #Bookofwhy.

2.12.20 6:19am - At the request of readers, I am sharing a transcript of my ACF talk on "Two New Weapons for Reclaiming Israel's Posture on Campus" Use them while you can in defense of equity, diversity and inclusion.

2.12.20 12:51am - For those of us still interested in Lord's paradox, and his question: "how to allow for uncontrolled pre-treatment differences between groups", I've added a comment to my post, examining why statisticians have given up the challenge prematurely #Bookofwhy

2.11.20 1:26pm - As Israel celebrates #WomenInScienceDay I can't help recalling how, in 1960, my EE class at the Technion had 4 female students among 120 males (1 of the 4 was my wife). Today the percentage of women is 40% ( 20% of them Arab). A more inviting culture for a male student.

2.11.20 6:26am - I used "head-to-head" in the Bayesian Network paper of 1985 and thereafter. But "collider" became more popular later, and I believe it was coined by the CMU group of Spirtes etal in 1994. #Bookofwhy

2.11.20 6:14am - I wish someone rewrites this wiki entry. I have never heard of a book that is "laying the foundations of the modern debate on causal inference". The writer tried so hard not to take sides that naive readers may take CI to be just one big "debate". #Bookofwhy

2.11.20 5:58am - Warning! Good paper, but I would not recommend expressions like: "A DAG is one method that investigators may use to clarify which variables have causal interpretations in statistical models."1st, NO method can clarify that. 2nd, DAGs are causal, not statistical models. #Bookofwhy

2.11.20 2:17am - It was a pleasure to present the logic of "Emancipating our Identity" and to see how the word Zionophobia gives people the power to express what they have been yearning to express all along. I am now convinced, Zionophobia WILL become the ugliest word on campus.

2.11.20 1:47am - (1/ ) This is one of the most perplexing question in Jewish history: to embrace or to shun. King Solomon said: To every thing there is a season... a time to embrace, and a time to refrain from embracing (Proverbs 3). I feel we are approaching a tipping point in our existence where
2.11.20 1:47am - (2/ ) we should CLOSE RANKS rather than EXTEND TENTS. Those who trade core values (eg, Israel) for social acceptance are becoming existential threats, thus making a painful community split or a unilateral secession almost inevitable.

2.11.20 1:26am - (Replying to @nicidob @maximananyev and @TradeandMoney) Recalling that "what's occurring" in not the data, but the ropes behind the data, "a sketch of what's occurring" is precisely what a causal model gives you, and what MHE pretends to ignore. #Bookofwhy

2.11.20 12:16am - I am not after definitions. I am intrigued by your accurate description of day-to-day practice of social scientists, a confession I could not obtain from practicing social scientists, or from authors of econ. texts. No qualms. #Bookofwhy

2.11.20 11:03pm - I think you have summarized the current practice of "natural experiment" fairly well, though MHE gurus would prefer to dress it up in more socially accepted terms, e.g., Note, matching cannot undo model-blindness. See #Bookofwhy page 274.

2.10.20 10:40pm - I could not let this phrase pass w/o a comment: "a pretty standard regression analysis. It's causal in that the coefficients show predictive power". Now I do not know where to begin. Perhaps here @Bookofwhy? Perhaps Wright (1920)? Perhaps the "microscope"

2.10.20 2:21am - Your tweet has enticed me to share a nostalgic piece from my ancient past. It is a popular exposition of the hopes that superconductivity inspired in the 1960's: . Enjoy. Nothing like #Bookofwhy though.

2.9.20 11:48am - Agree with @eliasbareinboim . This paper gives a more extensive and formal characterization of notions such as autonomy/locality/invariance that are usually attributed to causal relations. #Bookofwhy

2.9.20 6:50am - (Replying to @yudapearl and @attilacsordas) Sorry, I meant do(slippery), not do(wet).

2.9.20 6:48am - It also contains answers to: "Who needs causality if all you want is prediction". As I once tweeted, ask any pollster what the secret is to good prediction and the answer will soaked with causal vocabulary. Why? Integrating data from multiple sources (Fusion) is a causal exercise

2.9.20 6:24am - I watched this Ted talk in 2016, impressed by the capability of her system. Last month she gave an interview at the Technion, Israel, (our alma mater) and, when I watched it again I noticed that this is the best promotion for causal networks in existence (not counting #Bookofwhy)

2.9.20 6:36am - (Replying to @BenjaminShender) You mean conspiracy theories that emanate from hate are genuinely believed by the victim (eg., Roger Waters) and cannot be labeled "dishonesty", despite his rational knowledge of the truth. Interesting. My heart goes out to him.

2.9.20 6:24am - I watched this Ted talk in 2016, impressed by the capability of her system. Last month she gave an interview at the Technion, Israel, (our alma mater) and, when I watched it again I noticed that this is the best promotion for causal networks in existence (not counting #Bookofwhy)

2.9.20 6:05am - The journey to predict the future: Kira Radinsky at TEDxHiriya via @YouTube

2.9.20 3:55am - (Replying to @attilacsordas) Banana peel is a good instantiation of do(Wet), but the DAG does not need such specific external forces. Just tell it do(Wet) and it will imagine one on its own. On page 71 we will indeed represent do(Wet) by a force F(Wet), but why load the DAGs with redundant nodes? #Bookofwhy

2.9.20 2:45am - (Replying to @EpiEllie) Sorry, this tweet was meant for @EpiEllie and the toddler who was so impressed by what he thought he should have known. #Bookofwhy

2.9.20 1:55am - (Replying to @Irishchutzpah) Why give Roger Water the benefit of a "thinking disorder" or "hereditary disease". He knows the truth about Israel and still acts as if he believes the slogans against it. That makes him "dishonest", not "sick".

2.9.20 1:21am - (Replying to @Spinozasrose and @m_pezzolla) It is not "hatred" nor "antisemitism" but a straightforward continuation of the culture of deceit, as conveyed to The Guardian by Ehud Barak The more of it, the more its character is exposed and understood.

2.9.20 12:33am - (Replying to @Claire_Voltaire) Bernie and @ewarren have not learned from Corbyn's defeat, they think those grand "coalitions" represent anyone but their creative slogan writers. Like Corbyn, they underestimate voters' intelligence and voters' sense of right and wrong.

2.9.20 12:21am - (Replying to @PHuenermund) If you read the "linear microscopes", here and here, you will know more than what most Econ&PoliSci will ever know. And that's a lot. @Bookofwhy

2.8.20 11:45pm - Which raises a deep cognitive science question: Can dishonesty be localized? Can an artist be dishonest in his moral makeup and, simultaneously, honest in his artistry?

2.8.20 11:26pm - From my current reading of the article I do not see the super-breakthrough advertised. I see a serious attempt to narrow the set of data-compatible structures, using recently proposed (2006-) asymmetry-producing ASSUMPTIONS. Worth watching & reprimanding the advertiser #Bookofwhy

2.8.20 10:44pm - I cannot overemphasize the importance of page 24 for understanding: "intervention", "parents", "invariance" and other notions, often treated informally. Chapter 3 later explicates the ramifications of this one page, and gives us backdoor, do-calculus and much more, #Bookofwhy

2.8.20 10:04pm - My understanding of "measurement errors" is summarized in two papers on "effect restoration", linked via and I have the hunch the answers can be found there. #Bookofwhy

2.8.20 9:56pm - The advertisement for this super-breakthrough in causal inference is unfortunately too sketchy to comprehend. I hope someone attends the AAAI-2020 talk and tell us, in technical language, how they overcome the First Law: No causes in - No causes out. #Bookofwhy

2.8.20 10:20am - This is something (some) American Jews do not understand (e.g., New Israel Fund, post-Zionist Synagogues) as they accelerate the disappearance of American Jewry by weakening the moral basis for Israel's sovereignty.

2.8.20 9:46am - (Replying to @AngeloDalli) It is surely possible, because we can always fit a function to data. The question is whether the fitted function (w/o causally accounting for daily temperature) would correspond to our understanding of what banning ice cream will do to the world. #Bookofwhy

2.8.20 3:09am - (Replying to @ZionessMovement @JewishCJustice and 2 others) Anyone knows whose Bernie's surrogate is? Who makes up the questions? Whether anyone there is prepared to ask the hard questions? And where we can find the program?

2.7.20 8:20pm - It's final! Eden Alene will represent Israel in the Eurovision contest 2020. When I saw her perform: I could not control my goose pimples. Thinking how her parents walked the vast deserts of Ethiopia to give their children the opportunities they now have.

2.7.20 7:53pm - Thanks for retweeting this exercise. It was proposed in Sep 2019 but, in light of our recent discussion on DL and explainability, it remains doubly relevant. Let's do it together. #Bookofwhy

2.7.20 7:28pm - (Replying to @oacarah @JohannesTextor and @ProfMattFox) Agree, the "microscope" is a useful paper, I frequently consult it when linear models come up. I wonder why the dont teach it in Stat 101, since it can be taught as an innocent exercise in linear regression, not mentioning the C-word. #Bookofwhy

2.7.20 7:16pm - (Replying to @memosisland and @ylecun) Not so. Loops are allowed. See Causality page 315, and its complete solution on page 316. Dont miss footnote 10, on my funny experience with solving a problem that economists themselves posed. #Bookofwhy

2.7.20 8:34am - (Replying to @EpiEllie and @PHuenermund) It is alien in two senses: (1) The founding fathers of econ, Haavelmo, Koopman and Marschak would rather quit academia than teach from MHE. (2)Thinking about an economic reality as an RCT is convoluted and constraining, which is why they can't use DAGs, only IV. #Bookofwhy

2.7.20 8:22am - (Replying to @EpiEllie and @PHuenermund) This is the secret of Neo-econometrics. They quote and quote and requote themselves, which helps create an illusion of progress. But see what this echo-chamber quoting amounts to in my answer to Imbens; #Bookofwhy

2.7.20 8:07am - (Replying to @nano_unanue) It does not need to be a DAG. SCM may have a cyclic structure. But when the structure is acyclic certain properties come into being what facilitate analysis and identification and do-calculus. #Bookofwhy

2.7.20 8:03am - (Replying to @EpiEllie and @PHuenermund) The other way around (remember cause and effect?). I do not like Target Trial framework (do you still call it "framework"?) because I can see which direction it takes, not the other way. #Bookofwhy

2.7.20 7:47am - (Replying to @PHuenermund and @EpiEllie) I am not sure we agreed here, on tweeter, which direction it goes. Remind me, but I am fairly sure it was not the "right" direction. #Bookofwhy

2.7.20 7:15am - (1/ ) It is refreshing to read about "Causal diagrams: pitfalls an tips" by folks who actually use DAGs, as opposed to distant commentators. I recommend however that the paper be read in the context of (1) my critique of Imbens
2.7.20 7:15am - (2/2) and (2) the examples in which deconstruct myths about TWIGs. #Bookofwhy

2.6.20 4:38pm - If you are planning to attend AAAI-20 in NYC dont miss our humble contribution to causal discovery, Monday 2/10 Technical session 10: Reasoning under uncertainty. "For the want of a nail... a victory was won/lost" (BF) #Bookofwhy

2.6.20 8:21am - More on Fairness, this time in Nature accepting and explaining the necessity of causal models to define and quantify what we mean by Fairness or absence of Fairness. @Bookofwhy

2.6.20 7:46am - (Replying to @TheAngrySemite) Mena is where your language incubated, where your heroes sang their songs, and where your mind wanders when you say "homeland".

2.6.20 5:04am - (Replying to @joedotfaith @ylecun and @federalreserve) Very interesting! I wasn't aware of those regs. What backgrounds did the authors have?

2.6.20 2:39am - (1/n) Two comments on @ylecun thread, on doing engineering w/o causal explanations. True, flight engineers can improve wing design with Navier Stokes eq. But naive users need naive explanations on why a loan application was denied, gradient-based optimization in NN will not satisfy
2.6.20 2:39am - (2/ ) this need. I touched on this aspect of man-machine communication here 2nd comment: Every heat engineer knows the 2 laws of thermodynamics, albeit not their atomic explanations. These 2 laws have saved humanity many hours of searching for perpetual
2.6.20 2:39am - (3/ ) motion machines, and many many hours reviewing and rejecting proposals for such machines. In machine learning we also have two conservation laws: 1. Conservation of interventional information (the barrier on Rung 1) and 2. Conservation of counterfactual information (Rung 2
2.6.20 2:39am - (4/4) barrier). Taking seriously the analogy with engineering, one would insist on teaching ML thermodynamics to every ML class in the country. Would Facebook join UCLA in initiating this huge educational project? Just think of the savings in perpetual motion machines. #Bookofwhy

2.5.20 4:23am - (Replying to @shamoons) No! Jihadi music is Jihadi music. "From the river..." is a death chant to human beings living in the neighborhood. Not to mention the students whose meeting was silenced by the saintly "simple chant".

2.5.20 2:02am (Replying to @arxter and @Claire_Voltaire) Cultures void of history claim "indigenous is a silly notion" Others embrace it to heighten creativity.

2.4.20 10:58pm - I've heard this Jihadi music since I was 3 yrs old. What's new is the participation of gullible American students. We, academics, are partly responsible, by begging protection from anti-semitism instead of demanding a stop to anti-Zionism - the more lethal of the two racist cults

2.4.20 3:44pm - I do not give up and claim (as did Kruskal and Lord) that the solution lies beyond statistics. Rather, I am asking: (3.1) What information is needed for a solution? (3.2) What notation would this information be cast in? (3.3) Would I be able to read this notation and judge the
2.4.20 3:44pm - plausibility of the claims? (3.4) What would I do with it, once I agree with the plausibility? (3.5) Does it have any testable implications? (3.6) Unlike my semi-revitalized colleagues, I will begin critiquing specific models only when satisfied with (3.1)-(3.5). #Bookofwhy

2.4.20 2:02pm - The last question is answered fairly well in the conclusion section of the Lord Paradox posting A revitalized mainstreamer recognizes that (1) The two clashing intuitions are deeply entrenched in statistical thinking and should not be brushed off , and
2.4.20 2:02pm - (2) Both intuitions are causal, hence, to reconcile the apparent clash between them we need a causal language; statistics alone won’t do. A fully revitalized mainstreamer goes further: (3) accepting that every causal assertions must invoke untested causal assumptions I do not

2.3.20 5:40am - Replying to @Plinz and @ArcusCoTangens) I would like to believe that defenders of mainstream listen to me, not because I have an authority in any subject, but because my arguments for revitalizing mainstream make good sense and science would benefit from the revitalizatioin. #Bookofwhy

2.3.20 1:05am - Any disdain or conspiratorial undertones on my side are imaginary at best. In fact, as I articulate here: I respectfully invite mainstreamers to join me in the effort, by temporarily halting the question "What if the model was wrong?" and attend to another
2.3.20 1:05am - , equally important question: "Suppose it was right, what would we do with it?". Attending to this question is pre-requisite to resolving causal problems such as Lord's Paradox. (or, more generally: should we adjust for base-line conditions?) #Bookofwhy

2.3.20 12:07am - (Replying to @jtrecenti @athos_damiani and @agpatriota) Perhaps you can explain "the situation" w/o DAGs. But I have not seen an explanation of why we should come to a different decision, with the same data, depending on the story. #Bookofwhy

2.2.20 11:57pm - (Replying to @MarcusCrede) First, the data show that initial weight is correlated with diet. Second it stands to reason that over-weight students would choose a dining room differently than under-weight. #Bookofwhy

2.2.20 9:55pm - (Replying to @ArcusCoTangens) The question was: what information do we need to decide correctly. This can be answered independently of the question "do we have this information?" or "do we have sufficient evidence to support the needed information?" Separating tasks does not mean neglecting tasks #Bookofwhy

2.2.20 8:34pm - Correcting a link to the Lord Paradox posting. The correct link is and it should go to: Lord Paradox and the Power of Causal Thinking. (Thanks to Stephen Leroy for noting). #Bookofwhy

2.2.20 8:25pm - (1/ ) About a dozen or so readers have offered creative proposals for resolving Simpson's paradox in the X,Y,Color scatter plot example. I can't comments on each of the proposals, but I would beg the discussants to focus on my humble proposal: A causal model is both necessary and
2.2.20 8:25pm - (2/ ) sufficient for resolving the paradox, namely, for deciding if X increases Y for a person with unknown color. I would be happy to respond to anyone who thinks this statement is in some way incomplete. #Bookofwhy

2.2.20 8:08pm - (Replying to @jtrecenti @athos_damiani and @agpatriota) What explanation is "equivalent" to DAGs which does not use DAGs? perhaps PO? or "exchangeability"? or "higher resolution?" or "context-sensitive"? I am yet to see one. #Bookofwhy

2.2.20 5:22pm - (Replying to @jtrecenti @athos_damiani and @agpatriota) Why do you say "Causal models are useful" instead of "are necessary"? "Useful" is what economists use to justify not using (see When we have same data demanding two different conclusions depending on the model, we say "necessary" #Bookofwhy

2.2.20 5:08pm - (Replying to @agpatriota @eduardohorta and @michael_nielsen) Explanations are man-made, but we are facing a decision that has true/false value: Will Joe gain a higher Y with more X ? Philosophy aside; do we or dont we have enough information in the scatter plot to decide correctly? #Bookofwhy

2.2.20 4:59pm - (Replying to @DiogoFerrari) I would say: It is impossible to "deal with Simpson's Paradox" without a causal model, and it is impossible to specify a causal model in the language of probability distributions, however intricate. See which you've cited but not taken seriously.#Bookofwhy

2.2.20 4:37pm - (Replying to @eduardohorta @agpatriota and @michael_nielsen) I interpret your hesitation to mean: It depends on the causal relationships between X, Y and Color. Agree. Absent these relationships we cannot decide if X would increase or decrease Y for that person. No need to bemoan science, we need only glance at the causal graph. #Bookofwhy

2.2.20 7:33am - (Replying to @eduardohorta and @michael_nielsen) It surely helps. And the real puzzle begins here: A person with unknown color comes in, would X increase its Y or decrease it? #Bookofwhy

2.1.20 5:49pm - (Replying to @ArcusCoTangens) Too bad for the standard. Graphs are man made. Better to define things in their organic habitat - probabilities.

2.1.20 5:41pm - (Replying to @attilacsordas and @d_spiegel) It is not the "style" that gets things in order, but the introduction of the 3rd-dimension: causation. Hopefully a new chapter in @d_spiegel next edition.

2.1.20 5:31pm - (Replying to @agpatriota) It's the appendix to a prepublication report. It proves that reversal cannot occur in causal logic, in which the notion of "good for men" is expressed in do-operator. And it claims our intuition comes from this calculus, not from statistics, where reversal does occur. #Bookofwhy

2.1.20 11:45am - Cook & Campbell spent years enumerating threats to validity -- they could not do any better, and this paper explains why: Today, that we know how to establish external validity, it is a pleasure to see all the threats that are circumvented. #Bookofwhy

2.1.20 9:55am - (Replying to @agpatriota) No! do(x) does not appear in I just identified where the paradox comes from -- i.e., clash between causal intuition and statistical logic -- and resolved it by replacing stat logic with causal logic. Plain commonsense. #Bookofwhy

2.1.20 8:56am - I do not believe any of these countries thinks this "peace plan" was meant to be a peace plan. It was meant to be a sober reminder to Palestinians that time may no longer be on their side, to stop the tantrum and accept their neighbors as permanent, equally indigenous neighbors.

2.1.20 5:10am - A colleague alerted me to a new wikipedia entry on Market Blanket It is badly written, defining MB as a property of a graph, instead of a probability distribution. The most astonishing feature, uniqueness under positivity, is not mentioned. #Bookofwhy

2.1.20 5:04am - (1/ ) Curious me took a glimpse at Michael Nielsen's blog, which triggered Gelman's discussion of Simpson's paradox Michael is blunt: "[The paradox] shows that some of our ingrained intuitions about statistics are not just wrong, but spectacularly wrong." Yet
2.1.20 5:04am - (2/ ) Yet 90% of living statisticians still believe it can be "resolved" by re-visualizing data. My esteemed discussants on The American Statistician, (eg, Xiao-Li Meng) wouldn't even utter the word "causation". And it is 2020, and Statisticians get angry
2.1.20 5:04am - (3/3) when I remind them of the date, and their students still can't cope with a paradox that has haunted statistics for the past 120 years. Pearl unfairly bashes statistics, they say. #Bookofwhy

2.1.20 4:05am - (Replying to @zakkohane and @StatModeling) I am no longer convinced this is true, after seeing readers comments on "compatibility among the experts". The only way to grasp differences and commonalities is to take ONE toy problem and try to solve it using competing "approaches". The rest is Hollywood. #Bookofwhy

2.1.20 3:56am - (Replying to @dalmiaman) Don't let them "rehabilitate" you back into correlations.

