Skip to primary content

Statistical Modeling, Causal Inference, and Social Science

Main menu

  • Home
  • Authors
  • Blogs We Read
  • Books
  • Sponsors

Post navigation

Cheap talk and nukes 1, 2, 3, Many Tea Parties?

Lucia de Berk found not guilty

Posted on April 14, 2010 3:35 PM by Andrew

Maarten Buis writes:

A while ago you blogged on the case of Lucia de Berk:

Maybe you would like to know that her conviction has been overturned today.

This entry was posted in Miscellaneous Science by Andrew. Bookmark the permalink.
  • Art
  • Artificial Intelligence
  • Bayesian Statistics
  • Causal Inference
  • Decision Analysis
  • Economics
  • Jobs
  • Literature
  • Miscellaneous Science
  • Miscellaneous Statistics
  • Multilevel Modeling
  • Obituaries
  • Papers
  • Political Science
  • Public Health
  • Sociology
  • Sports
  • Stan
  • Statistical Computing
  • Statistical Graphics
  • Teaching
  • Zombies
  1. Chris on Forking paths in LLMs for data analysisJuly 21, 2026 11:51 AM

    The whole thing about AI discussions is weird. There seem to be three sources of critiques that highlight limitations and…

  2. Daniel Lakeland on Forking paths in LLMs for data analysisJuly 21, 2026 11:32 AM

    Dale, a chatbot is a function that maps from a sequence of tokens to a new token plus some "scaffolding"…

  3. Olaf Zimmermann on Was this USDA survey really “redundant, costly, politicized, and extraneous”?July 21, 2026 11:28 AM

    @Dale Lehman, who states "... when this administration does not like the way people react to information, its solution is…

  4. Joshua on Forking paths in LLMs for data analysisJuly 21, 2026 11:25 AM

    Anoneuoid, Tris - To the extent that I've been able to follow this discussion cleanly... In my views the math…

  5. Dale Lehman on Forking paths in LLMs for data analysisJuly 21, 2026 11:12 AM

    Daniel I know the AI is not "looking" at the graph. That is largely my point: it has done some…

  6. Daniel Lakeland on Forking paths in LLMs for data analysisJuly 21, 2026 11:11 AM

    > When you ask it to provide output, you are asking for trouble. I couldn't agree more. I just want…

  7. John G Williams on Was this USDA survey really “redundant, costly, politicized, and extraneous”?July 21, 2026 11:06 AM

    Or, on the issue of food security, how about this from The Bulwark? https://www.thebulwark.com/p/trumps-one-big-beautiful-hunger-crisis

  8. John G Williams on Was this USDA survey really “redundant, costly, politicized, and extraneous”?July 21, 2026 11:03 AM

    Given what the administration is doing with climate science (https://www.nytimes.com/2026/07/20/climate/climate-change-manual-national-academy-science.html), why should we be surprised?

  9. Andrew on Was this USDA survey really “redundant, costly, politicized, and extraneous”?July 21, 2026 10:56 AM

    Olaf, I added an Internet Archive link. And, yes, it's disgusting that our government is producing this sort of Soviet-style…

  10. Daniel Lakeland on Forking paths in LLMs for data analysisJuly 21, 2026 10:53 AM

    Dale, the AI is never "looking at the graph it produced" nor is the AI "looking at reports" It will…

  11. Olaf Zimmermann on Was this USDA survey really “redundant, costly, politicized, and extraneous”?July 21, 2026 10:52 AM

    I get "Access denied" when I click on that proposal' link ... Anyway, judging by the short snippet here, has…

  12. Dale Lehman on Forking paths in LLMs for data analysisJuly 21, 2026 10:52 AM

    Interesting - I also like that paper very much. You see it as an AI-killer and I see it as…

  13. Dale Lehman on Forking paths in LLMs for data analysisJuly 21, 2026 10:41 AM

    Daniel Your example prompted me to upload an image of a cat and dog and ask the AI what kinds…

  14. Daniel Lakeland on Forking paths in LLMs for data analysisJuly 21, 2026 10:36 AM

    This is great. Here is an actual use of AI where its randomness and its mercurial capacity to be blown…

  15. Dale Lehman on Forking paths in LLMs for data analysisJuly 21, 2026 10:28 AM

    Daniel No, I've never had that happen (though I don't doubt the many stories of people that have that experience).…

  16. Daniel Lakeland on Forking paths in LLMs for data analysisJuly 21, 2026 10:26 AM

    To give an example, I have seen chatbots being asked "please describe the contents of this image". You can then…

  17. Dale Lehman on Was this USDA survey really “redundant, costly, politicized, and extraneous”?July 21, 2026 10:19 AM

    I agree that the announcement does not make clear whether "survey" and "report" should be taken to mean the same…

  18. Daniel Lakeland on Forking paths in LLMs for data analysisJuly 21, 2026 10:14 AM

    Dale, have you never had a situation where you ask a chatbot to do something, it gives you an answer…

  19. Dale Lehman on Forking paths in LLMs for data analysisJuly 21, 2026 9:48 AM

    I find these comments (Chris and Matt) useful - more useful than admonitions that I am endangering myself by not…

  20. Daniel Lakeland on Rebecca Makkai points out: Fancy three-dimensional sets are easier to construct in books than in movies, but harder to explainJuly 21, 2026 9:44 AM

    As a kid I would become so absorbed in reading people would comment on it, and have difficulty interacting with…

  21. Matt Skaggs on Forking paths in LLMs for data analysisJuly 21, 2026 9:18 AM

    Chris, Thanks for posting the link to the paper. I read it last year and remembered it when reading Tris'…