1.31.20 11:04pm - (Replying to @KordingLab @danilobzdok and 2 others) 90% of living statisticians say the same:"I am interested in non-causal questions". Fine! But when a scholar titles a blog "causal inference" you expect to see some interest in causal questions on that blog, not "you do yours and I'll do mine" #Bookofwhy

1.31.20 7:56am - My disagreement with Gelman is fundamental, because his views represent an attitude that paralyzes wide circles of statistical researchers. My initial reaction was posted on Related posts: and #Bookofwhy

1.31.20 7:28am - (Replying to @KordingLab @danilobzdok and 2 others) That's one of them, thanks. As you can see, the causal character of Simpson's paradox is avoided like a plague. Why? I am asking you as one who knows the solution: Is it possible to solve w/o graphs? Those who can't solve a problem, avoid it, and cover up the void..#Bookofwhy

1.31.20 6:13am - (Replying to @danilobzdok @blei_lab and @StatModeling) Does anyone in 2020 believes that linear regr can be used for causal discovery?

1.31.20 6:10am - (Replying to @KordingLab @danilobzdok and 2 others) In person? You must be kidding. If you knew the number of hours I spent on his blog, you would not say "different language". The example that brings everything to the surface is Simpson's paradox. I dont have the strength to dig it again from his blog. If you can, please share.

1.31.20 5:47am - (Replying to @KordingLab @danilobzdok and 2 others) I am not so sure. Why? Because all my attempts to convince @statmodeling to show us how he solves a simple causal problem (one whose solution is known in advance) ended up in failure. I lost my charm. Simpson's would be a good example #Bookofwhy

1.31.20 5:22am - (Replying to @mf_schomaker) One argument is, indeed, that it should have been encoded in the SCM. Another, that it should be treated as "disjuctive action", and extrapolated by imaging: #Bookofwhy

1.31.20 5:11am - (Replying to @sbuhai) But look at the cost of this "standard". Invited papers tell readers: "Here is a body of work that has been neglected, catching up requires that we waive the "originality" requirement." If it were not for such editorial leadership, my work would be unknown outside CS. #Bookofwhy

1.31.20 3:59am - (Replying to @sbuhai) I beg to differ. An "invited survey" has a totally different status than ordinary submission. The requirements of innovation and importance are waived, b/c the editor decides on the latter. There remains only the requirement of "representation." #Bookofwhy

1.31.20 3:37am - Retweeting my rebuttal to Imbens's paper, because readers complained about bad @@ signs in the original post. The link is still the same: #Bookofwhy

1.31.20 1:43am - (1/ ) Another important clarification. In what way does SCM embrace the logic of PO? Ans. It supports the consistency rule Y_x = Y if X=x, which is the main inference engine of PO. Consistency a theorem in SCM and an assumption in PO (part of SUTVA, which
1.31.20 1:43am - (2/ ) deals with side effects in experimental settings). In this sense one can safely say that SCM provides legitimization for the logic of PO, but rejects PO as a "framework" or "approach" #Bookofwhy

1.31.20 1:08am - (Replying to @ThomSeaton) Don't give up so easily. "Ein Lecha Adam SheEin Lo Shaa" says the Mishna (Every person has untapped potential). Even "y=mx+b" is not trivial. I once asked a statistician what Y_x is, when b is correlated with x. Repeating this story would get me into another trouble. #Bookofwhy

1.31.20 12:30am - (1/ ) Important clarification of this point: SCM embraces the counterfactual notation Y_x and its logic, but not the "PO approach" which is a research methodology built around Y_x. The fundamental difference: SCM starts with what you believe (eg DAG), PO starts with what you need (e
1.31.20 12:30am - (2/ ) (eg., ignorability conditions). Why can't PO start with what we believe? Because it insists on expressing everything in the language of Y_x, no structure, no DAGs, while SCM starts with the language in which scientific knowledge is stored: "who listens to whom?"#Bookofwhy

1.30.20 11:47pm - A chapter in history. Readers asked when the relationship between PO and DAGs was first reported. My earliest record is this '93 Stat Conference in Florence (Section 6) As I recall, the audience had NO IDEA what I was talking about - blank eyed.#Bookofwhy

1.30.20 8:27pm - (Replying to @Jabaluck) Can we skip the "unfortunate" - the fault is all mine. And can we start rejoicing what we can do today that we couldn't yesterday? Rejoice man! Show your students how to do it. And write to Imbens to open Econometrica to those who want to learn. #Bookofwhy.

1.30.20 5:18pm - (Replying to @GregZ_MD @ipam_ucla and @GlockerBen) Would have loved to hear the talk, or get the paper, if available.

1.30.20 5:14pm - (Replying to @wagonomics @causalinf and 2 others) Clarification! @yudapearl has embraced the counterfactual notation Y_x, as did PO. But he could not embrace the "PO approach", which permits researchers to assume what they need, instead of what they believe. Not their fault, PO could not express what they believed @Bookofwhy

1.30.20 4:27pm (Replying to @btshapir and @Jabaluck) Research answer: Economist: Citation for me? Citation for whom? Non-economist: Rejoice man! Look what we can do today that we couldn't yesterday! Rejoice!

1.30.20 1:51pm - For the benefit of all readers, I have compiled my "Causal, Casual and Curious" articles in one searchable page, now posted here: . Enjoy. #Bookofwhy

1.30.20 1:38pm - (Replying to @Jabaluck) You are right. I was saying it more excitedly: "Rejoice! ALL (quantitative) Economists should find it useful" . I say the same today: Rejoice!

1.30.20 5:26am - (1/2) It is FAT2020 week, so I'll not spoil the fun. But I am not convinced by Issa's argument that counterfactual logic cannot capture fairness because it presumes autonomy which breaks down in social systems. Heckman made similar arguments for economic systems: "Shutting down one
1.30.20 5:26am - (2/2) equation might also affect the parameters of the other equations in the system and violate the requirements of parameter stability." (Heckman and Vytlacil 2007). I countered here, as well as Cartwright's critique of autonomy, earlier.#Bookofwhy

1.30.20 1:32am - @causalinf , @PHuenermund , @EconBookClub, @jeremyskog Speaking of heathens, converts and imposters in these last days of DAG-deprived economics, I've posted a brief rebuttal to Imbens paper on PO and DAGs: Comments are welcome. #Bookofwhy.

1.30.20 1:06am - (Replying to @attilacsordas) Can you rephrase the question in terms of information relevance, given that we "know". ???

1.29.20 3:37pm - (Replying to @shamoons) The question was addressed to colleagues who are dancing to the eliminationist tune of "from the river to the sea", not to Governments, organizations, or phenomena. I assume you are not one of those dancers.

1.29.20 7:20am - (Replying to @TheShubhanshu and @fatconference) It is still only one page long; is that all you have?

1.29.20 7:14am - (Replying to @causalinf) Remind me what our bonus agreement was. But for the life of me, where did find a "skeptical friend"? In 2020? In economics? You must have been searching really hard. #Bookofwhy

1.29.20 6:25am - (Replying to @TheShubhanshu and @fatconference) Thanks for posting. Does anyone has her compete paper? I saw it someplace, and it seems to challenge the capacity of counterfactual logic to capture "fairness", so I am wondering what logic can capture it. #Bookofwhy

1.29.20 6:08am - (Replying to @EmraniMd) I would be careful before I give a cold-blooded Zionophobe the benefits of hereditary disease.

1.29.20 5:18am - What is Causal Cognition? I thought this article will help me find the answer: But not seeing computational models of the key concepts I feel unable understand the goals and methodology of this intriguing discipline. My weakness. #Bookofwhy.

1.29.20 3:50am - For goodness sake, she really thinks apologies were made for self promotion.

1.28.20 10:35pm - Any peace plan that does not address the elephant in the room (i.e., the religion of "From the river to the sea") will be remembered, not for what it says but for the eulogies in its funeral.

1.28.20 8:36pm - (1/2) I had the thrill of discussing this counterfactual world with @DavidDeutschOxf yesterday. He viewed it as a swift allies victory in WW-II, assisted by a sizable army of dedicated Jewish volunteers from Palestine/Israel. I would add to it: a true Palestine-Israel co-existence
1.28.20 8:36pm - (2/2) un-impeded by traumatic memories of the Arab 1948 attack on the Yishuv, and the massive displacement of Arab population in that attack.

1.28.20 5:51pm - (Replying to @EinatWilf) Palestinian detriments are well documented, but their victories are not widely known. One such "victory" was the entrapment of European Jewry, with the British Navy, and its genocidal consequences. The memory of this successful "victory" makes "From the river..." appear feasible.

1.28.20 4:57pm - Everyone of my esteemed colleagues who is dancing and prancing to the tune of "From the River to the Sea" will have to answer to Lady History when she asks: What have you done for peace?

1.28.20 4:57am - (Replying to @RussInMtl @rlmcelreath and @Lester_Domes) Will follow your lead, as long as end up with the latter.

1.28.20 2:46am - (Replying to @rlmcelreath and @Lester_Domes) Why do we assume that students are "scared" by equations? I do not think this was on our minds in 1999. If it was, perhaps we should go back and sprinkle a few more equations there. "A formula is a baked idea. Words are ideas in the oven." #Bookofwhy page. 335.

1.28.20 1:54am - As some of you heard, UCLA is under investigation by the Education Department on charges of "discrimination and harassment," filed by two Jewish students. Here is a LA Times letter written by another student, explaining the background:

1.27.20 5:52am - I am compelled to retweet your pleadings @RealSarahIdan , partly b/c you grew up across the street from where my wife did, in Baghdad, partly b/c you sang Hatikvah with me, loud and clear, in the SWU Conference, but mainly b/c the Iraqi people deserve freedom we take for granted.

1.27.20 5:10am - Sharing exciting nanotechnology news from the Technion (my alma mater), Prince Charles, and Winston Churchill's Auditorium. The connection: Nostalgia: I was present at the 1958 dedication of this auditorium, and later attended lectures there. #Bookofwhy

1.27.20 1:33am - (Replying to @mohomran) And I thought I was the first to name this on-going pattern of lies as "CULTURAL" (not racial). Where did I get the idea of being first? On my campus, UCLA, where the word "Palestinian" means "always right" "holy of holy" "never to be criticized". Anti-Palestinian??? Come to UCLA

1.27.20 1:14am - (Replying to @mohomran) It is not "acceptable" at all. I just coined this expression today, seeing this culture grow from tiny sprinkles of denying Jewish history, to denying the Holocaust, to blood liable, to denying yesterday's snow. But now, having stained the US Congress, someone has to name it.

1.27.20 12:53am - (Replying to @SachaBaronCohen @EinatWilf and 2 others) @SachaBaronCohen , you are so right, and so is @JGreenblattADL . But don't those who delegitimize Israel also aim to encourage another one? What else do they aim to encourage? And watch how close they are getting to their aim when tolerated as "just something people get wrong".

1.27.20 12:30am - Why does everyone want Rashida to apologize? If she apologizes she would surely repeat the lies in different dressings. Let her stew in her own soup as a monument to Palestinian culture of deceit. Some Palestinians will eventually recognize what she does to their reputation.

1.26.20 8:04pm - Speaking about AI, machine learning, what DL can and cannot do? This analysis of 2800 yr old Hebrew inscription tells it all. Can CI be of help? Perhaps, in case we want to generate the explanations automatically. #Bookofwhy

1.26.20 3:43pm - (Replying to @jeffreywatumull @talyarkoni and 4 others) What humans do, we observe; what NN can't do we can prove. Eg, humans shout: "Impossible" when told of a drug that is good for men, good for women and bad for a person. What would NNs do if given Simpson's data? What would a statistician do? See #Bookofwhy

1.26.20 5:21am - Be careful. Hannan Ashrawi is the Darling of the West. She always tells reporters what they want to hear. E.g., When asked if PA recognizes Israel right to exist, she always says: "Arafat recognized Israel in 1988". None ever asks her: "Do you?" They know she can't say YES.

1.26.20 4:45am - Replying to @attilacsordas) I agree that Mackie's INUS is inadequate and full of inconsistencies. I mentioned them in a Tweet discussion on Rothman's "sufficient cause PIE", but Harvard Epis were so wedded to it that I felt like I'm depriving them of a seat in heaven. See #Bookofwhy

1.26.20 4:17am - (Replying to @DaveBrady72) This discussion intrigued me b/c it challenges the capacity of counterfactuals to define "race", but I could not find the paper. Is it available? It raises the question: Are human capable of ever agreeing on the distinction between "racist" and "non-racist" practices? #Bookofwhy

1.26.20 12:53am - (Replying to @talyarkoni @dileeplearning and 3 others) Should we really cast them as "open empirical questions" when we have mathematical results on what is doable and undoable irrespective of NN architectures ?? Not sure. Empirical claims are subject to interpretation when we get to higher level AI tasks #Bookofwhy

1.26.20 12:41am - We spoke about innate templates that help reasoners do prediction intervertions and counterfactuals. This new paper in Developmental Psychology explores how young children identify "causally relevant variables" #Bookofwhy

1.25.20 10:52pm - Not entirely surprising. A culture rooted in deceit spawns leaders of like character.

1.25.20 9:57pm - (Replying to @HannesMalmberg1 and @HarryDCrane) I think of SCM in econ as reclaiming commonsense in economics (hesitating betwee foreign invasion or home-grown abberation).

1.25.20 7:10pm - The more requests I get to participate in live debates with critics of CI, the more I wish to ask if there is anything missing from the "Dialogue with A Hostile Examiner"( p. 369, that we can learn from a live debate or a staged fist-fight. #Bookofwhy

1.25.20 6:36pm - Nor will I get tired of commending those who recognized the educational power of Primer While educators debate where to start - estimation, correlation, PO, or ML - Primer starts where knowledge resides: Structure. The rest follows organically. #Bookofwhy

1.25.20 2:36pm - Replying to @HannesMalmberg1 and @HarryDCrane) Yet despite common ancestry, the new culture of "natural experimentalists" does view "structural economics", or even mere "structure," as an existential threat. If you think this is an exaggeration, I'll dig out some juicy quotes. #Bookofwhy

1.25.20 7:01am - (Replying to @cecilejanssens) Sorry if I misunderstood. My question was also simple: To decide if a {hypothesis + data} supports a conclusion takes more than "reasoning 101". We need a non-standard logic to do it if the conclusion is causal. Will wait patiently for a written summary when available. #Bookofwhy

1.25.20 6:27am - (Replying to @dr_benner) Hats off!!! Still, the entrapment of 1936-1940 needs to be taught, especially in Rashida Tlaib's school district, in Michigan.

1.25.20 6:17am - (Replying to @trumanfrancis) Great question!!! Textbooks tell us: No way! correlation is all that's needed for prediction. Butask any pollster what the secret is to good prediction and the answer will invoke causal vocabulary. Why? Data Fusion is a causal exercise. See #Bookofwhy

1.25.20 5:55am - The most memorable moment I carry from this somber Holocaust memorial is listening to the President of Germany recite the Hebrew prayer "Shehrchianu" ("to the living") instead of "in memory of the dead". So clearly telling the world what Israel is all about. I wish the hundreds,
1.25.20 5:55am - perhaps thousands of Holocaust Museums around the world will follow his insight and dedicate one exhibit room to celebrate the "to the living" message that Israel sends to the rest of the world.

1.25.20 5:25am - Replying to @cecilejanssens) What kind of logic have you used? My reading into the philosophy of science says that, if the hypothesis is causal, classical logic will not do, and we need some sort of causal logic. Which did you find most useful? #Bookofwhy

1.25.20 5:12am - Replying to @MaartenvSmeden and @trumanfrancis) Agree. It holds for any statistical model, any ML algorithm of any form, regardless how smart, and regardless how big the data. It's hard for ML folks to swallow, but I thought statisticians got over the trauma in 2002, when Denis Lindley confessed to this effect. #Bookofwhy

1.25.20 4:34am - (Replying to @trumanfrancis) Indeed. Logistic regression in itself, without extra-statistical causal assumptions, is as helpless as linear regression. It reveals NOTHING about causality. Please alert the thousands of "logistic regressionists" of this fact; textbooks lure them to think differently. #Bookofwhy

1.25.20 4:19am - (Replying to @btolgao @nntaleb and 2 others) What element of irrational human behavior do you think the introduction to #Bookofwhy should have emphasized more?

1.25.20 4:12am - Same goes for readers who ask "could you please explain what causal inference is about?". Please read: "A Dialogue with a Hostile Examiner", page 369, in The entire miscommunication between statistics and CI is encapsulated here in 3 entertaining pages

1.25.20 3:55am - To readers craving to watch a fist-fight: What's the point? You can learn so much more from solving one toy problem in CI. Still, for those who are incurably blood-thirsty, I got one for you. "A Dialogue with a Hostile Examiner", page 369 here #Bookofwhy

1.25.20 3:30am - (Replying to @BenjaminMenashe @nntaleb and @HarryDCrane) I believe @nntaleb had in mind causal inference in probabilistic setting which, of course, #Bookofwhy and each of my writings is advertising in big megaphones. Not to be confused with "probabilistic causality" as an outdated branch of philosophy. See

1.25.20 2:11am - (Replying to @CasualBrady) Sensitivity analysis with graphical models has not made it to textbooks yet." But see this paper for a solid introduction to the framework. #Bookofwhy

1.25.20 1:55am - (Replying to @CasualBrady and @Peninha_13) Agree. "can respond to" is as good. I happened to like anthropomorphic metaphors, which irritates statisticians, but they dig it better. "can respond to changes in" is also useful, unless you talk to a mathematician, for whom "non-trivial function of" is sufficient. #Bookofwhy

1.25.20 1:14am - (Replying to @HannesMalmberg1 and @HarryDCrane) If natural experimentalists where interested in methodology they would rejoice finding their work perfectly aligned with SCM, enriched with more powerful machinery. Unfortunately, they see "structure" as an existential threat. #Bookofwhy

1.25.20 12:40am - (Replying to @avicenna @joedotfaith and 2 others) Moreover, this assumption is "causal", not statistical, so it cannot be expressed as some restriction on parametric family of distribution functions, as we normally find in the p-value literature. Its an assumption of a new dimension. #Bookofwhy

1.24.20 9:02am - (Replying to @mathalytics @BatteryHorse and @HarryDCrane) No shame in self promotion, especially when institutions expected to promote new ideas prefer to perpetuate old ones. I have done a few tutorials at JSM, but a 1-day course should be more effective. Good luck. #Bookofwhy

1.24.20 8:26pm - (Replying to @joe_shipman @nntaleb and @HarryDCrane) Sorry if I misinterpret you meaning of "meaning", it reminded me how I used to argue with Stats who kept insisting on "SEM has no meaning" or "the only meaning of y=ax+eps is E[Y|x]=ax" [Holland 1988] etc. etc. The influx of Tweets from @nntaleb followers reminded me of old days.

1.24.20 8:16pm - (Replying to @HannesMalmberg1 and @HarryDCrane) The partial derivative definition is a good alternative to do(x). It has obvious problems with discrete x but, more importantly, we do not really have a calculus based on partials that can deliver something like the backdoor/frontdor adjustment formulas. Do we? #Bookofwhy

1.24.20 8:04pm - (Replying to @joe_shipman @nntaleb and @HarryDCrane) CI is orthogonal to the Bayesian vs. Frequentiss debate. See here Why I consider myself only half Bayesian. #Bookofwhy

1.24.20 8:00pm - (Replying to @joe_shipman @nntaleb and @HarryDCrane) No offense, but I can tell you are statistician. How? For my stat colleagues things has "meaning" only if they can be expressed in the language of statistics. Causal questions cannot be expressed thus, and I presume this is what is so frustrating to @nntaleb and @HarryDCrane .

1.24.20 7:51pm - (Replying to @analisereal and @HarryDCrane @analisereal) Thanks for reminding us of this ancient thread. Today I would answer it differently: "Bashing Statistics? I was generous! Count 2020 stat textbooks that mention CI, stat dpts offering CI courses, PhD's who can solve a toy problem in CI (eg, Simpson prdx?) #Bookofwhy

1.24.20 7:03pm - (Replying to @HarryDCrane) You are not missing a thing, if you don't have a causal question to answer. If you have such a question we can examine if you can answer it in some other mathematical framework, alternative to CI. I do not know of any, but am always open to learn. How about Simpson's.#Bookofwhy

1.24.20 6:54pm - (Replying to @pastramimachine and @HarryDCrane) I will never get tired of recommending Primer

1.24.20 6:52pm - I concur with @eddericu recommendation, though it depends on readers background. Some are impatient with Primer's gentleness and prefer Causality with all its proofs. #Bookofwhy

1.24.20 6:49pm - Many readers ask me this question about Primer But publishers are driven by greed, and this is one secret they won't divulge. They want you to buy an uncorrected edition by hiding all marks identifying the corrected one. Future authors, Beware! #Bookofwhy

1.24.20 6:37pm - (Replying to @HannesMalmberg1 and @HarryDCrane) Almost! It is logically equivalent to PO, but infinitely more transparent. Similar to Haavelmo? Yes, see the parallels in As to "econ defn" I have not seen one since Strotz & Wold (1960); it has been buried by the "natural experimental" craze. #Bookofwhy

1.24.20 6:26pm - (Replying to @HarryDCrane) You told me you are curious, not antagonistic, so I am searching for your questions, and all I see are statements about what causality is, and what do(x) is good for. So, what are you curious about? I am curious myself #Bookofwhy

1.24.20 6:15pm - (Replying to @nntaleb and @HarryDCrane) I assume you are asking these questions as a curious observer, and you are not in antagonistic mood, like I've mistaken Harry to be. If so, permit me to ask how familiar you are with the CI literature (it is all in the probabilistic dimension) & I will start from there.#Bookofwhy

1.24.20 6:05pm - (Replying to @RotemEren @ShekatkarSnehal and 2 others) Who is blocking? Who is great? Why block? I thought I mentioned Granger causality in #Bookofwhy, but was not sure. Thanks for quoting.

1.24.20 6:02pm - (Replying to @quantadan @ShekatkarSnehal and 3 others) I dont understand. Who is deleting? who is blocking? who is dismissive? I just got assurance that this thread is motivated by curiosity, not antagonism.

1.24.20 5:49pm - (Replying to @agpatriota @nntaleb and @HarryDCrane) Beautiful example. I use it often to demonstrate the difference between Rung-2 and Rung-3 (see Causality, Section 1.4.4). Glad you brought it up. #Bookofwhy

1.24.20 4:33pm - (1/ ) (Replying to @HarryDCrane) Since you are in such antagonistic mood, I'll answer your question using "CI for the Infidel". You are unhappy with the Economist's article on minimum wage, and you want to expose its weaknesses by pointing out how vulnerable the conclusions are to unwarranted assumptions.
1.24.20 4:44pm - (2/ ) (Replying to @yudapearl and @HarryDCrane) So far you are in good company. But now you want to pick the critiqued assumptions effectively, making sure that they are CRUCIAL, not tangential, to the conclusions. CI logic helps you decide whether a given assumption is necessary (aka sufficient) for the conclusion. Nice!
1.24.20 4:51pm - (3/ ) (Replying to @yudapearl and @HarryDCrane) I say "Nice!" b/c no branch of statistics or classical logic can give you such powerful machinery to criticize papers you don't like. But, to be more constructive, here is another chapter in "CI for the infidel" I called it a "survival kit". #Bookofwhy

1.24.20 4:16pm - Anyone with a determination to read Primer is already "up to the task", and will soon join CI. But make sure you've got the "corrected printing". How? Page ix should cite #Bookofwhy. This is the litmus test.

1.24.20 3:54pm - (1/ ) (Replying to @nntaleb and @HarryDCrane) This one is easy. In 1991, I had a quiet dinner with Clive Granger in Uppsala, Sweden. Between the 2nd and 3rd glass of wine, he confessed to me that he feels embarrassed by the name: "Granger causality", since it has nothing to do with causality, but he can't stop people from
1.24.20 3:59pm - (2/3) (Replying to @yudapearl @nntaleb and @HarryDCrane) using it; they need some way to express what they wish to estimate. I think we should honor him by echoing his understanding. An easy way to see that GC has nothing to do with causality is to look at the defining equations and note that they comprise only conditional
1.24.20 4:04pm - (3/3) (Replying to @yudapearl @nntaleb and @HarryDCrane) probabilities, no do(x) expressions, nor counterfactual terms Y_x. Bingo! We are done! Whenever a concept is defined in terms of a distribution of observable variables it can't be "causal". No causes in - no causes out (N. Cartwright) #Bookofwhy

1.24.20 3:36pm - (Replying to @littlebode) I wrote #Bookofwhy and opened a Twitter account so that each one of our 28K followers should be able to, and excited to create a MOOC on Causality. Are you there?

1.24.20 3:23pm - (Replying to @hubertpaulo and @Peninha_13) The invariance of the axioms under changing definitions is important to mathematics, but the question was why not accept ONE of those definitions and work with it as THE definition?

1.24.20 6:45am - The cover looks very familiar. I hope the translators fixed all our errors. Oh, BTW, we just commissioned a Spanish translation. #Bookofwhy

1.24.20 6:39am - (Replying to @ShekatkarSnehal @HarryDCrane and 2 others) People think I'm kidding when I say that most statistically-trained folks are still unaware of what CI is all about. Glad I have this Twitter post from which to watch the glacial progress of science. #Bookofwhy

1.24.20 6:22am - nother forgotten chapter of history. Glad it sees the light of day. My grandparents were stranded in Poland as a result of this Arab-British-Nazi circle of appeasement. It was "Genocide by entrapment", still not taught in Palestinian schools.

1.24.20 3:41am - Nice example of how SCM can serve as a laboratory to test various interpretations of familiar and colloquially used terms, in this case "incentives". #Bookofwhy

1.24.20 3:13am - (Replying to @Meetasengupta) Hilarious, if it wasn't so true. But don't blame AI. Blame big business for enslaving us to premature AI. #Bookofwhy

1.24.20 3:01am - (Replying to @Peninha_13) Euclid "defined" lines and points in terms of other undefined notions, "part of" "length" "breadth" etc. This reductive definition must stop at some point where we choose the irreducible primitives and declare them "self evident". #Bookofwhy chose "listens to" as primitive.

1.24.20 1:48am - (Replying to @tdietterich @BethCarey12 and @GaryMarcus) Let me explain. Interventional data cannot distinguish between two competing Rung-3 hypotheses. Eg, "No effect" in RCT cannot be distinguished from "treatment kills some and cures others". See gentle introduction here: A true Rung-3 Primer. #Bookofwhy

1.23.20 6:44am - Your dream is my dream. Just signed. #Bookofwhy

1.23.20 6:31am - A new dawn in sensitivity analysis! A systematic approach based on meaningful assumptions. #Bookofwhy

1.23.20 6:16am - The brightest light I see in this somber Holocaust memorial is its taking place in Israel, a country that symbolizes a metamorphosis from tragedy to hope, death to renewal and historical injustice to inspiration for all oppressed minorities

1.23.20 2:52am - Clarification: Although the paper referenced does not mention do-calculus explicitly, explainability criteria are defined in terms of interventions on a structural model M meeting condition C. So readers can see what data+assumptions would enable do-calculus to prove C.#Bookofwhy

1.23.20 2:10am - Explainability is ML's Achilles heel. This paper presents formal criteria & measures for various aspects of explainability. Its use of do-calculus gives readers immediate view of the data and assumptions needed to meet each of the criteria. #Bookofwhy

1.22.20 1:48am - (Replying to @udansk) The issue is, if I am not mistaken, whether DAGs could be taken seriously as computational models of human causal inference, as well as of mental store of scientific knowledge. This is a stronger claim than the correctness of the math. #Bookofwhy

1.22.20 1:06am - (Replying to @kerstingAIML @mitbrainandcog and 4 others) Agree. I used "we" figuratively, referring to the trendy community dominated by DL. I am still under the influence of the DL vs. CI discussion we had here a week ago. #Bookofwhy

1.21.20 11:37pm - Speaking for myself, I've gained a lot of insight about myself by playing around with various computational models. Example, I'd never guess that I had a causal diagram in my mind's eye until realizing that no other representation can account for my swiftness and versatility.

1.21.20 10:49pm - In the 1970-80's, AI was all about "computational models of mental processes". With the advent of ML, many now view AI as "algorithmic replacement of mental processes". We lost something in this transition, b/c the fun part of doing AI is gaining an understanding of ourselves.

1.21.20 2:31pm - (Replying to @ArcusCoTangens @bariweiss) and 3 others Enormous difference! The former is emotional, the latter calculated and genocidal. See

1.21.20 1:50pm - (Replying to @udansk) I said DAGs (SCMs) make causal reasoning POSSIBLE. To refute it, you need to present another mathematical object that allows the derivation of predictions interventions and counterfactuals simultaneously. I am not aware of any. Is anyone? #Bookofwhy .

1.21.20 1:40pm - (Replying to @WiringTheBrain) Do we have computational models of that "deep knowledge", ie, how it is represented in the mind and how it is accessed and transformed into a "causal diagram"? Is it not possible that causal diagram are already the deepest form of knowledge accessible to an agent? #Bookofwhy

1.21.20 1:31pm - (Replying to @andrewthesmart) Disagree. "Computational models" can be emulated on a computer. "familiarity" or "fieldwork", even "design", cannot!

1.21.20 1:27pm - (Replying to @SMBrocklehurst and @WiringTheBrain) The SCM (structural causal model) supports feedback loops (see Causality chapter 7). People use the term DAG because acyclicity gives us many computational advantages. #Bookofwhy

1.21.20 6:15am - (1/ ) Has AI done that? To a limited extent YES. Not through DL, sadly. But if you agree that "understanding" means having a representation that allows you to predict and control a phenomenon then DAGs do a fairly amazing job. Where else have you seen a compact mathematical object
1.21.20 6:15am - (2/2) from which you can derive all four: predictions, redrodictions, interventions and counterfactuals? It's primitive, yes, but it offers us a laboratory in which we can test ideas about understanding. Refinements? By all means! Knitpicking? Don't slow us down, please. #Bookofwhy

1.21.20 3:18am - (Replying to @WiringTheBrain @mendel_random and @causalinf) It is science indeed. But AI (at least me) has endeavored to ask a slightly more ambitious question: What has changed in our mental representation of a phenomenon that makes us feel: "We understand something well enough to define a hypothesis that can be tested.". #Bookofwhy

1.20.20 7:07am - I summarize this discussion thus: Progress in our century amounts to building computational models of mental processes that have escaped scrutiny under slippery terms such as "design" "discovery" "background knowledge" "familiarity" "fieldwork". etc. Saying "DAGs are not enough"
1.20.20 10:25pm - (2/3) produces no progress until one is prepared to propose a mathematical model for what IS enough. The IV criterion I presented is necessary for justification, so keep it in mind and teach it to natural experimentalists who should be jubilant seeing their IV's justified.
1.20.20 10:25pm - (3/3) We should always be open to extensions and refinements, but mystification holds us back. We had enough of that in the 20th century. #Bookofwhy

1.20.20 7:07am - (Replying to @deaneckles @thosjleeper and @aecoppock) What makes a promising candidate differ from an unpromising candidate if not features that would enable it to pass the IV criterion against beliefs about the world? #Bookofwhy

1.20.20 7:03am - (Replying to @ryancbriggs @causalinf and @mendel_random) I dont see any debate here. I see an attempt to demystify what some people describe as a mysterious process of obtaining new domain knowledge and spotting a "natural experiment" in the knowledge obtained. #Bookofwhy

1.20.20 6:55am - (Replying to @thosjleeper @deaneckles and @aecoppock) Sure. And then? Eventually, after forming the new simplified DAG, someone has to say: "Hey I got a new IV, and it's a good one too!" #Bookofwhy

1.20.20 6:51am - (Replying to @causalinf and @ez_angus) Agree. I would only equate DAGS with "theory, model, intuition, and deep institutional knowledge". Is there something in the last four that is missing in the DAG? If there is, lets add it, after deciding of course how DEEP we wish to go. #Bookofwhy

1.20.20 6:43am - (Replying to @causalinf and @ez_angus) All circularities disappear if we equate DAGs with "an encoding of what we know". Difficulties may surface if you can postulate a more refined, or more natural encoding of knowledge than DAGs. I can't, but am open to suggestions. #Bookofwhy

1.20.20 6:30am - (Replying to @thosjleeper and @deaneckles) Talking with people is always enlightening, but I am trying to demystify the process of "finding new variables, arrows etc. etc." and see how it leads us at the end to "Hey, I got a new IV". Reaching such a conclusion requires checking an IV criterion on some model. #Bookofwhy

1.20.20 3:49am - (Replying to @deaneckles) I do not see any disconnect. I see a quest to understand (semi formally) what "fieldwork" gives us that we did not have before, and where a "fieldworker" puts the newly learned variables if not in a some revised model. #Bookofwhy

1.20.20 3:07am - Let us try to give "a situation" a somewhat technical interpretation. You do "fieldwork" for 3 months, 24/7, you come back to your office and reflect: "Have I seen a "situation" that satisfies a definition?" What's in memory are you interrogating with this question? #Bookofwhy

1.20.20 12:43am - To focus the discussion: Z is a good IV if there exists a set S of measured variables such that: (1) S does not separate Z from X, and (2) S separates Z from Y after removing all arrows entering X (Causality p.248). We see that the criterion does not require data. #Bookofwhy

1.19.20 11:57pm - Preparing myself for the #WorldHolocaustForum. My grandparents were murdered in Auschwitz in 1942, my son has not been spared the wrath of this hate, and my colleagues in Israel are still under the threats of Ayatollahs designs and Palestinian deligitimization. Preparing myself.

1.19.20 11:23pm - Replying to @PHuenermund Best news from econ. since Haavelmo.

1.19.20 11:20pm - Replying to @maximananyev But you are unfair to people who can only talk "natural experiments", for whom the whole economy is a huge salad of natural experiments, and he who chooses one and ignores the others is called an "experimentalist". #Bookofwhy

1.19.20 9:13pm - (Replying to @DrMikeH49 @StandWithUs and 5 others) Amazing day indeed. My highlights: 1) Singing Hatikva with Sarah Idan (Miss Iraq 2017) and discussing the psychology of Palestinian Rejectionism with Hussein Aboubakr (Egyptian scholar, CCA)

1.19.20 8:35pm - (Replying to @AdanZBecerra1 @PHuenermund and @laura_tastic) Good detective work, thanks. The thing to do now is to petition Susan R. Bailey, AMA President Elect, to bring up this issue for public discussion. My banner would be: "Censor the claims, not the language. Causes have produced effects Myrs before RCT was invented." #Bookofwhy

1.19.20 5:45pm - (Replying to @yudapearl and @EinatWilf) "Culture or mentality?" Sadly, it holds the key to ME peace: If Palestinian Mentality becomes the center of conversation, chances are the embarrassment will sober up minds and plans. Here is a beginning:

1.19.20 5:23pm - (Replying to @EinatWilf) Agree. That's why Omar Barghouti started his lecture at UCLA (Jan 2014) with "They (Jews) are not a people..." Strange, Edward Said insisted on Arabs' exclusive right to define themselves; his disciples now insist on exclusive right to define others. Is this culture or mentality?

1.19.20 6:35am - (Replying to @EinatWilf) If EU has any obligation at all it is to ask Dr. Saeb Erekat if he ever thought whether Israel's has a right to self-determination. Last I heard from him he said: NEVER! Can a people seek nationhood on its neighbor's tomb?

1.19.20 4:15am - (Replying to @yudapearl and @causalinf) For readers asking how one can spot a natural experiment in a model, the answer is very simple and is given here among the toy problems I presented to economists in 2015. To the best of my understanding, they are still working on it. Are they? #Bookofwhy

1.19.20 3:46am - (Replying to @attilacsordas @sciencescanner and @freesci) Discrepancies between RCT and observational studies usually point to confounding (ie spurious correlation) in the latter. However, certain kind of discrepancies may indicate selection bias in the RCT. See Causality page 294. #Bookofwhy

1.19.20 1:03am - (Replying to @aceyuan) Admitting ignorance and overwhelmed by acronyms: What is BERT?

1.19.20 12:58am - (Replying to @causalinf) I am not an economist, nor social scientist. But if I were one, I would first learn how to spot a natural experiment in a well specified model of the phenomenon. Only then I'll try to apply the method to the mental model I have of the phenomenon, however familiar it is.#Bookofwhy

1.19.20 12:07am - (Replying to @HenMazzig) The antisemites are poor victims of hereditary disease. It's the Zionophobes that are doing it in cold blood, destroying every movement that gives them a nod of approval.

1.17.20 10:15pm - (Replying to @AdanZBecerra1 @Andrew___Baker and 3 others) Are you sure the PO guardians of "well definedness" (say @_MiguelHernan ) would agree with your suggestion (and my position in that if you can find it in the DAG it is automatically "well defined"? #Bookofwhy

1.17.20 9:52pm - (Replying to @Andrew___Baker @PHuenermund and 2 others) The notion of "potential outcome" in the PO framework requires "treatment assignment" w/o which you cannot define Y_x. That is why PO folks are debating whether non-manipulative x (say an earthquake) can have counterfactuals -- the RCT roots are still strong & stifling.#Bookofwhy

1.17.20 9:28pm - (Replying to @Andrew___Baker @PHuenermund and 2 others) Many authors are still using "as if randomized" instead of "unconfounded", and the whole enterprise called "target trial" is based on the idea that if you try to emulate some RCT you would do better than trying to think about the underlying cause-effect relations. #Bookofwhy

1.17.20 6:27pm - (Replying to @IzaTabaro) We are partly responsible for fermenting this aberration, by exposing it ONLY when it crosses into anti-Semitism. Aren't they ugly enough, and racist enough, just being anti-Zionists?

1.17.20 6:07pm - (Replying to @PHuenermund @laura_tastic) , Where is this page proof taken from?

1.17.20 6:05pm - (Replying to @christopherruhm @PHuenermund and @causalinf) RCT draws its legitimacy from a proof that, if conducted properly, it delivers a sought-after quantity called "causal effect". This means that causal language must exist BEFORE RCT. The #Bookofwhy makes this point on pages 146-150.

1.17.20 12:02pm - (Replying to @IntuitMachine and @jm_alexia) Examples?

1.17.20 11:55am - (Replying to @MaxALittle and @jm_alexia) Who said it is hard? Only those who have not tried it. Those who have, can’t understand the hesitation of the others. #Bookofwhy

1.17.20 4:58am - Readers who, like me, believe that symbols make history should understand why i felt compelled to retweet this photo. One of the few hopeful signs in the past week, after Omar’s congress refused to support iranian protestors, another history shaping symbol.

1.17.20 4:25am - (Replying to @DKedmey) “listening” is a good and accurate NL metaphor, and so is “source of variation”. But these are good for pre-scientific discourse, or for education. Are we in a similar pressing need for NL definition of “correlation”? #Bookofwhy

1.16.20 5:32pm - (Replying to @PhilosopherMD1) Keep up informed either way. Its been a long time since we heard from philosophers of statistics.

1.16.20 5:27pm - (Replying to @PHuenermund @Lester_Domes and 6 others) The question was not about path-specific effects (for which we have the Avin etal result) but about the Total Effect, ie, E[Stroke|do(Age)]. The miracle of DAGs (not widely acknowledged) is that expressions of type E(Y|do(x), z)] do not need summation over paths.#Bookofwhy

1.16.20 12:40am - Do we need a NL definition of a “cause”? We are blessed with a compact and meaningful mathematical object from which we can derive NL utterances on all three Rungs, association, intervention and counterfactuals! Why chase the unreachable? #Bookofwhy

1.15.20 11:30pm - (Replying to @Lester_Domes @RUBENSARO and 6 others) If we insist on doing path tracing Wright style then, yes, all paths from AGE to STROKE should be counted. Fortunately, we have DAGS to save us: The effect is simply E[STROKE|AGE] , no paths, no multiplication, no additions. Welcome to the magic world of DAGs-land #Bookofwhy

1.15.20 4:58am - (Replying to @PHuenermund and @Andrew___Baker) Indeed, what’s in “design”? Pressing hard, l’ve found that “design” is a word used by economists to cover decisions they prefer to do informally, away from piers scrutiny, thinking old-fashionally that they cannot be made formally. #Bookofwhy

1.15.20 3:47am - Highly recommended watch. And while watching, worth keeping in mind what #Bookofwhy says about Kahaneman’s dictum, and his “undoing project”.

1.14.20 2:48pm - (Replying to @JDHaltigan) If you can be specific, I'll be glad to demonstrate "no malice" in #Bookofwhy

1.13.20 11:48pm - (Replying to @Physical_Prep and @Alan_Couzens) I have great faith in AI, and I have read #Bookofwhy long time ago. The book, too, has great faith in AI.

1.13.20 11:26pm - (Replying to @GodfreySnorgyrs) Why dont you write one for anthropology? Epidemiology owes a lot to that pioneering article, and I am sure anthropology will be grateful to you, if you write one. #Bookofwhy

1.13.20 11:11pm - (Replying to @yudapearl @ShalitUri and @omaclaren) expect a "methodology" to deliver, not guesswork about the dark side of the moon, which we can generate absent a "methodology", still consistent with the standard mathematical definition of "solution. #Bookofwhy

1.13.20 11:06pm - (Replying to @ShalitUri and @omaclaren) That's exactly my point. Before we submit our causal questions to the mercy of DL we need to ask whether we possess sufficient knowledge to prevent contradictory solutions, or in prediction, for which curve-fitting has unique solution. This notion of "unique solution" is what we.

1.13.20 10:51pm - (Replying to @richard_landes @DKedmey and @EinatWilf) I am thinking of fake feminists like Linda Sarsour who is probably sincere in her admiration of Jews like Sen. Sanders, and fake liberals like Corbyn who hate Jews because they support Israel more than he hate them for rejecting Jesus.

1.13.20 9:49pm - (Replying to @omaclaren) A "solution" and a contradiction:
X,Y,Z = 1,2,2
X,Y,Z = 0,3,2
You just told me X=1, now you telling me X=0, and you want me to call it a "solution"? Are DL folks willing to accept both: "the drug is helpful" and "the drug is dangerous" as a legitimate "solution" of DL methodology

1.13.20 9:05pm - (Replying to @richard_landes @DKedmey and @EinatWilf) Your video is very convincing, though I've been exploring another theory, that antisemites like Richard Wagner and Mel Gibson can be forgiven for sucking it with their mother milk, not so Zionophobes, like BDS activists, for whom Israel elimination is cold and calculated agenda.

1.13.20 8:31pm - (Replying to @DKedmey and @EinatWilf) I was not familiar with @DavidDeutsch pattern, probably because I grew up in Israel, with a firm conviction that all past patterns are a thing of the past -- the New Jew will put an end to irrationality and inhumanity. It is 2020, and I am anxious to study the "pattern". Thanks.

1.13.20 8:21pm - (Replying to @DonBeham @RandomlyWalking and @raamana_) Yes, the diagram imposes constraints on the data, so data can refute or satisfy a diagram. This the 2nd fundamental law of causal inference which, fortunately, ML folks are beginning to internalize (unlike economists) .#Bookofwhy

1.13.20 1:13am - (Replying to @TheLeanAcademic @tdietterich and 2 others) The classical (and easiest) case is d-separation; it tells us if some members impose conditional independence constraint on the data. But speaking of Rung-3, things are more involved, and Section 8.5.2 of Causality (2009) has beautiful examples how data shape priors. #Bookofwhy

1.13.20 12:48am - (Replying to @tdietterich @BethCarey12 and @RaajeevVerma) The right way to think about it is, first, to ask if the "evidence" is capable of ruling out members of the model space. The hardest thing for statisticians to swallow is that data can be equally compatible with each member of the space. So, we must act as if we've NO evidence.

1.13.20 12:34am - (Replying to @nitalon and @ylecun) Chibuk L'Katin

1.13.20 12:31am - (Replying to @InnovatArt and @ylecun) This still begs the question: "mathematical models of WHAT?", and what do we do with it once we have it? #Bookofwhy

1.13.20 12:25am - (Replying to @SinghalApoorve) I sometimes feel I need to re-read my own fables to reinforce basic truth. That's what months in the trenches do to you. #Bookofwhy

1.12.20 10:25pm - (Replying to @ylecun) I am the last person to doubt your definition of DL, but I'll have very hard time explaining it to my students; they will rebuke me if I ever say "construct a model" without saying a model of WHAT. Is it a model of the dark side of the moon? Any data from that side? #Bookofwhy

1.12.20 10:05pm - (Replying to @roydanroy @suchisaria and 2 others) Last time (1990) I examined Dempster-Shafer theory, it couldn't even do Monty Hall correctly. See and Anything new to consider? #Bookofwhy

1.12.20 9:54pm - (Replying to @roydanroy @suchisaria and 2 others) We, in the trenches, need more gentle instructions on how you implement SEM/SCM as POMDP when you dont have the state transitions, not even actions, just passive observations. Have you tried it? Has anyone you personally know? #Bookofwhy

1.12.20 9:46pm - (Replying to @InfoSymmetries) Explanation: This is Amazon announcement of the paperback edition, which our publisher decided to release in August 2020. Good news, it will contain all the errata that readers labored to find. Thanks. #Bookofwhy

1.12.20 9:39pm - (Replying to @BethCarey12 @RaajeevVerma and @tdietterich) In my little corner of the universe the answer is: no amount of data can lift us from Rung-1 to Rung-2, or from Rung-2 to Rung-3, no matter how smart we are, and no matter if we call it DL or NN or AI or stochastic optimization. #Bookofwhy

1.12.20 4:44pm - (Replying to @roydanroy @suchisaria and 2 others)

1.12.20 4:44pm - (Replying to @roydanroy @suchisaria and 2 others) Please walk us by the hand through some simplified issues. Take us with POMDP to decide which drug is better, the one that is good for men and good for women, or the one that is good for people. Let's start with the model-free brethern. #Bookofwhy

1.12.20 4:33pm - (Replying to @suchisaria @RandomlyWalking and 2 others) How do we provide them with more talk space than twitter with occasional encouragement to "Please rebel!"

1.12.20 3:29pm - (Replying to @simonbchen) The implicit assumption you are talking about, has hardly been touched in DL circles, which is an interesting social phenomenon: A whole civilization discussing how to get a better fit to data, rather than "what can we do once we get a perfect fit" #Bookofwh

1.12.20 3:16pm - (Replying to @VeronicaEpi) I've not met anyone saying "causal inference is still confusing" after reading #Bookofwhy How can anything be confusing if it merely reflects human thought?

1.12.20 3:08pm - (Replying to @suchisaria @RandomlyWalking and @raamana_) I like your distinctions, for they are cast in terms of the questions we aim to answer, rather than the mathematics used to answer them. I hope they bring DL students to realize: "Gee! We never dealt with questions which we could not answer! The world IS 3-dimensional" @Bookofwhy

1.12.20 2:56pm - (Replying to @RandomlyWalking and @raamana_) "Snark" is not in my vocabulary. The high-school algebra example simply demonstrates that, if we do not have the data to answer certain questions, we should not just guess at random. It has nothing to do with inductive bias, which is a Rung-1 (function-fitting) concept #Bookofwhy

1.12.20 7:03am - (Replying to @davidcnorrismd @venkmurthy and 5 others) I would not call it "maligns". It is a very accurate description that most authors agree with, and only few dare share, which helps perpetuate the mystery. #Bookofwhy

1.12.20 1:50am - (1/ ) (Replying to @raamana_) JP TWEETS (updated 1.13.2020)

1.13.20 1:13am - (Replying to @TheLeanAcademic @tdietterich and 2 others) The classical (and easiest) case is d-separation; it tells us if some members impose conditional independence constraint on the data. But speaking of Rung-3, things are more involved, and Section 8.5.2 of Causality (2009) has beautiful examples how data shape priors. #Bookofwhy

1.13.20 12:48am - (Replying to @tdietterich @BethCarey12 and @RaajeevVerma) The right way to think about it is, first, to ask if the "evidence" is capable of ruling out members of the model space. The hardest thing for statisticians to swallow is that data can be equally compatible with each member of the space. So, we must act as if we've NO evidence.

1.13.20 12:34am - (Replying to @nitalon and @ylecun) Chibuk L'Katin

1.13.20 12:31am - (Replying to @InnovatArt and @ylecun) This still begs the question: "mathematical models of WHAT?", and what do we do with it once we have it? #Bookofwhy

1.13.20 12:25am - (Replying to @SinghalApoorve) I sometimes feel I need to re-read my own fables to reinforce basic truth. That's what months in the trenches do to you. #Bookofwhy

1.12.20 10:25pm - (Replying to @ylecun) I am the last person to doubt your definition of DL, but I'll have very hard time explaining it to my students; they will rebuke me if I ever say "construct a model" without saying a model of WHAT. Is it a model of the dark side of the moon? Any data from that side? #Bookofwhy

1.12.20 10:05pm - (Replying to @roydanroy @suchisaria and 2 others) Last time (1990) I examined Dempster-Shafer theory, it couldn't even do Monty Hall correctly. See and Anything new to consider? #Bookofwhy

1.12.20 9:54pm - (Replying to @roydanroy @suchisaria and 2 others) We, in the trenches, need more gentle instructions on how you implement SEM/SCM as POMDP when you dont have the state transitions, not even actions, just passive observations. Have you tried it? Has anyone you personally know? #Bookofwhy

1.12.20 9:46pm - (Replying to @InfoSymmetries) Explanation: This is Amazon announcement of the paperback edition, which our publisher decided to release in August 2020. Good news, it will contain all the errata that readers labored to find. Thanks. #Bookofwhy

1.12.20 9:39pm - (Replying to @BethCarey12 @RaajeevVerma and @tdietterich) In my little corner of the universe the answer is: no amount of data can lift us from Rung-1 to Rung-2, or from Rung-2 to Rung-3, no matter how smart we are, and no matter if we call it DL or NN or AI or stochastic optimization. #Bookofwhy

1.12.20 4:44pm - (Replying to @roydanroy @suchisaria and 2 others) Please walk us by the hand through some simplified issues. Take us with POMDP to decide which drug is better, the one that is good for men and good for women, or the one that is good for people. Let's start with the model-free brethern. #Bookofwhy

1.12.20 4:33pm - (Replying to @suchisaria @RandomlyWalking and 2 others) How do we provide them with more talk space than twitter with occasional encouragement to "Please rebel!"

1.12.20 3:29pm - (Replying to @simonbchen) The implicit assumption you are talking about, has hardly been touched in DL circles, which is an interesting social phenomenon: A whole civilization discussing how to get a better fit to data, rather than "what can we do once we get a perfect fit" #Bookofwh

1.12.20 3:16pm - (Replying to @VeronicaEpi) I've not met anyone saying "causal inference is still confusing" after reading #Bookofwhy How can anything be confusing if it merely reflects human thought?

1.12.20 3:08pm - (Replying to @suchisaria @RandomlyWalking and @raamana_) I like your distinctions, for they are cast in terms of the questions we aim to answer, rather than the mathematics used to answer them. I hope they bring DL students to realize: "Gee! We never dealt with questions which we could not answer! The world IS 3-dimensional" @Bookofwhy

1.12.20 2:56pm - (Replying to @RandomlyWalking and @raamana_) "Snark" is not in my vocabulary. The high-school algebra example simply demonstrates that, if we do not have the data to answer certain questions, we should not just guess at random. It has nothing to do with inductive bias, which is a Rung-1 (function-fitting) concept #Bookofwhy

1.12.20 7:03am - (Replying to @davidcnorrismd @venkmurthy and 5 others) I would not call it "maligns". It is a very accurate description that most authors agree with, and only few dare share, which helps perpetuate the mystery. #Bookofwhy

1.12.20 1:50am - (1/ ) (Replying to @raamana_) Hilarious! But the more we listen to DL talks, the more it sounds like "everything". I am still hoping one of the faithfuls would remember how the idea that we need n equations to solve for n unknowns changed how we did algebra in high-school. We first asked ourselves:
1.12.20 2:02am - (2/ ) (Replying to @yudapearl and @raamana_) "Do we have n equations?" If not: "Have we forgotten one?", if not: "Perhaps we can impose one?" and then, only then, we submitted our equations to the algebraic machinery that we could trust. The analogous process for CI is spelled out in #Bookofwhy

1.11.20 10:59pm - (1/ ) (Replying to @raamana_) What I did learn from the discussion of "What is DL" is that, though no definition is in sight, DL folks are utterly intoxicated by the seemingly unlimited potentials of their programs, techniques and vocabulary. I remember our second or third week of high-school algebra,
1.11.20 10:59pm - (2/ ) (Replying to @raamana_) utterly intoxicated by its unlimited solution-finding capabilities, the teacher told us that you can't solve two equations with three unknowns. Humbled and disappointed, we were nevertheless excited by this revelation; it saved us hours upon hours of chasing after solutions
1.11.20 10:59pm - (3/3) (Replying to @raamana_) that do not exist. But these were high-school days. #Bookofwhy

1.11.20 2:34am - The intense discussion on "What is DL" has evidently not converged on a consensus among DL practitioners. Nor has it convinced me to reconsider any of the impediments reported in or the ML/CI division of labor recommended. Ready to be educated. #Bookofwhy

1.11.20 1:27am - (Replying to @DKedmey and @tangled_zans) Reply to your tweet removed: @EinatWilf is my hero. I wish Israeli leaders will be as clear as she is on the kind of peace Israel society strives to achieve and what kind of "peace talks" are a waist of time, because they do not entail "equally legitimate and equally indigenous".

1.10.20 4:58am - (Replying to @tyrell_turing) Not really. The other reason for the Q "what is DL" was that good ppl identified DL with everything that AI can ever hope to achieve, which naturally led to the question: "What is NOT DL?" #Bookofwhy

1.10.20 4:54am - (Replying to @witbrock @zaffama and @ylecun) The problem is not what classes of computable functions can be effectively learned by a DL system? But rather, whether the answer you expect your algorithm to deliver is a computable function of your data. #Bookofwhy

1.10.20 4:41am - I am re-tweeting with the hope of reaching one or two econ. students and enticing them to ask some hard questions. #Bookofwhy

1.10.20 4:02am - (Replying to @csilviavr) Your paper makes the point very clear. I was first hesitant seeing the term Causal Bayesian Network CBN, which is defined on interventional distributions (Causality p. 24). But you corrected for it through path-specific, ie. counterfactual notation. #Bookofwhy

1.10.20 1:50am - (Replying to @javisamo) I couldn't access Lily Hu's paper. Can you summarize the limitations of counterfactual reasoning? A punchy example, perhaps? #Bookofwhy

1.10.20 1:41am - (Replying to @PHuenermund @causalinf and @Dunkin_Donuts_2) I think viewing these anecdotes as "personality cult culture" misses the real crisis of econ -- the only data-driven discipline the majority of whose students are still deprived of 21st-century tools such as graphical models, and are not rebelling, not even striking. #Bookofwhy

1.10.20 1:01am - I'm reading this extensive review of Fairness in Machine Learning and am happy to see that it acknowledges (though not strongly enough imo) that "fairness" is a causal notion, and that DAGs plus counterfactuals are needed to make sense of it. #Bookofwhy

1.9.20 7:28pm - As a Special Anointer of saints, I hereby commit the next sainthood to the Editor of the first econometric journal to invite a survey paper on "graphical models in econometric". Spread the word. Candidates may include previous saints. #Bookofwhy

1.9.20 7:03pm - The discussion on "What is DL?" reminds me of identical discussion (among Bayesians) on "Who is a Bayesian?" ending up with a dead-end: "Do it! Dont ask!". Good for justifying what you have been doing, not so good for science, as I confess here #Bookofwhy

1.9.20 3:27am - (Replying to @AndrewLBeam and @ylecun) The division of labor between CI and DL is mathematically defined by every CI exercise. See Each probabilistic expression that appears in the derived CI estimand calls for a DL exercise. Data Fusion entails a similar philosophy

1.9.20 2:33am - (Replying to @_mb46_ and @ylecun) Your simple definition of DL "using ANNs to do stuff" is too broad for me. Is a computer chip an ANN? It is certainly a "network" of wires and gates. What does it take to qualify as "neural" once we allow "artificial". #Bookofwhy

1.9.20 2:27am - (Replying to @tangled_zans) What for you may seem "Politics" is for me "moral imperative". My take: I am a student of Middle East history and a long-time peace activist towards "two equally legitimate states, for two equally indigenous peoples".

1.9.20 2:06am - (Replying to @tangled_zans) The silencing takes the form of protecting this woman from criticism, for fear of being called "Islamophobe". I feel I have earned credibility as a fighter against Islamophobia to articulate what others see but are afraid to condemn.

1.9.20 1:46am - (Replying to @_mb46_ and @ylecun) Surely "considerable differences" exist. But are they judged to be fundamental, or merely temporary limitations of current DL techniques? If the former, then we need indeed to better understand what the the definition is of DL. #Bookofwhy

1.9.20 1:31am - (Replying to @tangled_zans) Random strangers have immense influence on me, but I can't understand what you see wrong in me alerting readers to the danger that this woman poses, through her hate speeches. It is my responsibility to share my analysis with readers who, silenced by PC, do not dare speak out.

1.9.20 12:58am - (Replying to @zaffama and @ylecun) I am not sure DL advocates would be satisfied with your modest definition. They would probably wish to extend it to include "every success story in the next century". But I should not speak for them. Curious: "What is NOT DL?" #Bookofwhy

1.9.20 12:47am - (Replying to @_mb46_ and @ylecun) I believe it was #tdietterich who defined DL as "the science of intelligence systems". And the answer to "Are ANN enough" is trivial: "Of course! Because we know that Organic NN are enough." I am waiting for a less ambitious clarification of what DL is.

1.8.20 11:56pm - (Replying to @tangled_zans) The "bigger picture": To assure descent readers that I share their disgust with the racism of certain Zionophobes, that the educational origin of this aberration is explained, and that it is OK to speak against it. No INSULT here, since Zionophobes are proud of their disorder.

1.8.20 11:32pm - I wish I could comment on the question "Is DL enough?" But I can't, because we do not know what DL is. Some define it as an aspiration to emulate intelligence. Others, eg @ylecun , define it as a specific NN implementation. We are waiting for clarification. #Bookofwhy

1.8.20 11:10pm - Asking a Zionophobe for logical consistency is like asking a snake to walk on two. According to Mogadishu-born Ayan Hirsi Ali, the only semblance of consistency in her (& Omar's) school curriculum in Mogadishu was the consistent blaming of Jews for bad grades and water shortage.

1.8.20 10:46pm - (Replying to @sherrirose and @RhubbBstat) Muchos Congratulationes. #Bookofwhy

1.8.20 10:40pm - OOPPS. I've just answered it here:

1.8.20 10:24pm - It's a bit more that "offsetting paths". Consider Y=f(X,U)= 1 iff X=U, and let X and U be two fair coins. No offsetting paths in X-->Y<--U. Yet X||Y and Y||U. Statistical magic. #Bookofwhy

1.8.20 12:28am - Just bumped into this wonderful Ted Talk of @harari_yuval from whom I've learned about the origins of counterfactuals and imagination, and where I've found a connection between the structural theory of counterfactuals and futuristic robotics. #Bookofwhy

1.7.20 11:00pm - (Replying to @EpiEllie @ruschenpohler and @AndrewJDBell) I would dare say that to ensure generalizability under changing location or time we need to postulate a data-generating-model of some kind. I don't think we can get by with frequency information alone. #Bookofwhy

1.7.20 7:26am - (Replying to @carlgieringer) And how is opposing "killing frogs" a victory for commonsense? The charge varies from year to year, and is becoming more "benign" and "universal". Who does not love frogs? BDS scores it victory from the music "Israel is on trial" and the megaphone, not the libretto.

1.7.20 6:55am - Am. Hist. Assoc. scored victory for commonsense, but the BDS circus will continue. Next year the resolution will call on Israel to stop killing frogs on Tuesdays. Do they? Questions later! First listen to the sweet music of "Israel killing" - BDS symphony #99 in E-flat major.

1.7.20 4:44am - Clarification. MR, IV, and conditional-IV methods belong to bullet 5 (More elaborate policy evaluation.) These are available in DAGGITY, if I recall correctly. It is important to emphasize that DAGs should be used to bestow defensibility upon these methods. #Bookofwhy

1.7.20 12:26am - (1/ ) Comments on "Use of DAGs". Commending you on an important survey. I was worried that DAGs will be introduced too narrowly, but the route of "data generating mechanism" avoids misconception that gets Epi people into trouble when talking "treatment-assignment" jargon.
1.7.20 12:26am - (2/ ) Readers noticed that the papers selection criterion was not clearly defined. E.g., did you include papers where DAGs were used merely to communicate context? or was the causal component necessary?
I would suggest the following taxonomy of DAG usages:
1.7.20 12:26am - (3/ )
1. Communicate contextual information
2. Communicate and defend causal assumptions
3. Selection of covariates for adjustment
4. identify testable implications.
5. Elaborate policy evaluatioin
6. Transportability and selection bias.
7. Missing data
1.7.20 12:26am - (4/4) I was surprised to find no papers in bullets 4-7 Dont people ever test their models? Is Epi still in the adjustment era ? Much work for educators.
Overall, a useful panoramic view of a field gone scientific. #Bookofwhy

1.6.20 6:29pm - (1/ ) I can't wait to see the faces of our two smearing experts when they find that the person whom they labeled "conservative" "religious" "blinded by hatred" etc. etc. is in fact a (1) devout liberal, (2) a voting registered Democrat, (3) a devout atheist, (4)award winner for
1.6.20 6:29pm - (2/ ) anti-hate programs.(5) active writer on Middle East history. What these smear peddlers could not stomach was my criticism of the untouchable saintly Rashida Tlaib. I therefore repeat what I know about her. She has been spewing hate, deceit and outright racism since being
1.6.20 6:29pm - (3/ ) elected. For readers who respect my sincerity, domain knowledge and ability to read beneath the surface, I have also commented on her sweet anti-war statement after soleimani death, and noted that she cannot afford to tell her audience what kind of mass-murderer Suleimani was
1.6.20 6:29pm - (4/4) because a large segment of her constituency sympathizes with Suleimani's agenda of weakening the US and wiping out Israel -- two holy aims that justify all means.

1.6.20 7:25am - (Replying to @timminglab) Moreover, the more you get to know colleagues at Ivy league universities, the more you appreciate the tribal games you did not have to play being yourself.

1.6.20 7:15am - Or, at the very least, @SenSanders , stand up like a Mench and answer questions about what chunks of Sarsour's ideology you DON'T share. I am willing to do the interview, since none of the reporters behind the microphones is willing to touch on this issue.

1.6.20 5:45am - (Replying to @KordingLab) Thanks for the update. Interesting interpretation. The sequence of presenting a method first and asking for veracity second is what triggered my simplistic interpretation.

1.6.20 4:51am - (Replying to @KordingLab) Is there a new "Athey tutorial" in circulation? The last I have seen said essentially: "Whatever economists do is automatically causal" (my humble interpretation). Is there a re-freshener? #Bookofwhy

1.6.20 1:50am - New sounds: "naive, ignorant, unschooled, pasty". What makes some academics so allergic to new information from a new perspective? Put differently, what makes Rashida Tlaib more knowledgeable, educated, revered or morally compelling than humble me? Ready for the first test? is automatically causal" (my humble interpretation). Is there a re-freshener? #Bookofwhy

1.5.20 9:13pm - (Replying to @KordingLab) Agree! Most of the discussed issues are rung-1. Still, they all eventually lead to questions of policy and/or explanation, and the paper gives the impression that DL is well prepared to tackle these issues too. Your honest opinion: Is it? #Bookofwhy

1.5.20 8:32pm - (1/ ) Fair question! My aim was not to "interpret" Rashida's statement, but to express my reaction to her rhetoric in light of what she has been doing in the past. Specifically, in light of the hate, deceit and outright racism she has been spewing since being elected.
1.5.20 8:32pm - (2/ ) 3/3 which do not tell us who Suleimani was and why Rashida can't speak about it. She can't because she would lose her support base, a big chunk of which sympathizes with Suleimani's agenda of weakening America and wiping out Israel. Sad. But rhetoric won't change reality.
1.5.20 8:32pm - (3/3) which do not tell us who Suleimani was and why Rashida can't speak about it. She can't because she would lose her support base, a big chunk of which sympathizes with Suleimani's agenda of weakening America and wiping out Israel. Sad. But rhetoric won't change reality.

1.5.20 8:01pm - (Replying to @bariweiss) Kol Hakavod

1.5.20 6:21pm - (Replying to @fayyazhere)
1. The limitations of rung-3 reasoning are that it relies on information about the functions behind the arrows. Lacking it, gives us bounds, not point estimates.
2. It is useful to assume that my ML colleagues can extract the optimal predictive information available.#Bookofwhy

1.5.20 8:29pm - Beautiful 1-pager. Now I see my math teacher winking to me: Should I be pedantic and spoil the beauty? No. I'll let it go. Just a tiny spoiler: All "statistically dependents" in bullet 5 should read "likely statistically dependents". Why? Some other time. #Bookofwhy

1.5.20 4:52am - (Replying to @ktmud) What you are saying is: It's TOUGH to see someone who has spent lots of time studying the middle east conflict, and worked day and night spreading love and understanding for whole mankind express ideas that differ than MINE. So I'll dismiss him as "blinded by born identity".EASY!

1.5.20 3:48am - This paper encourages ML folks to tackle problems of Climate Change, yet neglects to mention the unique causal and counterfactual tools needed for the challenge. In particular the evaluation of sufficient and necessary causes, eg:

1.5.20 3:27am - (Replying to @TheDavidSJ) If I were the murderer of Rina Shernav (17) I'd say: "Job worth repeating! Even when we kill a girl in the heart of Tel Aviv, we now have someone in the US Congress to blame "Israeli occupation" for the crime. It's a new dawn!" [Note how she did not say "equal rights to Israelis"

1.5.20 1:02am - (Replying to @TheDavidSJ) If she truly cares for "loss of innocent lives", including those hundreds of thousands of innocent lives destroyed by Suleimani, Congresswoman Rashida Tlaib has had ample opportunities to weave it into her speeches since his departure. She has not done so & we know why. She cant!

1.4.20 11:41pm - Congresswoman Rashida Tlaib "cannot stay silent". She is deeply shaken by the unexpected loss of an ideological idol, mass murderer Qasem Suleimani. Her sudden interest in "innocent home and across the globe" now humors her American voters in Michigan.

1.4.20 8:55pm - This is one of the most futuristic, yet sensible futuristic conversations I have heard: I'm only surprised they treat consciousness as a mysterious inaccessible concept, laden with pain and suffering. #Bookofwhy

1.4.20 4:30pm - Glad we agree. I believe, however, that it goes deeper than "origin". The refusal of the PO community (including Econs and Harvard) to adopt the intervention/counterfactual distinction represents refusal to accept structural models as a basis of scientific thinking.#Bookofwhy

1.4.20 3:29pm - (Replying to @jrgptrs @3ieNews and 13 others) If I were able to contribute to your blog, I would ask: "Why quote Deaton who laments careless handling of #externalvalidity instead of promoting methods that properly handle the problem?" For example: or #Bookofwhy

1.4.20 2:56pm - I highly recommend Primer:, which has a wealth of beautiful examples. Plus, the recent survey by Paul and Elias, geared to economists: #Bookofwhy

1.4.20 6:02am - (Replying to @attilacsordas) Let's discuss by email.

1.4.20 5:54am - (1/ ) What is "counterfactual prediction"? The quantity estimated in: turns out not a Rung-3, but a Rung-2 interventional expression. To avoid confusion & false expectations, I recommend that ML folks adopt the hierarchy in #Bookofwhy. We need to distinguish
1.4.20 5:54am - (2/2) "effects of causes" from "causes of effects" (see and else ML students will come to believe that DL has found a magic way of inferring counterfactuals from data, thus risking an inevitable disappointment.

1.4.20 4:02am - I welcome this new review of #Bookofwhy in American Mathematical Monthly It is written by a hard core statistician unafraid to confess the nature of statistics discomfort with causation.

1.4.20 3:34am - Rabbi Sacks is always insightful and to the point. I would only add that the irrational animosity towards Israel originates from the same swamp: failure to accept a cohesive thriving society not based on class struggle. For elaboration, see

1.4.20 12:12am - (Replying to @attilacsordas) What did you find incomplete in the formal definitions of "sufficient cause" and "necessary cause" as given here: or here: I find them quite satisfactory. Plato.stanford is outdated. #Bookofwhy

1.3.20 6:01am - (Replying to @jonathanborows2) What is "clickbait marketing"? I love the sound of it.

1.3.20 5:58am - (Replying to @mendel_random) I noted, with sadness and humor, that Johnson will have hard time finding qualified applicants for those jobs. (ps. where did you dream up "faithfulness"?) #Bookofwhy

1.3.20 12:50am - I see a much more immediate risk, that @downingstreet will barely find any applicants. I can hardly name a dozen who would qualify, can you? Not because it is hard to explain, but b/c DS folks are too busy doing what they are so good in doing. #Bookofwhy

1.3.20 12:39am - (Replying to @eliasbareinboim) I am trying to interpret the expression "using another DL model". I thought DL prides itself on being "model free". Perhaps you mean "using the estimation prescribed by the CI model" ? #Bookofwhy

1.2.20 11:19pm - (Replying to @UlrichJunker and @sahilsingla47) Thanks

1.2.20 11:18pm - The wrong link, to the "counterfactual prediction" paper was not meant to be an endorsement. I am still not sure the authors are doing counterfactuals in the Rung-3 sense. There is a tendency among OP folks to call interventions counterfactuals; too bad ML folks follow #Bookofwhy

1.2.20 10:31pm - Correcting link to the video: Though it will never correct the mentality of the department that invited this pervert to speak in an "Institute of higher learning". We ARE responsible.

1.2.20 10:21pm - A must watch VIDEO, from San-Diego State University: And what are we doing, as refined academicians, to prevent this poison from igniting our campuses? Have you called out the Chair of that idiotic department? History records both action and apathy.

1.2.20 3:01am - (1/ ) Reflecting back on our DL/CI discussion with @tdietterich, @ylecun, @GaryMarcus etal, we should note that CI folks are invoking the same scientific argument that has brought DL its fame and success. Back propagation and subsequently gradient-based optimization were inspired by
1.2.20 3:01am - (2/ ) machine known in the 1970's. As we go to higher levels of cognition, eg causal inference, it would make sense to ask (and imitate), how the mind preforms this mode of inference, that is, what cognitive
1.2.20 3:01am - (3/ ) templates enable it to manage association, interventions and counterfactuals simultanewously, compactly and swiftly. I know of only one such template: Structural Causal Model. Thus, instead of hoping that CI templates would pop up spontaneously from DL, let us ask first
1.2.20 3:01am - (4/4) what should they look like? How should we represent them, and what we would be able to do once we get them. #Bookofwhy

1.2.20 12:36am - Glad to see that subtitles were added to my podcast with Lex Fridman I can now understand my Hebrew accent. #Bookofwhy

1.1.20 3:53pm - (Replying to @AdanZBecerra1 @Lester_Domes and @nature) I was called "disruptive" in 1st grade of school, sent home, and was back only three weeks later, after seeing an oppressive psychiatrist who vowed that I was ready. Evidently, I wasn't. #Bookofwhy

1.1.20 2:49am - (Replying to @maximananyev) How many Masters does it take to correct what some Metrics refuse to do? #Bookofwhy

12.31.19 11:29pm - (Replying to @BariFaisal) I assume you refer to #Bookofwhy? My thoughts on empirical research are summarized here:, and on RCT, here: Warning: Graph-shunning economists are depriving their students of ever combining experimental and observational studies.

12.31.19 7:49pm - (Replying to @Isalomaki) Yes, "Be" means "We", not "Me, Me, Me".

12.31.19 4:56am - (Replying to @HenMazzig)` Isn't it what congress-woman Tlaib has been howling all along?

12.31.19 4:45am - (1/ ) My New-Year greeting this year is again in a form of a poem/song, through which I wish all readers:
"Make it be, Make it be, Pleading: "Make it be", All for which we pray Make it be."
In 1973, Nomi Shemer's, set out to translate the Beatle's song "Let it be" into Hebrew. But
12.31.19 4:45am - (2/ ) then the Yom Kippur war broke out, and what came out her pen was a prayer that swept the country with immense yearnings for peace. It still resonates today in every street and on every occasion:
The summer ends, the journey's over, Let them please at last come home...
12.31.19 4:45am - (3/3) All for which we pray: "Make it be".
You can hear the entire song about 7:10 minutes into this video.
Happy New Year! and "Make it be!"

12.30.19 1:55pm - (Replying to @AwokeKnowing @ylecun and @LeCun) Sure. The math of woodwork is not too complex, so I dont see why it cannot be expressed as output of the math of metallurgy, given the unique metallurgical properties of wood. #Bookofwhy

12.30.19 1:39pm - (Replying to @vatsal_maru @ylecun and @lexfridman) Solving ONE toy problem in causal inference is worth 1000 conversations, interviews, debates & what-have-you about what others do. Take, say, Simpson's paradox and analyze it through the math of "gradient-based optimization", but stick to the toy! #Bookofwhy.

12.30.19 1:06pm - (Replying to @ylecun) Please listen carefully to what the carpenter is saying:
* Carpenter: Sure, you can do the nails, I'll do the woodwork, together we can build a house.
He should also add that the math of metallurgy is not very helpful in carpentry, a new math is needed. #bookofwhy

12.30.19 6:38am - (Replying to @ElliotMalin) Wishing you and your lucky bride many years of happiness. And may your commitment to Israel be a constant inspiration to you and to your children.

12.30.19 4:48am - (Replying to @PolandCherieM and @oralassila) In #Bookofwhy I was just summarizing half a century of zero-progress in external-validity research, stuck for lack of language. DL, in contrast, has shown immense progress, but lack of language is still a fundamental impediment, and is beginning to show its face.

12.30.19 4:06am - A strange thought on what DL can and cannot do.
` If "X can do Y" whenever we need X to do Y, then transistor experts "can do" AI, and nail-makers "can do" furniture.

12.30.19 3:33am - (Replying to @ArnaudMegret) And naive me assumed that every ML researcher read #Bookofwhy, with just a few hard cases who remained unconvinced. No wonder @ylecun 's followers keep asking: "Can't we just "extend" DL a bit to cover causality?" I don't believe any reader of #Bookofwhy would ask such a question.

12.30.19 3:13am - (Replying to @ArnaudMegret and @LeCun) This paper summarizes our latest findings on how much information one can get out of missing data, both MAR and non-MAR As you point out, the question is causal, and it is hard to imagine how any statistical analysis can obtain these results. #Bookofwhy

12.30.19 2:05am - (Replying to @AngeloDalli @ylecun and @LeCun)
* Metallurgist: My nails just did a table, they can do a house?"
* Carpenter: Sure, you can do the nails, I'll do the woodwork, together we can build a house.
* Metallurgist: No! The nails should do it. They can perhaps be extended to handle the woodwork.

12.29.19 4:13pm - (Replying to @ylecun and @LeCun) If we have interventions, we do not need CI, we just fit the observed data (eg by DL) and we're done. CI is needed to answer questions that cannot be answered directly from data, eg "Find the effect of X, when I only have interventions on Z". A new math is needed here. #Bookofwhy

12.29.19 6:48am - (Replying to @ngutten @jsusskin and @LeCun) Sure! Everyone (except perhaps Yann) understands that the goal of DL is to fit a function to 10K+vars, and that CI has a totally different goal (requiring totally different math): to answer causal questions about reality ASSUMING that DL folks succeed in their fitting efforts.

12.29.19 6:36am - Today's madness could not be more vividly portrayed than through this record of @KenRoth , the Orwellian watchman of (the world's) Human Rights.

12.29.19 6:27am - (Replying to @SussexFriends) Is this moral pervert still talking? After all that he did to embolden hate against Israel? May my last words to him be: Nes Gaddol Haya Po (in UK).

12.29.19 5:33am - (1/ ) (Replying to @ngutten @jsusskin and @LeCun) We should discourage researchers from thinking that 'to do CI with regression requires an extra step'. This would lead them into believing that doing CI with regression is not an oxymoron, that the extra step is trivial and would pop up spontaneously if only they study more
12.29.19 5:48am - (2/ ) (Replying to @yudapearl @ngutten and 2 others) and more regression. It won't. On the contrary. Today, the best experts on regression are the least likely to understand what is needed to do CI. I foresee the same fate awaiting students of DL. My advice: start afresh, with CI, and resort to DL only if needed. #Bookofwhy

12.29.19 3:59am - I've hoped to celebrate the 8th night of Chanukka on a joyful note. In the wake of the Monsey stabbing, I must retreat to what I wrote in 2009 (in the WSJ): The normalization of evil begins with the normalization of hate, and our leaders are not watchful:

12.29.19 2:32am - (Replying to @Imamofpeace) That's essentially what Linda Sarsour told Bernie Sander. He believed her, assuming she represents the next wave of votes. The consequences are in the streets of NYC.

12.28.19 5:38pm - (Replying to @KordingLab @GaryMarcus and 7 others) As you can see here:, I love everything about the economists tradition, except their latest betrayal of that tradition, and their cultish avoidance of tools that operationalize that tradition. See eg #Bookofwhy

12.28.19 5:14pm - (Replying to @KordingLab @GaryMarcus and 7 others) I am sincerely trying to understand what it IS, or what it means to YOU, and you are slapping me with "denying the validity". Unfair! #Bookofwhy

12.28.19 5:08pm - (Replying to @KordingLab @GaryMarcus and 7 others) For the sake of amazing clarity I have been trying to avoid the term "different flavor", because it has been misused to create the impression that there are differences in substance where there are none. @Bookofwhy

12.28.19 4:06pm - (Replying to @KordingLab @GaryMarcus and 7 others) The arguments are perfectly solid. I am just not clear what econ. practitioners call "OV bias equation", and how they use the equation to reach new heights. #Bookofwhy

12.28.19 3:54pm - (Replying to @edwardsjk @EpidByDesign and 11 others) I take it that the idea is to "combine" observational and experimental studies, rather than exclude one in favor of the other. Are the results of "combining" different than those of "fusing"? As in eg ?#Bookofwhy

12.28.19 3:39pm - (Replying to @KordingLab @GaryMarcus and 7 others) A sincere question: What is the "omitted variable bias equation"? How does one realistically "apply it to real world reasoning?" I honestly havn't seen it applied, but I might have missed the equation title. #Bookofwhy

12.28.19 3:24pm - (Replying to @KordingLab @GaryMarcus and 7 others) I would love to join you in admiration of "econ style CI" if I only knew what it is, what it is "complementary" to, and how it "beautifully complements" pieces that are missing elsewhere. Curious. #Bookofwhy

12.28.19 6:14am - (Replying to @oralassila) It is not an implicit prediction as much as it is an explicit concern. If we continue to invest most of our resources in the science of nuts and bolts, engrossed and intoxicated by its successes, our ability to build engines will be delayed. #Bookofwhy

12.28.19 4:16am - (Replying to @PolandCherieM) Thanks for posting this illuminating paper on the history of regression. Biased by the lens of causation we emphasized the odd fact that Galton sought a causal explanation to Darwin's theory and ended up abandoning causation to the mercy of correlation. What a miss! #Bookofwhy

12.28.19 3:55am - In addition to introducing new methods of generalization, this paper also identifies and solves a new and important problem of learning causal models from both observations and interventions. Highly recommended. #Bookofwhy

12.28.19 3:14am - To all who celebrate the 6th night of Hanukkah with my friends and colleagues in Israel, I am sharing a song that we used to sing since Kindergarten:
A miracle did not happen to us, We have not found a vessel of oil, We carved the rock till we bled, And there was LIGHT!

12.28.19 3:03am - Variation on a painful, left-leaning theme: The only thing less politically convenient for leftists to recognize than left-wing antisemitism is left-wing Zionophobia and Palestinian elimination-ism.

12.28.19 2:47am - I am retweeting this reply, for it crystallizes my position in the latest conversation on the relationships between DL (deep learning) and CI (causal inference) with @tdietterich , @ylecun , @GaryMarcus , @rodneyabrooks and significant others. #Bookofwhy

12.28.19 2:06am - (Replying to @moultano @bradpwyble and @GaryMarcus) It is extremely difficult to make precision nuts and bolts, and it is by no means "trivial". But the science of making nuts and bolts is not the science of making engines, and a community engrossed in the former will be limping on the latter. #Bookofwhy

12.27.19 8:08am - (Replying to @kudkudakpl @GaryMarcus and @moultano) A different approach is needed. Now it is my turn to ask "out of curiosity": Why do you keep saying "doing DL"? Is it because you enjoy approximating functions? Or because you want to ride the DL hype? Seriously, I am truly curious. #Bookofwhy

12.27.19 8:01am - (Replying to @databoydg and @tdietterich) Same can be said about linear algebra. Current "approaches" are insufficient but, eventually, everything AI will achieve in the future will use linear algebra to some extent, So, linear algebra will spontaneously spawn intelligence. Tough? Don't bet against DL. #Bookofwhy

12.27.19 7:39am - (Replying to @kudkudakpl @GaryMarcus and @moultano) Good question. What I mean by "mathematical impossibility" is w/o causal (extra-data) assumptions of some kind. The paper u cite makes such causal assumptions and the study of what assumptions are needed has been the central focus of CI research in the past 3 decades. #Bookofwhy

12.27.19 7:29am - (Replying to @databoydg and @tdietterich) Why assume anyone pretends they can't exist together? I just wrote that I have always assumed DL is almighty and can perfectly approximate any function and, so, it can exist together with CI, approximating galore all functions that CI says should be approximated. #Bookofwhy

12.27.19 7:05am - (Replying to @kishkushkay)` When was that, Kay?

12.27.19 6:55am - (Replying to @memosisland @GaryMarcus and 2 others) @tdietterich wrote that DL's aim was to develop "science and engineering of intelligent systems." I removed "engineering" from his quote, because I want to focus on the "science", where CI has made some modest contributions. #Bookofwhy

12.27.19 6:41am - (Replying to @moultano and @GaryMarcus) My expectations about DL never ended up wrong, because I have always assumed DL is almighty, and can approximate perfectly every function it is given, regardless how complex. But for DL to spontaneously produce CI, it's not implausible, its a mathematical impossibility #Bookofwhy

12.27.19 5:26am - (1/ ) There is more to @GaryMarcus argument than taking credit. If the goal of DL was indeed as ambitious as you describe it ("science of intelligent systems") than it failed miserably by not realizing the barriers to rising above Rung-1 (ie., function fitting). If, however, the
12.27.19 5:26am - (2/ ) goal was more modest, say to perfect the art of function fitting, then there is no room for claims that CI will eventually become part of the DL umbrella. The distinction between "CI will be under the DL umbrella" vs. "DL is part of the CI umbrella" is not about credit. It's
12.27.19 5:26am - (3/ ) about the agenda of AI research in the next decade. Saying "CI is merely an extension of DL methodology" amounts to telling educators, funders and the public: "Let us continue doing what we have been doing all along, CI will emerge organically." It won't! And the result:
12.27.19 5:26am - (4/ ) AI progress will be slowed, and the disappointment will be painful. In contrast, saying: "DL is a component of CI" amounts to telling funders and educators: "We must educate a new breed of DL researchers, capable of tackling new challenges. Things as usual just won't work!"
12.27.19 5:26am - (5/5) I believe the second alternative is more constructive. "Things as usual" leads to "Things as usual" unless there is a paradigm change & a community awakening to the limitations of "Things as usual". Luckily, we can formalize those limitations. Let's overcome them! #Bookofwhy

12.27.19 3:46am - Happy new year! To all readers who are hoping for one, me included. Glad Israel remains the world's beacon of moral clarity, something the world has forgotten exists, and finds it hard to stomach. Happy new year!!

12.27.19 3:15am - Thanks for posting this interesting paper. I've never imagined that causal graphs will find applications in such esoteric fields as Visual Dialogue. But I am known to suffer from occasional lapse of imagination. #Bookofwhy

12.27.19 2:38am - (1/ ) (Replying to @tdietterich) Disparity in levels is one source of the confusion. You congratulate DL engineers for their achievement in approximating such complex functions, and I say: it is ONLY "curve fitting". We are both right! In the past 3 decades I took function approximation to be a done deal.
12.27.19 2:45am - (2/ ) (Replying to @yudapearl and @tdietterich) How? Every time my math ended up with a probability expression, say E(Y|x,z), I labeled it "solved", namely, DL folks will find a way of approximating this function from data. Done deal! Let's look at the more challenging task of reducing our causal questions to probabilities!
12.27.19 2:56am - (3/ ) (Replying to @yudapearl and @tdietterich) Today, that we know almost everything about what can or cannot be reduced (meaning what DL can or cannot do) we find DL folks awakening to, and dismissing the main challenge with: "We can fit everything!" It's true, they only need someone to tell them what to fit.#Bookofwhy

12.27.19 2:19am - (Replying to @eyad_nawar) If we have 1&3 you do not need controlled experiment. If we don't, we can't move beyond "curve fitting". This is something DL folks are beginning to realize, in their own unique pace. #Bookofwhy

12.27.19 2:05am - (Replying to @steventberry @PHuenermund and @ChrisAdamsEcon) The confusion about "what to control for" is symptomatic of a fundamental methodological neglect econometric is currently recovering from. I compiled a few more symptoms in: For survey and remedy, see: #Bookofwhy, @ben_golub

12.26.19 6:54am - Arnold Roth is both a friend and a grieving comrade. His daughter Malki and my son Daniel were victims of the same wave of hate that swept our planet in 2001-2002. Please listen to his plea. Judea Pearl

12.26.19 6:17am - (Replying to @guillefix @maier_ak and 5 others) What is not available to DL training is not "data" but "information" on how to get the appropriate data. For example, that "age" is a confounder, hence, estimate the age-specific effects, not aggregate effects. See Simpson's paradox, #Bookofwhy p.211, or

12.26.19 6:05am - (Replying to @bioinfochat @tdietterich and 3 others) Agree, "regression is here to stay" does not imply "DL is here to stay", I should have said "Some form of function approximation is here to stay", which is tautological and elevates DL to a special status it rightly deserves. #Bookofwhy

12.26.19 5:44am - The Middle East Studies Association (MESA) sent a letter to President Trump in which they decide how Jews should define themselves. The verdict: Jews are not a people of common national origin. Their logic betrays their wisdom --a true academic gem.
12.26.19 5:44am - (2/ ) The logic goes: Since not ALL Jews share a common national origin, therefore, NO Jew, including the vast majority who DO, should be treated as such. May the 5th candle of Hanuka remind those academic "experts" how Jews have been defining themselves for the past 2200 years.

12.26.19 3:06am - (Replying to @LuisMateusRocha and @LeCun) Thanks, but this is over my head. Think about a naive AI fellow who wants to implement one of those "non-inductive" machine. Where should she start?

12.26.19 2:55am - Who said John Searle will never pass the Chinese Room Turing Test? See #Bookofwhy page 38-39.

12.26.19 2:35am - (Replying to @LuisMateusRocha and @LeCun) I'm not familiar with the technical details of this literature. Can you Tweet what information "anticipatory system" has that #induction machines do not, where it gets it, and how it represents it. Thanks #Bookofwhy.

12.26.19 2:02am - (1/ ) (Replying to @tdietterich @Plinz and @GaryMarcus) The point is that the term "functional approximation" is more informative to the engineer than an account of flexibility. It tell the engineer to stop asking how good the approximation and start asking whether she approximated the right function, and how to get the correct
12.26.19 2:12am - (2/ ) (Replying to @yudapearl @tdietterich and 2 others) input-output pairs. It also means that @ylecun statement "DL is here to stay" is tautological, akin to "programming is here to stay". What counts is whether we write the right program. And when we do, we don't label it "Turing Machine" #Bookofwhy

12.25.19 4:39pm - (Replying to @ylecun @BethCarey12 and 2 others) Just for the music: Replace DL with "regression", and see how things still stick, and the music is still flowing soundly and harmoniously, even more so, for we know so much about "regression" and its limitations. BUT: Is "regression" is the right scientific model? NO! #Bookofwhy

12.25.19 2:30pm - (Replying to @tdietterich @Plinz and @GaryMarcus) You'r speaking from a function-builder perspective. But from the end-user, input-output, what have we got? A function that approximates the actual relationship between input and output. It need not capture the pain, labor, flexibility and sophistication of the builder. #Bookofwhy

12.25.19 5:19am - (Replying to @maier_ak @hardmaru and 4 others) When we talk "algorithms" everything seems to be doable in DL or NN etc. But when we talk about "information" the limitations become clear -- information cannot be created by processing data, it is either there or not. As Cartwright said: "No causes in, no causes out." #Bookofwhy

12.25.19 3:25am - (Replying to @recursus @basorot and @LeCun) Fine, but "working out the entailments of a model" requires a data structure to represent that model. Once you commit to such representation, Bingo! You are doing CI! And if you still want to keep your memebership in the DL club, fine, we wont excommunicate you. #Bookofwhy

12.25.19 2:15am - (Replying to @recursus @basorot and @LeCun) Sorry to disappoint, but the "model" cannot unfortunately "be found through a generate-and-test process", b/c there is nothing in the training data that can refute a bad model. It can only be tested when we make a bad policy decision and, then, it's too late. #Bookofwhy

12.25.19 1:24am - To all readers who celebrate it, Merry Christmas from me and from my family in Jerusalem. I fell in love with Christmas the moment we arrived to US, 1960, primarily b/c it reminded me so much of Hanuka as it was celebrated in my home town, with songs and lights from every window.

12.25.19 12:56am - (Replying to @hardmaru @ngutten and 3 others) These proposals reinforce my conclusion that some DL folks have not internalized the Ladder of Causation and its implications. One of these says: No matter how you squeeze the data, no matter how smart the squeezing, you can't get causal information out of it. Sorry #Bookofwhy

12.25.19 12:11am - (Replying to @ngutten and @jsusskin) Interesting viewpoint. Trouble is, regression is such a small part of CI that to say "To do CI with regression requires an extra step" is almost like saying: "To do CI with algebra requires an extra step". Its better to say "Do CI first, add NN if needed" @ylecun #Bookofwhy

12.25.19 12:11am - (Replying to @ngutten and @jsusskin) Interesting viewpoint. Trouble is, regression is such a small part of CI that to say "To do CI with regression requires an extra step" is almost like saying: "To do CI with algebra requires an extra step". Its better to say "Do CI first, add NN if needed" @ylecun #Bookofwhy

12.24.19 11:10pm - (Replying to @ylecun and @LeCun) And there is another reason why I want to convince YOU, rather than Leon. I have noticed a day-and-night difference between folks who actually solved a toy problem in CI and those who know all about how others do it. If you are among the former, Leon would do. #Bookofwhy

12.24.19 10:57pm - (Replying to @ylecun and @LeCun) And there is another reason why I want to convince YOU, rather than Leon. I have noticed a day-and-night difference between folks who actually solved a toy problem in CI and those who know all about how others do it. If you are among the former, Leon would do. #Bookofwhy

12.24.19 10:57pm - (Replying to @Plinz and @GaryMarcus) And what will be lost if we drop the "compositional" and call it just "function approximation"?? or, God forbid "curve fitting"? (see footnote b in #Bookofwhy

12.24.19 10:47pm - (Replying to @GaryMarcus) Great questions, @BaryMarcus! May I add a 5th question: What will be lost to AI if we replace the term DL with "super-efficient way of doing regression" or simply "regression"? A lot will be gained, b/c we know so much about what can and cannot be done w/ regression. #Bookofwhy

12.24.19 10:22pm - (1/ ) (Replying to @ngutten @jsusskin and @LeCun) Sure. A NN trained on specially prepared data would capture the features that the preparation meant it to capture. How much preparation can RL do? See Tweet: By deploying interventions in training, RL allows us to infer consequences of those interventions, but ONLY those
12.24.19 10:22pm - (2/ ) (Replying to @yudapearl @ngutten and 2 others) interventions. A causal model is needed to go BEYOND, ie, to predict consequences of actions not used in training, or even combinations of actions used in training. Are NN useful in CI? Sure! Same as asking "is regression useful in CI?" Sure!. But is NN different? #Bookofwhy

12.24.19 9:12pm - (Replying to @basorot and @LeCun) Wait, Wait! This is an interesting point!. You propose to call the model "part of processing". Fine. Have you seen a DL system given a model and, after processing it with the data, answer questions it could not answer before? This is a critical ingredient ML folks lack #Bookofwhy

12.24.19 9:03pm - (Replying to @jsusskin and @LeCun) If we go beyond the definition, we find Yann saying: Don't say "DL doesn't do causal inference" when you really mean "a plain, supervised neural net does not spontaneously discover causal relationships." The correct answer is: "DL just doesn't! Regardless of the NN!" @Bookofwhy

12.24.19 8:51pm - (Replying to @jsusskin and @LeCun) If we go beyond the definition, we find Yann saying: Don't say "DL doesn't do causal inference" when you really mean "a plain, supervised neural net does not spontaneously discover causal relationships." The correct answer is: "DL just doesn't! Regardless of the NN!" @Bookofwhy

12.24.19 8:23pm - (Replying to @basorot and @LeCun) If you cannot get answers for questions by processing the data, and those questions ought to be answered by a sensible agent, it means that the sensible agent uses some information that is not in the data itself. This extra-information is called "model", not "data."#Bookofwhy

12.24.19 7:01pm - This bold defense of deep learning says that some DL folks have yet to internalize the Ladder of Causation, eg Can any DL-insider think of a good way to convince @Lecun that some questions cannot be answered from data alone, no matter what? #Bookofwhy.

12.24.19 5:12am - (Replying to @HenMazzig) Well said. However, I was informed that Arafat started the erasure of the Jewish history already in 2000, at Camp David. See . Also, the Palestinian claims are weakest when "indigeneity" is defined as an intellectual-cultural state of mind.

12.24.19 4:43am - (Replying to @jrgptrs and @JustinSandefur) In 1987 David Freedman wrote "As others see us", which had a soul-searching impact on social scientists. Econometrics is waiting for an insider to write such a paper. I volunteer to play one of the "others", but the field is begging for a courageous insider. #Bookofwhy

12.24.19 2:27am - (1/ ) (Replying to @jrgptrs and @JustinSandefur) 1/ I perfectly understand the culture that reinforces your arguments, but I beg to question its logic. The fact that "no referee requires us to use [the] Data Fusion approach." does not negate the possibility that the methods used by that culture are grossly outdated. In fact,
12.24.19 2:37am - (2/ ) (Replying to @yudapearl @jrgptrs and @JustinSandefur) readers on this Twitter post have been decrying the insular, echo-chambered character of economics literature in general, and in the area of external validity in particular. So, I am glad you are looking into this issue from a non-cultish perspective. And, BTW, "Data Fusion"
12.24.19 2:42am - (3/ ) (Replying to @yudapearl @jrgptrs and @JustinSandefur) is not "my approach" nor is it something you may choose to apply or ignore. The findings of "Data Fusion" analysis are universal, regardless of how you choose to extrapolate across environments. The environments either permit extrapolations or not and, ignoring the latter
12.24.19 2:49am - (4/ ) (Replying to @yudapearl @jrgptrs and @JustinSandefur) may lead to errors regardless of what referees say or not say. I trust that, after looking into the state of art, you will join your colleagues' efforts to elevate econometrics to the age of modernity. #Bookofwhy

12.24.19 12:14am - (1/ ) This paper introduces an interesting refinement of selection diagrams: If one is interested in transporting a single contrast (say Risk Difference) and is willing to make parametric assumptions to match this aim, then selection diagrams can be pruned
12.24.19 12:14am - (2/ ) so as to reduce the number of covariates involved. This refinement is similar in spirit to that proposed in, which leverages knowledge of how mechanisms interact to produce the outcome. #Bookofwhy
12.24.19 12:49am - (2/ ) (Replying to @yudapearl and @DavidHarrisAJC) This means that, seeing any such claim opens up an opportunity to expose, with facts, figures and dates, the glaring recentness of the claimant's historical heritage.

12.23.19 10:38pm - (Replying to @DavidHarrisAJC) And let's not forget that claims of "no Jewish historical tie w/ Israel & Jerusalem!" are fabricated for a reason; to justify an embarrassing 'no historical tie' on the side of the claimants. As described here

12.23.19 7:31pm - We, who are bound to earth, salute you @Astro_Jessica for showing us how cosmic imagination can enrich a humble tradition, of a tiny people, who did not have much to offer the world, except learning, memory and imagination.

12.23.19 7:16pm - The legitimization of proxies is treated here and here We call it "Measurement Bias and Effect Restoration". BTW, the conditions established are not unique to Pearl-stale analysis; they govern ALL analyses, no escape. #Bookofwhy

12.23.19 4:26pm - (Replying to @AngeloDalli @GaryMarcus and 2 others) I'll second the conclusions. Now, let's do it. That's what the do-operator calls on us to do. #Bookofwhy

12.23.19 3:22pm - (Replying to @EpiEllie and @smueller) If I were a mentor to 5-yr old, I would say: Isn't it amazing? With the same tools we used to determine backdoor conditions you can now determine how to re-weight data to estimate target effects? This possibility will not occur to an economist in the next 17 years! #Bookofwhy

12.23.19 3:09pm - (Replying to @emzanotti) She doesn't. Hanuka represents Jewish sovereignty in some part of the middle east, an idea that @IlhanMN Omar has vowed to eradicate, in blunt betrayal of her Minnesota voters.

12.23.19 6:31am - (Replying to @stuartbuck1 @Jabaluck and @metrics52) Inverting matrices is a good analogy. I would use "solving equations algebraically" vs. trying out all possible solutions and see which satisfies the equations. We can do it by hand, true, but where would science be today w/o algebra? #Bookofwhy

12.23.19 5:10am - (Replying to @charleendadams) Ocho Candelikas is my favorite Chanukah song. Inviting all our Spanish speaking readers to join us in lighting the candles and celebrating the defeat of darkness. Here:

12.23.19 2:13am - Readers elevated by the Holidays spirit are invited to a sing-along in Hebrew, ending with a Chanukah song we used to sing in kindergarten: Sevivon Sov Sov Sov, about 11 minutes into this video Hallelujah!

12.22.19 5:16am - (Replying to @totteh @jttiehen and @keithfrankish) I wrote "sifting", not "shifting", because I am always hoping for some reader to show me a nugget that I have neglected in my combing of this literature. #Bookofwhy

12.21.19 5:52pm - For readers baked in traditional philosophical theories of causation, who wish to see how these gel with modern formalisms, this thesis provides a panoramic view: I am still sifting this tradition for a gold nugget worth adding to SCM #Bookofwhy

12.21.19 12:49pm - (Replying to @mattshomepage and @Elias) More than that! The completeness of the do-calculus is nice, but it does not provide us a procedure for finding an identifying pattern. The algorithm (summarized here: actually finds it, whenever such a pattern exists. #Bookofwhy

12.21.19 11:05am - (Replying to @mattshomepage) The answer is YES. @Elias algorithm can be proven to be "complete" namely, guaranteed to discover ALL doors, where by "doors" we mean patterns that allow identification. #Bookofwhy

12.21.19 10:38am - (Replying to @mattshomepage) I do not buy the "very specific structure" limitation. The class of structures satisfying the front-door condition is not smaller than those satisfying backdoor condition. It is only that the latter is in more common use. #Bookofwhy

12.21.19 10:33am - (Replying to @mattshomepage) Well said, and it entices me to add: When you wish to express WHAT you know, you better express it the WAY you know it, that is, the way it is stored in your mind, not the way it would justify favorite statistical routines. #Bookofwhy

12.21.19 4:41am - (Replying to @PWGTennant and @medrxivpreprint) I am anxious to see the first paragraph, where DAGs are introduced. eg. DAGs are .....

12.21.19 4:20am - (Replying to @ang_hermann) Good luck @ang_hermann on your French translation project. I know it is a work of love, else you would not have undertaken it. I'm looking forward to see @Bookofwhy on French book-shelves, and hear France communicate with Portugal in DAGs. PS. a Spanish one is in the making too.

12.21.19 3:04am - (1/ ) (Replying to @ErakatSaeb) This is not a joke! It is in fact the cause, the root, and the essence of the Palestinian tragedy. Unable to celebrate any holiday connected to the land to which they claim sole ownership. Unable to chant a single hymn authored in the days of Jesus or Judas Maccabeus,
12.21.19 3:09am - (2/ ) (Replying to @yudapearl and @ErakatSaeb) lacking any cultural connection to those days, Palestinians been laboring to fabricate such connection by molesting the heritage of their neighbors, and hoping to be taken seriously and well-intentioned if/when they decide to come to the negotiating table again. If only ...
12.21.19 3:14am - (3/3) (Replying to @yudapearl and @ErakatSaeb) If only @ErakatSaeb understood what this circus acrobatic does to Palestinian posture.

12.21.19 2:04am - (1/ ) Halleluya, It's Hanuka! Sunday night we will be lighting the first candle to commemorate the Maccabees revolt, Jerusalem, 161 BC. Join me in a holiday that inspires all who cherish the ideas of freedom and self-determination. My grandson asked me why we make such a big fuss l
12.21.19 2:04am - (2/ ) about this holiday. I told him Hanuka is our TRUST DEED to the birthplace of our history, more solid even than the ancient synagogues they excavate in Israel, or the arch of Titus in Rome. Why? Because stones can be faked, not so a continuous celebration for 110 generations.
12.21.19 2:04am - (3/3) I wrote about this aspect of Hanuka when the LA Times asked me "what does Israel mean to you?" and why understanding Hanuka is so essential for dreaming any peace prospect between these two equally indigenous peoples.

12.21.19 12:36am - Halleluya!! Good tidings! Kindle has finally uploaded a revised edition of PRIMER with all the errata smoothed to perfection. Enjoy. As to the hard copy, publishers try to keep this information very very secret. But I'll Tweet when it happens #Bookofwhy

12.21.19 12:14am - (1/ ) Gratified to see researchers committed to decision making adopt modern tools of causal inference. I can tell right away that the authors have not graduated from UCLA. How? (1) Consistency is an "assumption" in PO, not so in SCM, where it is a "theorem",
12.21.19 12:14am - (2/2) and (2) Graphs/potential-outcomes symbiosis is not sacrilegious in SCM; it is in fact the norm; the former to express what we know, the latter what we wish to know. See #Bookofwhy (though it may be sacrilegious in some islands of PO, eg Lilliput and secret others)

12.20.19 11:41pm - Yes, worth keeping in mind, and add to it: Hypothesizing means unabashedly committing to a representation (e.g., a model) from which we can deduce how things operate, when the need arises. #Bookofwhy

12.20.19 3:33am - (Replying to @maximananyev @Jabaluck and @metrics52) Do you think this great paper gives potential users any idea about how to select variables for valid reweighing? (even if we knew exactly how study-participation was determined in MA and in NJ). #Bookofwhy

12.20.19 1:28am - In the island of Lilliput they don't teach addition. Why? Because the problems they truly truly wish to solve involve multiplication; addition alone cannot handle. That's in the legendary island of Lilliput. #Bookofwhy

12.19.19 8:25pm - (Replying to @Jabaluck and @metrics52) Ask not what I did wrong nor what I did right, ask what YOU can do to elevate your field to new heights. #Bookofwhy

12.19.19 7:36am - (1/ ) True, Journals normally do not bid and do not solicit. However, what makes an enlightened editor ENLIGHTENED is his/her ability to spot promising new develpments and invite authors to write a review article for readers who have not been exposed to it. Phil Dawid, for example
12.19.19 7:36am - (2/ ) invited me in 1995 to write an article to Biometrika, because he felt that it was about time that Stat readers will learn something about graphical models. He probably faced stiff resistance by reviewers and the established elite. He did it, and his leadership made an impact.
12.19.19 7:36am - (3/ ) I wrote at least a dozen papers thus invited by enlightened editors. You don't have to be a genius to see that your field lags behind in a given area, you need only be a leader to do something about it. This is what I meant by "bidding." Perhaps econ. editors are still
12.19.19 7:36am - (4/ ) fighting for the honor to be the first? Or perhaps they have not been informed of the paper. Alert them. Some editors need a jolt. Play on their patriotic duties to their readership. Econ. students can't afford another decade of arrow-phobic prohibition. Can they? #Bookofwhy

12.19.19 3:49am - (Replying to @SadiqKhan) Commending you on your courage and good will. But to be effective, please heed to an advise of someone who has studied the anatomy of anti-Semitism. Change your statement to read: "a hostile environment for anti-Semites, Islamophobes and Zionophobes. @EinatWilf

12.19.19 2:42am - I bet you already received 4-5 invitations from top Econ journals. Seriously, one can argue till dawn (as @jabaluck & @metrics52 ) that economists do not need new identification strategies. But I dont know one who'd argue they don't need external validity or data fusion.#Bookofwhy

12.19.19 2:32am - (Replying to @Meetasengupta and @prem_k) Would love to discuss those possibilities with you, but my family will be descending on us on Chrismass, so I can't make any plans in advance. Let's play it by ear.

12.19.19 12:12am - (Replying to @Meetasengupta @prem_k and 3 others) I believe we can go a step beyond "try to narrow the gap", since we are in the age of causation, where "trying" is just not good enough. We can at least formalize when gaps are narrowable, what information is needed to narrow them, etc. We have the language to do it! #Bookofwhy

12.18.19 11:40pm - (Replying to @prem_k @Meetasengupta and 3 others) I am not sure what "Evidence Based" people count as "evidence". I know that N. Cartwright, one of the champions of EB has also complained bitterly about external (non)validity of RCT's, and I've tried to introduce her (and Deaton) to S-diagrams: #Bookofwhy

12.18.19 11:28pm - Evidently, wines do taste better with age. As I re-read the 1995 paper on instrumental inequalities,, I can't help but wish that IV's would be introduced today with the same clarity and freshness as they were treated then, in 1995. #Bookofwhy

12.18.19 10:47pm - To readers who asked whether I heard back from Miguel why he goes back to PO, given that it is so easy to generalize in DAGs,, the answer is not yet. I think Harvard folks love DAGs, but still have problems swallowing do-calculus. Who knows? #Bookofwhy

12.18.19 5:20pm - (1/2) Glad to see this ancient inequality finding useful applications in Mendelian Randomization. Truthfully, I knew already in 1995 that it would be resurrected some day, but I wrote it for economists, who loved IV and bemoaned: "it can't be tested". I thought they would jump
12.18.19 5:20pm - (2/2) from joy and celebrate the testability of their beloved IV. Little did I know that economists have been waiting for epidemiologists to take the lead. #Bookofwhy

12.18.19 3:01pm - This paper may just be what econometrics has been waiting for in its effort to catch up with the causal generation. I envision dozens of editors bidding on the right to publish it in their journals. Imagine what such paper could do to the reputation of Econometrika. #Bookofwhy

12.18.19 1:24am - (1/ ) Our success of showing readers how re-weighing works in moving across populations entices me to show how it works in correcting sampling-selection bias. Assume we preferentially select subjects for a study based of an arbitrary set of characteristics (some measured & some not) 12.18.19 1:24am - (2/ ) , and we wish to estimate P(y|x) in the target (unsampled) population. Can the DAG tell us when we can do it? Yes. Whenever we can measure a set Z of variables such that: (1) {X,Z} separates S from Y and (2) The conditional distribution P(z|x) can be estimated for the target 12.18.19 1:24am - (3/ ) (unsampled) population. If these hold, then the re-weighing formula P(y|x) = SEM_z P(y|x,z, S=1) P(z|x) gives us an unbiased estimate of P(y|x) in the target population. Illuminating examples, generalizations and refinements are given in 12.18.19 1:24am - (4/4) Note that the only difference between this re-weighing formula and the one used in transportability is the weighing term P(z|x) vs. P(z). Note also that even @EpiEllie 5-yr old can enjoy and appreciate the power of selection diagrams; how else would one select Z?? #Bookofwhy

12.17.19 6:36pm - Our success of showing readers how re-weighing works in moving across populations entices me to show how it works in correcting sampling-selection bias. Assume we preferentially select subjects for a study based of an arbitrary set of characteristics (some measured & some not)

12.17.19 5:59pm - (1/ ) Great to see Uber joining the causal era. There is one glitch in this article which might be misleading to the novice - the word "approach" appears 19 times. Why misleading? Because "approach" connotes an "option" under the analyst control. Yet the 19 methods discussed
12.17.19 5:59pm - (2/ ) are not under our control but are "opportunities" made available to us by (what we believe is true about) reality, ie, our model of reality (DAG). Once we have a model of reality we cannot speak about the 19 methods as "approaches," ie, as if they were our options. #Bookofwhy

12.17.19 6:31am - (Replying to @jeremyphoward and @calabi_and_yau) Students of causality would read this paper with a constant question in mind: How do you climb the Ladder of Causation, from predictions, to effect of actions, to counterfactuals (eg would this customer be swayed by this discount). I guess it is all in the model used. #bookofwhy

12.17.19 6:25am - (Replying to @HenningStrandin) Thanks for posting this interesting summary. Just in case you find it relevant, I have found comfort in escaping the circularity trap through another route. It is in Section 33.5.3 of this paper: #Bookofwhy

12.17.19 12:54am - (1/3) I am retweeting Charlene's insightful and personal testimony from the holy city of Bristol. The city where anti-semitism does not exist, and whatever does exist has nothing to do with Corbyn, and whatever has to do with him is Jewish invention or Israel's fault.
12.17.19 12:54am - (2/3) Bristol is a city of highly devoted Corbynites who helped turned a true "Labour Party" into a dangerous and un-electable cult of directionless excuse makers. A personal question to each member of your faculty, including prof. Harvey Goldstein and those who like his report:
12.17.19 12:54am - (3/3) What have YOU personally done to fight BDS activities on YOUR campus? Have you too decided that hostilities connected with Israel are kosher hostilities, not affecting students like me who are, or are perceived to be, proud of Israel? Wasn't it YOUR academic duty to act?

12.16.19 11:55pm - (Replying to @calabi_and_yau and @jeremyphoward) I listened to @jeremyphoward talk and kept asking myself: When was it given? It turned out it was given in 2012, before analysts realized that most of his aspirations can be algorithmatized. So, yes, it is related to #Bookofwhy

12.16.19 11:01pm - (Replying to @jouni_helske) Thanks for saving me an n-th attempt to understand what Gelman is talking about. It is not his fault, it is mainly mine. We live in two non-intersecting universes. I have solved a few toy problems, Gelman refuses to try even one. #Bookofwhy

12.16.19 10:56pm - (Replying to @calabi_and_yau) It would be of interest to document the various contexts in which business people yearn for "insight" and see if it confirms my theory that what they yearn for is "causal understanding", or at least some chunk of a causal model compatible with the data. #Bookofwhy

12.16.19 2:58pm - (Replying to @jouni_helske) I will comment on Gelman's post if you tell us why you find it to be an "interesting take." Or, how it could possibly change the way we/others are doing Causal Inference, however minutely. #Bookofwhy

12.16.19 4:28am - I consider myself progressive, so some academic colleagues call on me, resentful of Israel's courting right-wing politicians to pass anti-BDS laws. My answer to them: What have YOU personally done past 10 years to fight BDS activity on YOUR campus? Wasn't it YOUR academic duty?

12.16.19 3:34am - (Replying to @thosjleeper) We have considered, but S-admissible is reminiscent of backdoor admissible for identification. Moreover, it is almost 9 years since its first publication. What would you suggest? Reweighing-admissible"? perhaps. S-separator? "S-Equalizer"? Suggestions? #Bookofwhy

12.16.19 2:31am - (1/4) In view of the dominant role that re-weighing plays in extrapolating effects across populations, and the many Twitter requests for a concise graphical criterion that gives re-weighing its legitimacy, I am retweeting the criterion (called "S-admissibility"), in next 4 tweets.
12.16.19 2:31am - (2/4) It works on a selection diagram in which S nodes represent disparities between the target (*) population and study population (experimental). Z is a set of measurements. To test if Z is S-admissible (1) Remove all arrows pointing to X (2) Check if {X, Z} d-separates S from Y
12.16.19 2:31am - (3/4) If Z passes this test, then the reweighing formula is valid: P*(y|do(x)) = SUM_z P(y|do(x),z)P*(z) In words: Effect at target equals the Z-specific effects at study, averaged over Z, using the target distribution P*(z) as weight. Warning, this is merely a sufficient test.
12.16.19 2:31am - (4/4) Many more opportunities are available for situations in which a S-admissible set cannot be found. see and #Bookofwhy (And dont dismiss "trivial transportability," described in Graphs are fun! #Bookofwhy

12.16.19 1:50am - I am also unsure of what "rainfall became to Economics", but I am retweeting ONE sentence, because it is a GEM: "Mendelian Randomization is just another observational study, with a different set of assumptions." Quasi-experimentalists shoud keep this in mind. #Bookofwhyk

12.16.19 1:38am - (Replying to @davidsirota) You may be Jewish by birth, but that does not prevent you from conflating the issue, perhaps intentionally. Bernie Sanders is criticize for surrounding himself with Zionophobes like Linda Sarsour, consumed by genocidal agenda to eliminate Israel. Jews do not escape issues.

12.15.19 1:35pm - (Replying to @EpiEllie) Yes. (I thought I tweeted it before, but can't find it.)This criterion (called S-admissibility) is sufficient for transporting the effect by simple re-calibration on Z. But this is the simplest kind of transport. We can do much more even if such a Z is not measured.#Bookofwhy

12.15.19 1:23am - (Replying to @DrMikeH49 @markshiffer00 and @YousefMunayyer) Mike, thanks for using Zionophobia, our only fighting word which, for some reason, is still not "the ugliest word in town" but it's getting there. See my reasons for preferring it over anti-semitism:

12.15.19 12:45pm - (Replying to @EpiEllie) Which diagram specifically gives you hard time to decide if {X,Z} separates Y from S (in the revised DAG from which all arrows entering X are deleted)? This is a simple d-separation test -- duck soup in your Epi class. What's the problem? #Bookofwhy

12.15.19 12:21pm - (Replying to @YousefMunayyer) Agree! Equating anti-zionism with antisemitism is misguided; it gives the former a speck of legitimacy that racist ideologies do not deserve. Zionophobia is more dangerous on its own character and genocidal aims. Here is why: It's the ugliest word in town.

12.15.19 12:01pm - (Replying to @yudapearl and @wtgowers) This is another reason why I viewed Corbyn as a danger to humanity: Rabbi Sacks is a champion of universal values. I know him personally. We went together to Muslim schools in London and talked to kids on our common heritage. I weigh his word over Guardian

12.15.19 4:05am - (Replying to @Howard_Lovy and @lsarsour) Which proves that there IS a fundamental difference between "left-wing" and "right-wing" anti-Semitism. The latter is emotional and mother-milked, the former is calculated, strategic and eliminationist, ala @lsarsour .

12.15.19 3:50am - (Replying to @wtgowers) Admittedly, I was unaware of some these very concerning allegations. However, this report was prepared by dedicated scholars whom I know for years. In contrast, I know NONE of the academic signatories on the Guardian. Strange! I should know at least ONE.

12.15.19 3:21am - (Replying to @AndreaSaltelli @RonKenett and @fhuszar) I have read your Nature paper with great interest, yet it is not clear to me where "causal models" reside? In the statistical or mathematical category? My taxonomy of models (Ladder of Causation) is based on what language the assumptions must be expressed. Thoughts? #Bookofwhy

12.15.19 3:06am - (Replying to @RonKenett @AndreaSaltelli and @fhuszar) Pearl has this habit of celebrating what we can do today that we could not do yesterday. Ron has the habit of playing Solomon the wise: "Nothing is new under the sun" relevant or not. I'd go with Pearl; celebrations are fun and yield new results. Solomon doesn't #Bookofwhy

12.15.19 2:22am - (Replying to @james_r_lucas @roydanroy and 3 others) To predict a "wall" and overcome it, you need to have some idea of what's behind it. ML per se has no language to describe "what's behind". But see #Bookofwhy @tdietterich

12.15.19 1:59am - Little Jack Horner, sat in the corner... Oh What a Good Boy Am I. !!! Singing Oh What a merry land is England Singing Oh What a merry land is England

12.15.19 12:51am - (Replying to @mendel_random @stephensenn and 2 others) The #Bookofwhy unveils several appearances of causal notions in the periphery of statistical analysis, it even celebrates with great fanfare David Cox's (1958) explicit expression "quite unaffected". Those peripheral sparks can now be made explicit and unleashed. R U unhappy?

12.14.19 11:46pm - (Replying to @RonKenett and @AndreaSaltelli) I am celebrating a new ability: To take 10 meaningful examples and solve them formally from A to Z, and you are inviting me to go back to the hand-waving days of "gaining understanding" w/o solving any of those examples.What is gained by reminding people of those days? #Bookofwhy

12.14.19 11:05pm - (1/2) (Replying to @RonKenett and @AndreaSaltelli) I am open to be convinced about the existence of "different ways to generalise findings" as soon as you show me (not 'so and so', but YOU) how ONE of those "ways" decides which of the ten simple examples in is generalizable. #Bookofwhy
12.14.19 11:26pm - (2/2) (Replying to @yudapearl @RonKenett and @AndreaSaltelli) 2/2 Elias has compiled dozens of quotes on "different ways" and "so and so", from Campbell etal, to David Cox, to Manski. I don't see why the belief that "there r different ways" gives you more comfort than my conclusion (10 yr exploration) that there r'nt such ways. #Bookofwhy

12.14.19 8:21pm - (1/ ) Good question: "So how do researchers generalize experimental results if they don't know selection-diagrams or, worse, if they (eg econ.) don't even use DAGs?" Ans. They dont. They publish tons of paper with "generalizing" in the titles but, inside, they assume the analyst can
12.14.19 9:11pm - (2/ ) do the S-admissibility test in her/his head and come out with some "conditional ignorability" expression that justifies the result they want. Bingo! They did it! Reminiscent of the way folks did adjustments before DAGs. In we decry this culture in
12.14.19 9:11pm - (3/3) no ambiguous terms. But to those in the cultural bubble, decrying amounts to non-existing. And to economists, it may even amount to: "Don't show our students." Why the resistance? Selection diagrams highlight the needs to start with what we know - too hard for PO #Bookofwhy

12.14.19 7:52pm - I echo Democritus (~400 BC) who said: "I would rather find one cause than be a King of Persia". My bit: "You learn more by solving ONE toy problem than watching 10 panels"- even when everyone tells you "it ain't REAL-DATA" or "its textbook econ." It ain't! Try one! #Bookofwhy

12.14.19 6:54pm - (Replying to @EpiEllie) As they say in the Mishna "A child that does't ask, lead him to a question" (At Ptach Lo). So, which of the following can you read as "S-admissible"? (i.e., the x->y effect is transportable by reweighing on Z) #Bookofwhy

12.14.19 6:42pm - (Replying to @vatsal_maru and @BernieSanders) Poor Bernie wants us to believe he is a "man of principles," but is appointing advisers (eg Sarsour) that are consumed by unprincipled agenda. Why? Because other advisers tell him "this is what young voters go for" -- an unprincipled "principle", proven a disaster by Corbyn.

12.14.19 6:21pm - (Replying to @paulpharoah) You are absolutely right to demand English, but you need to say more than just "gibberish". Tell us where you get stuck.

12.14.19 6:18pm - (Replying to @nathankallus @SusanMurphylab1 and @Susan_Athey) My take: (1) “read the stats literature!” but only after taking "causal inference 101", eg (2) “pay SOME attention to (3) “read the empirical literature! but only after (1)&(2) (4) Solve one "toy problem" in CI, say from #Bookofwhy

12.14.19 3:00pm - Seriously. What did the panelists think when the words "the ML community's approach to causal inference" were spoken? Did they take the word "approach" to mean "aspirations" or "buzz words" or "pay attention to" or (my favorite) "careful thinking" #Bookofwhy

12.14.19 2:42pm - (Replying to @analisereal @EpiEllie and 2 others) The Mishna says: "The shy cannot learn and the strict cannot teach". We are facing a 5-yr old that is too shy to ask, and teachers that are open to any question. Will the shy help with: "I would like to know how...?" or "What do we do in case..." or... or... #Bookofwhy

12.14.19 2:30pm - Great thread! As an outside observer, the punchiest punch-line was: Contrast traditions that start with *the data generating process* versus those starting with *the algorithm*. The most perplexed line was "the ML community's *APPROACH* (???) to causal inference" #Bookofwhy

12.14.19 1:59pm - (Replying to @EpiEllie) What's the problem? Constructing a selection diagram? Or reading it to determine how to transport things? We can go over it one at a time, assuming, of course, that your 5-yr old can read DAGs. & while we are at it, other readers can enjoy its power on favorite probls.#Bookoofwhy

12.14.19 9:39am - (1/ ) Before parting from the beauty of selection diagrams and the S-admissibility criterion, it is important to empower @EpiEllie 5-yr olds with the proper historical perspective. In the same way that backdoor holds the key to the adjustment formula P(y|do(x)) = SUM_z P(y|x,z)P(z)
12.14.19 9:39am - (2/ ) which was used informally since Yule (1899), so does S-admissibility hold the key to the re-calibration formula (or re-stratification, or re-weighing, or re-standardization) P*(y|do(x)) = SUM_z P(y|do(x),z) P*(z) which was used informally by 18th century demographers
12.14.19 9:39am - (3/3) to transport mortality rates across populations (from PI to PI*). We make this historical connection in footnote 13 of and where readers can find additional magics of the graphical S-admissibility criterion. Enjoy. #bookofwhy

12.14.19 2:41am - A new paper deserving our attention, on alternative CI methods: My first reaction: Not clear why triangulating IV with backdoor methods should give us more information than, say, one backdoor method with two distinct admissible sets. #Bookofwhy

12.14.19 1:02am - Sincere apologies to my colleagues in the dpt of Archaeology at UCLA. It is the dpt of Anthropology that deserves the shame, as described here and here: Sorry for the confusion.

12.14.19 12:29am - Corbyn to Pellosi: Nancy, they all lied to us; about the angry mob, the shifting voting base, the radical millenials of Omar, Tlaib and JVP --they are just not there! Voters can't stand liberal politicians when they betray liberal values.

12.13.19 10:16pm - (Replying to @omaclaren) Where can we read more about Fraser and Bunke and what they related the fiducial argument for?

12.13.19 7:13pm - (Replying to @charleendadams) And my only unhappiness with Corbyn's defeat is that (according to analysts) he was not defeated for the reason he should have been: Racism, but for being wishy-washy on Brexit. Still, his defeat removes fears of shame & darkness for all my colleagues in the UK - Congratulations!

12.13.19 4:35pm - Personally, I have NOT encountered antisemitism, as confessed here But my students HAVE, especially in dpts of History and Archaeology, shamefully mentioned in It is the duty of every professor to expose those pockets of hate.

12.13.19 10:46am - (Replying to @charleendadams and @pablo_gps) University of Bristol!!! What a coincidence!!! I believe George @mendel_random is a professor there, and I presume he spends hours fighting the phenomenon you are describing. As I am doing in my university. Our reputation is at stakes, not to speak of our moral commitments.

12.13.19 10:23am - (Replying to @mendel_random and @pablo_gps) The understatement of the year: "Corbyn should have apologised for the behaviour of a smallish number of people mascerading as labour supporters." Compare to a comprehensive, documented report, by professionals that I know personally: "Smallish Number"!!!

12.13.19 1:37am - (Replying to @pablo_gps) The disaster I feared was not Labor (I loved Blair), nor Brexit. I feared seeing a racist like Corbyn as a PM of a European country that would be turned into a safe haven for Hamas and other anti-Western movements.

12.13.19 1:24am - This is the normal controversy we find in the literature. Then we have the added dimension of causality, see, and the fact that Bayes was chasing causes. #Bookofwhy

12.13.19 12:37am - (Replying to @ulrichspeck On the nail. Hot-headed students hijacked the volume on social media, and our politicians believe their voting base is really shifting. Look at what Linda Sarsour has done to Sanders and what Omar and Tlaib are doing to Pelosi.

12.13.19 12:21am - (Replying to @pablo_gps) Upon facing a sure disaster and an unknown, you dont "choose" the unknown but you sure escape the disaster.

12.13.19 12:16am - The pleasure was mutual and I was thrilled by the "philosophical twist" of today's talk. First time I speak about Bayes' pool table (p. 98 #Bookofwhy) realizing that most people (including Bayesians) dont really know what Bayes did and what the controversy is all about. New book?

12.12.19 11:19pm - On such historical events, my grandpa used to bless: Oh God! Hagomel Chassadim Tovim L'Amo Israel. As a secular Jew, but one who does believe in miracles, all my lips can say: Oh God! please give me the illusion that my Tweet has swayed at least one voter to defeat this racist!

12.12.19 9:40pm - (1/ ) (Replying to @EpiEllie) If DAG rules take a 5-year old, then selection takes a 6-year old. First, a tiny correction! Only a tiny fraction of DAG rules (ie, backdoor) is taught to 5-yr. The entire spectrum of identification opportunities (ie, do-calculus) awaits to be embraced in Epitweet #Bookofwhy
12.12.19 10:08pm - (2/ ) (Replying to @yudapearl and @EpiEllie) You are probably hoping for a backdoor rule when you say "5-year old". Don't worry, we've one for your 5-yr old in transportability. It is Def-8 & Cor-1 in If Y can be separated from S in G_X, Bingo! The effect is transportable! (by Eq 3.1) #Bookofwhy
12.12.19 10:19pm - (3/3) (Replying to @yudapearl and @EpiEllie) You see, It's duck-soup easy!! Even to a 5-yr old. All it takes is a simple language to say how populations differ, and a deep sense of belief that, if things are done the hard way in PO, they can be done the easy way in our natural language. #Bookofwhy

12.12.19 9:26pm - (Replying to @EpiEllie) The new name reminds users that the diagram stands for TWO populations, not one, and that the selection nodes [S] just switch from one population to another. This in mind,anyone who understand DAGs should understand 2-DAGs, ie, S-diagrams, as Elias shows .

12.12.19 7:27pm - (Replying to @EpiEllie) If you are happy with the functions behind the arrows, just follows the beautiful slides by Carlos and Elias, here: and tell me if/where their explanation of selection diagrams is less than simple. We will continue from there. Is it a deal? #Bookofwhy

12.12.19 3:15am - (Replying to @eddericu @EpiEllie and 2 others) This are great slides indeed, but they still require one to understand that behind the DAG's arrows there are FUNCTIONS and, not less important, that writing y=f(x,u*) is another way of writing y=g(x,u), with g neq f. #Bookofwhy

12.12.19 1:20am - My My! I almost forgot I did this podcast. Now the counterfactuals are beginning to play tricks on you: "You should have answered it this way, or that way." No, I shouldn't; we know that free will is just an illusion #Bookofwhy

12.11.19 11:00pm - (Replying to @yudapearl and @Corey_Yanofsky) But, for havens sake, even if p-Jews (People-hood defined Jews) are just a minority (they are a vast majority) aren't they entitled to dignity and respect like other p-minorities? Especially by the Office of (don't smile) "Equity, Diversity, Inclusion" at my university, UCLA.?

12.11.19 10:09pm - (Replying to @EpiEllie) I've tried to get inspired by your Dags Cartoons, to explain *selection diagrams* but, OOPS, I got stuck on agreeing about what a DAG is. "A Research Tool"? True. But a DAG also represents a collection of functions that defines a population. Can we agree, to continue? #Bookofwhy

12.11.19 9:28pm - (Replying to @EpiEllie) You know that I am the last person to "change topic" on Tweeters. For me, a selection diagram is just a DAG's way of representing two populations that differ in some of their underlying mechanisms (ie, equations). If you know DAGs, you know selection diagrams. #Bookofwhy

12.11.19 8:55pm - (Replying to @Corey_Yanofsky) And Ruth said: Your people will be my people, and your God will be my God. People-hood first, religion second.

12.11.19 7:05pm - I am retweeting, because this question comes up again and again: Why are diagrams "natural", "transparent" and "communicable"? Why? It also touches on the sanctity of RCT and the obsession with "well defined intervention" that some brilliant scholars refuse to abandon #Bookofwhy

12.11.19 6:51pm - (Replying to @EpiEllie) For ordinary folks who represent knowledge the way it is stored in our brain, in terms of FACTORS influencing one another, who have learned (eg. #Bookofwhy or that Potential Outcomes are DERIVED properties of that knowledge, not its elementary particles

12.11.19 5:13pm - Which reminds me. Tomorrow I'll be speaking to History of Science Reading Group (3pm UCLA Royce 314) where I will resurrect Bayes, Galton, Pearson, Wright and Haavelmo and ask again: Where have all the causes gone? Will Haavelmo's legacy survive in Econometrics ?#Bookofwhy

12.11.19 5:01pm - Remember Ander's "Russian Roulette" and all the arguments why one needs to use cross-world cryptic like P(Y(1)|Y(0)) to generalize experimental results? Well, Carlos and I took another look, Good news: You can do better in common language. #Bookofwhy.

12.11.19 4:37pm - It is a victory for BDS and an indictment of American universities that it should take a Presidential Order to define "Jewishness" the same way as Jews are defining themselves today, and the same way they have been defining themselves since the exodus from Egypt.

12.11.19 4:14pm - (Replying to @mstephens999 and @y2silence) Sure, because "salah" is not the only factor that determines the outcome of the game. Recall, however, a "unit" u, in both SCM and in PO stands for the sum total of all factors necessary to make the outcome 0 or 1, leaving nothing to uncertainty, thus no Pr(*). #Bookofwhy

12.11.19 3:50pm - (Replying to @mstephens999 and @y2silence) What makes "u" different than "i" ? If there is no difference, and both are just "anyone", we can remove the index. But if "u" have an allergy to zinc or some specific experience with zinc, we say: Pr (cold|do(zinc), specific knowledge). Still no index. #Bookofwhy

12.11.19 5:11am - This thoroughly-documented report by the Simon Wiesenthal Center chases sleep from my eyes. I imagine UK slipping into Corbynism, what it would do to my people, to my former country, Israel, and to the moral compass of the civilized world. Extremely worried.

12.11.19 1:44am - (Replying to @mstephens999 and @y2silence) I haven't thought about what Pr(Y(u)=y | do(X(u)=x)) could mean. But now that I think about it, the answer would be either 1 or 0. Because the Y(u) takes some value, say y' , which is either equal to y or not. If u is picked up at random we get P(Y=y|do(X=x)) w/o u. #Bookofwhy

12.11.19 1:33am - (Replying to @patrickkloesel @ASSAMeeting and 2 others) If invited, I'll break my vacation and go to San Diego to remind economists how Haavelmo defined "economic model", as I started to do here:, and of all the new things they can do by restoring his vision. #Bookofwhy

12.10.19 12:08am - This is a powerful presentation of the message I am trying to send to ML folks: "You and I know that we must climb the ladder of causation at some point. Therefore, to be safe, let's look at the rungs before we climb, and surely before we make believe we are already up"#Bookofwhy

12.9.19 9:52pm - (Replying to @mstephens999 and @y2silence) I see it as an advantage, to be able to distinguish between questions that can be answered by intervention, do-expressions, and those that cannot, because they are Rung-3 counterfactuals. Look how important this distinction is: , #Bookofwhy

12.9.19 4:31pm - (Replying to @mstephens999 and @y2silence) Do-expressions are probabilisitic, involving populations. To go to individual level, we go to SCM (Stractural Causal Model), where U=u is the individual. All counterfactuals, PO, ignorability etc etc can be derived from a fully specified SCM.(Including "consistency")#Bookofwhy

12.9.19 3:22pm - (Replying to @VartanKA @haldaume3 and 3 others) Attempts to synthesize have been successful. All is unified now and we fully understand where each concept stands in the scheme of things. eg. "ignorability" is logically implied by 'backdoor." No "synthesis" in Mostly Harmless where graphs mean disloyalty.#Bookofwhy

12.9.19 3:10pm - (Replying to @mstephens999 and @y2silence) When we write P(y|do(x), z) it means the full distribution of Y (not "mean") for every intervention do(X=x), for every z-specific class of individuals. It is so heterogeneous that calling it "heterogeneous" only diminishes its heterogeneity. #Bookofwhy

12.9.19 4:09pm - This is one merit of social networks: Reviewers and editors are no longer operating with impunity, they know that history will judge them, albeit not personally. They do not want to shame their Journal or their Alma Mater. #Bookofwhy

12.9.19 1:25am - (Replying to @RonKenett) Recall that transportability theory covers statistical transportability as well. So, what exactly changed that needs to be "transferred"? Population? Environment? distribution? #Bookofwhy

12.9.19 12:45am - (Replying to @RonKenett) The fact that they do not deal with causal effects does not make it "much wider". What does? Can you name a problem area (not an algorithm) that makes the scope of this survey "wider"? #Bookofwhy

12.9.19 12:27am - Where did you say this "editor" got his education? The quote is a vivid example of how research is impeded by those who slap careless labels on other people works (yesterday it was NPSEM-IE). As to content, I have not met a more heterogeneous object than a DAG. #Bookofwhy

12.8.19 11:30pm - I was also surprised to find the context worse. Modern statisticians are so embarrassed by his text that some resist: No, he wan't one of the founders of statistics. Some even told me: "Well, he was a socialist." Which proves that dogmas & politics still rule science. #Bookofwhy

12.8.19 5:24pm - (Replying to @AsraNomani) When Linda Sarsour says a Zionist can’t be a feminist....#zioness #WW84 #GalGadot, she is just confessing to being a FAKE feminist (see unfit to lead any progressive movement. When will Bernie see the harm she is causing him?

12.8.19 4:29pm - (Replying to @AndersHuitfeldt) Many people (unnamed) interpret the term "independent errors" (IE) to mean "Causal Markov condition", which is not just "semantics", but a severe restriction. This, IMHO, is a "gross misunderstanding," perhaps (?) unintended by the authors of the term "IE". #Bookofwhy

12.8.19 4:04pm - (Replying to @AndersHuitfeldt) The "misunderstanding" is on the part of readers who take "IE" as a restriction on SCM. Giving researchers the OPTION to express independence should not be construed as a restriction, it is LIBERATION when compared to systems that do not provide such option. #Bookofwhy

12.8.19 3:33pm - (Replying to @AndersHuitfeldt) That's a gross misunderstanding. The Causal Markov condition is NOT assumed to hold a priori in nonparametric SCM. Labeling it NPSEM-IE (IE ="independent errors") is misleading. A notational system that gives you the OPTION to assume IE should not be labeled IE. #Bookofwhy

12.8.19 3:08pm - (Replying to @hazqiyal) The bi-directed arrows denote correlated errors for some reasons, e.g., latent variables with names, un-named latents, selection bias, etc. If you suspect correlation, use them. Absence of those bi-directed arrows implies error-un-correlatedness for any reason. #Bookofwhy

12.8.19 8:06am - (Replying to @hazqiyal) Almost every diagram in Causality (2000) has bidirected arrows, X< - - ->Y, which stand for correlation between eps_X and eps_Y. #Bookofwhy

12.8.19 3:03am - "Transfer learning" (another word for "transportability") is what called my attention to this paper The title: "Learning subject-specific causal effects" made me doubly curious: Is this possible? Isn't "class-specific" the most we can hope for?#Bookofwhy

12.8.19 2:09am - Causal discovery enthusiasts among our readers should be interested in this symbiotic "discover-identify" approach, just posted on our website: #Bookofwhy

12.7.19 11:41pm - (Replying to @steppenjiff) Corbyn cannot be excused as lacking the brain to foresee the genocidal consequences of his words and actions. Do you genuinely believe the call for "from the river to the sea" is anything less than genocidal?

12.7.19 10:27pm - Agree with @GaryMarcus on the insufficiency of causal inference. But the "type-token distinction" is IMHO well formalized in structural models as "unit level vs. population level distinction" (Causality p. 310). The firing-squad example in #Bookofwhy shows token counterfactuals.

12.7.19 8:28pm - (Replying to @jon_y_huang) Thanks for posting @diedre_tobias . And I am happy to see that Jamie and I agree on at least one thing: “We are in the midst of a Causal Revolution!". Our disagreements dwarf when we remember what it took to get here and the huge work still ahead of us. #Bookofwhy

12.7.19 8:01pm - Luciana Berger is my UK heroine, for standing up to Corbynism.

12.7.19 7:44pm - (Replying to @aminjorati) "voices are not heard?" You must be kidding! Come to UCLA, my campus, where Palestinian victim-hood is celebrated day and night with megaphones and marching bands, and where BDS activists chant: “We don't want two states, we want 1948!" You must be kidding!

12.7.19 7:01pm - (Replying to @peder_isager @dingding_peng and 3 others) In cyclic equations, the d-separation may breakdown when non-linearities are introduced. [It holds for linear cyclic systems]. For "Identifying Independencies in Causal Graphs with Feedback" see #Bookofwhy

12.7.19 6:49pm - "Zionophobic" is the only word I use for Zionophobes -- a much uglier word than "Anti-Semite". The latter may often be a victim of misguided upbringing. The former has no excuse, it's a cold & calculated call for genocide.

12.7.19 6:30pm - There are (at least) two inaccuracies in this paper:(1) Nonparametric SEM does not assume error-independence (see<-->) therefore, no one should use the acronym NPSEM-IE. (2) NPSEM does NOT assume that all variables (V ) can be intervened on. See #Bookofwhy

12.7.19 12:13pm - (Replying to @wtgowers) She will always find a reason, or a paragraph, or an adjective to prevent her from voting for a two-state solution. Why? Such a vote would turn her into a Zionist, thus losing her voting base. Watch and see how next time she says: It does't include the "right of return".

12.7.19 5:30am - (Replying to @Kaniska_Mohanty) I do recommend Primer. Chapters are available here, and it is a really good book!!! #Bookofwhy

12.7.19 5:26am - Omar, Tlaib, AOC & Pressley -- the four elephants in the room. American law makers have not internalized a simple fact: Zionophobes cannot utter the words "two-states solution" and Israel's neighbors suffer from the same speech impairment.

12.6.19 11:00pm - (Replying to @geomblog @L_badikho and 4 others) These people are wonderful, very broad, and technically superb. But all the King's horses and all the King's men can't turn a mathematical impossibility into a possibility. "Fairness" requires acknowledging certain impossibilities, else all King's men would be wasted.#Bookofwhy

12.6.19 1:20am - (Replying to @markcannon5 @AnimaAnandkumar and 5 others) I hope you are not suggesting that DL is beyond enrichment, or that DL+CI symbiosis is not a compelling way of overcoming DL's shortcomings. see #Bookofwhy

12.6.19 1:06am - Continuing this track on Bernie Sanders & Linda Sarsour marriage. Poor Bernie thought Zionophobes could be kept under leash for some semblance of progressive values. Now he is wondering who owns the leash, who's "surrogate" to whom?

12.5.19 11:23pm - (Replying to @L_badikho @timnitGebru and 6 others) Agree totally. Logic too originated from moral disputations. At the same time we should not ignore the fact that current ML proposals for dealing with "fairness" are dominated by a model-blind thinking, while "fairness" demands a model-based approach. #Bookofwhy.

12.5.19 11:08pm - (Replying to @L_badikho @timnitGebru and 2 others) I was explicit about "expand their expertise" which is different from "lack technical expertise". The former is respectful and hopeful, so am I. #Bookofwhy

12.5.19 10:51pm - Very well put. I am seeing this "ignorance above all" phenomenon first hand at UCLA, Department of Archaeology. I, too, cast the conflict as a morality play, a simple choice between co-existence and elimination. A very simple choice.

12.5.19 10:13pm - (Replying to @timelessdev) CI stands for "Causal Inference" as defined and described in or

12.5.19 10:06pm - Agree! And the question is: Are ML folks flexible enough to understand that "fairness" demands expansion of their expertise? Or will they insist on handling it the model-blind way, and waste the funds and trust that society now pours on them. Same to "explainability #Bookofwhy

12.5.19 7:20pm - "Fairness" involves CI issues such as mediation (ie distinction between direct and indirect effects), undesirable side effects, and protected variables. It is hard to see how these issues can be managed without a causal model of reality. #Bookofwhys

12.5.19 7:14pm - Why are "fairness" and ML so often mentioned in the same breath? True, the former can be compromised by irresponsible practices of the latter, but this does not mean that the tools available to ML folks are adequate for preventing those compromises. #Bookofwhy

12.5.19 7:01pm - A glimpse at the NeurIPS 2019 program shows that CI is well represented. IBM, for example, seems to be taking the DL+CI symbiosis quite seriously: I fail to see though why "fairness" and ML are so often mentioned in the same breath. #Bookofwhy

12.5.19 6:05pm - I can try and contribute a couple of papers, but I am afraid the title will invite "what computers cannot do" type of speculating philosophers (eg Penrose, Searle) with whom I do not wish to be identified. #Bookofwhy

12.5.19 2:41pm - (Replying to @IzaTabaro) Bernie can't decided between the Sarsourish crowd that flocks his band wagon and the principles he learned at home.

12.5.19 1:57pm - Viewing history from a causal lens is an exciting adventure for me. I will be giving this talk to the History & Philosophy of Science Group, at UCLA. December 12, 3 pm, Royce Hall Room 314. If you happen to be around that day, just drop in. #Bookofwhy

12.5.19 6:36am - (Replying to @BlakeFlayton) On the contrary. She just confessed to being an incurable racist, hence unfit in any progressive movement. Poor Bernie, choosing "surrogates" turns out more hazardous than one thinks. Has he given up on attracting non-racist voters. ???

12.4.19 11:29pm - Linda Sarsour (Bernie Sanders' surrogate) is my muse this week. Friday she said: "Israel is based on supremacy," Yesterday, she apologized: "[No, No]...not the Jewish people. I apologize for the confusion." I had to tweet her: Sister Linda! Sister Linda!

12.4.19 7:24pm - (Replying to @lsarsour) Sister Linda, Sister Linda! A decent way to take back what you said is to say what you meant: e.g., "Zionism is NOT what I said! It's a noble dream, a people's homecoming journey, one we should revere, respect and assist --That's what I meant." Then apologize for the confusion.

12.4.19 5:33pm - (Replying to @RabbiWolpe) Here is the context: "I love Jews, and Bernie is my proof, but I was bought up to hate, and Zionophobia is safe, and so much chic"

12.3.19 10:03pm - I have thoroughly enjoyed re-reading @EinatWilf 's article a day after Linda Sarsour, the darling of Fake Feminism (see, tried again to contaminate Zionism with her racist thoughts.

12.3.19 12:56pm - (Replying to @EpiEllie) I do not recall any paper rejected for having been pre-printed. But I may be too young to recall all cases, I'll check with my buddies on the Editorial Board.

12.3.19 8:20am - (Replying to @nickchk) This may be a new IV, but the graph-void presentation makes me commiserate with econ. readers who must make sense of cryptic sentences like: "an instrument correlated with the confounders, but which itself is not causally related to the direct effect of the treatment." #Bookofwhy

12.3.19 1:44am - (Replying to @DemMaj4Israel) "Israel’s built on idea that every people is entitled to self determination" - How true! But did you ask Bernie Sanders and Nancy Pelosi if they think about that truth when they choose advisers and spokespersons?.

12.3.19 1:17am - Being part of a condemned minority is not new to Jewish history, but being a PROUD member of such minority IS new and, strangely, you feel fortunate not sharing the embarrassment of the Sapiens delegates voting tomorrow, having to face their families the next day.

12.2.19 3:25am - (Replying to @juli_schuess @thosjleeper and 6 others) On first scan, this is a very valuable paper, connecting the huge survey-inference literature with causal graphs and showing how crucial design issues in the former are rooted in causal thinking and can now be formalized and made transparent in graphical models. Kudos! #Bookofwhy

12.2.19 1:29am - (Replying to @curtatkisson) I'm not familiar with Meng 2018, should I? And who are "they"? Sorry, I can't guess.

12.1.19 10:40pm - (Replying to @JDHaltigan) I rarely agree with Gelman but, in this case, he is right that the true effects can be higher or lower than what they found; the same holds under "natural experiments" which are just observational studies with different assumptions, see, #Bookofwhy

12.1.19 9:43pm - (Replying to @RonKenett) What is not clear to me about the "second dimension" in your paper is whether the "matching of data resolution/aggregation and analysis goal" is a desirable feature to have, or an algorithm that actually establishes this matching in some formal setting. #Bookofwhy

12.1.19 7:47pm - (Replying to @AviMayer) It's worse than pure antisemitism, its ugly Zionophobia. It's time we stop treating the latter as the lesser of the two evils. See And let's deny its legitimacy using its real name, Zionophobia, not antizionism, which sounds almost as innocent as anti-tax

12.1.19 4:55pm - Granularity in Scientific Explanation has been a favorite topic of discussion for many philosophers, but when cast in a specific representation (DAG), capable of producing explanations, it now assumes a new level of clarity: #Bookofwhy

12.1.19 4:16pm - A new paper on my desk (i.e., screen), calling for attention, combining linear or nonlinear conditional independence tests with a causal discovery algorithm to estimate causal networks from real-world data. #Bookofwhy

12.1.19 3:22pm - (Replying to @brianoflondon @EinatWilf and 2 others) Fairly gloomy picture, I've suspected this kind of defeatism for quite some time. Any healthy organization trying to pressure the @BoardofDeputies into some level of pride and conviction?

12.1.19 8:21am - (Replying to @yudapearl @XiXiDu and 2 others) And too often they treat Antizionism as a semi-acceptable form of racism, less dangerous than Antisemitism. It isn't.

12.1.19 7:53am - (Replying to @XiXiDu @EinatWilf and @HananyaNaftali) Part of the reason Antizionism has metastasized among the political left is that OUR leaders and OUR spokespersons so rarely challenge its legitimacy and barely expose its moral stench.

12.1.19 7:39am - For heaven's sake, why is everyone questioning Corbyn on antisemitism, a question he is so skillful in deflecting. Why not charge him squarely with incurable Zionophobia, a moral deformity he cannot possibly hide. @EinatWilf @HananyaNaftali

12.1.19 1:31am - (Replying to @MaccormickIan and @ewenharrison)