  22. Chris on Forking paths in LLMs for data analysisJuly 21, 2026 8:48 AM

    I haven’t added to the very interesting discussion here but I might add my own use of AI (basic free…

  23. Chris on Rebecca Makkai points out: Fancy three-dimensional sets are easier to construct in books than in movies, but harder to explainJuly 21, 2026 7:05 AM

    One thing they love to do in Hollywood is save money when possible. You need two people to have a…

  24. Andrew on Limitations of “Limitations of Bayesian Leave-One-Out Cross-Validation for Model Selection”July 21, 2026 6:50 AM

    Michael: Someone should point Danielle to our 1995 paper, Avoiding model selection in Bayesian social research, which I think she'd…

  25. Chris on Rebecca Makkai points out: Fancy three-dimensional sets are easier to construct in books than in movies, but harder to explainJuly 21, 2026 6:40 AM

    I'm surprised you're dubious about a Dicken's movie that matches Dickens - I gave you a couple of examples. The…

  26. Michael Bedward on Limitations of “Limitations of Bayesian Leave-One-Out Cross-Validation for Model Selection”July 21, 2026 5:36 AM

    For anyone looking for Danielle's great blog post in 2026, you can find it here... https://blog.djnavarro.net/posts/2023-04-12_bayes-factors/

  27. Anon on Forking paths in LLMs for data analysisJuly 21, 2026 3:52 AM

    Maybe this has been noted somewhere in this thread already, but it seems directly relevant to the original blog post:…

  28. AAAnonymous on Rebecca Makkai points out: Fancy three-dimensional sets are easier to construct in books than in movies, but harder to explainJuly 21, 2026 2:21 AM

    Ah interesting, I had not heard of aphantasia. I just looked it up. I can imagine that the ideal amount…

  29. David, a Bostonian in Tokyo on Forking paths in LLMs for data analysisJuly 20, 2026 11:34 PM

    Hey! Tris! Now you're cookin'. You're begining to sound like me!!! "Measuring the “predictive skill” of a syntactic engine that…

  30. Meese on Forking paths in LLMs for data analysisJuly 20, 2026 10:39 PM

    Tris, I can’t really understand the jargon but I think agree with you. AI hallucinates all the time and it’s…

  31. Tris Simondsen on Forking paths in LLMs for data analysisJuly 20, 2026 8:59 PM

    Anoneuoid, this is precisely not about replications and predictive skills. That is basically asking a closed loop to validate itself.…

  32. Anoneuoid on Forking paths in LLMs for data analysisJuly 20, 2026 8:26 PM

    @Tris recent peer-reviewed research in Nature (Shumailov et al.). First, 2024 isn't recent for llms. They are used totally different…

  33. Tris Simondsen on Forking paths in LLMs for data analysisJuly 20, 2026 7:35 PM

    Anoneuoid, your own example perfectly illustrates why you are not refuting my premise. Finding a counterexample to an 85-year-old math…

  34. Tris Simondsen on Forking paths in LLMs for data analysisJuly 20, 2026 6:57 PM

    Daniel, because of the Halting Problem, it is provably impossible to build a perfect, a priori static analyzer that guarantees…

  35. Tris Simondsen on Forking paths in LLMs for data analysisJuly 20, 2026 6:39 PM

    Matt, I do not pay for any LLM service either. Perhaps the day where I begin the actual construction of…

  36. Daniel Lakeland on Rebecca Makkai points out: Fancy three-dimensional sets are easier to construct in books than in movies, but harder to explainJuly 20, 2026 6:29 PM

    As a person with aphantasia I can also assure you that the ideal amount of descriptive detail varies wildly from…

  37. Daniel Lakeland on Forking paths in LLMs for data analysisJuly 20, 2026 6:17 PM

    Dale, Let me just try to give some concrete less wild-eyed version of what I mean. There are multiple ways…

  38. Dale Lehman on Forking paths in LLMs for data analysisJuly 20, 2026 6:13 PM

    Daniel If you are worrying about how I (and others) will be impacted by the many improper uses of AI,…

  39. Jonathan (another one) on Rebecca Makkai points out: Fancy three-dimensional sets are easier to construct in books than in movies, but harder to explainJuly 20, 2026 6:13 PM

    Far be it from me to disagree with Andrew when he says I have a point. I'll give you Remains…

  40. Daniel Lakeland on Forking paths in LLMs for data analysisJuly 20, 2026 5:21 PM

    Also Dale, to clarify a bit. I also agree it's unlikely you personally will be targeted to trash your econ…

  41. Dale Lehman on Forking paths in LLMs for data analysisJuly 20, 2026 5:19 PM

    Daniel If only humans were so trustworthy. You made me laugh. We can't live without trust, but you are asking…

  42. paul alper on Rebecca Makkai points out: Fancy three-dimensional sets are easier to construct in books than in movies, but harder to explainJuly 20, 2026 5:13 PM

    Ah yes, New Haven and Broadway tryouts. The "wrongest" review of a musical is attributed to Avrom Hirsch Goldbogen, aka…

  43. Daniel Lakeland on Forking paths in LLMs for data analysisJuly 20, 2026 5:04 PM

    Dale, You rightly bring up that there are trust issues at every level of software. One aspect of that trust…

  44. Anonymous on Forking paths in LLMs for data analysisJuly 20, 2026 5:00 PM

    Dale, you may be interested in this: https://www.wired.com/story/fast16-malware-stuxnet-precursor-iran-nuclear-attack/ There was a bug called Fast16 created in 2005 for the purpose…

  45. Dale Lehman on Forking paths in LLMs for data analysisJuly 20, 2026 4:15 PM

    Daniel How do I know that when I do a linear regression using regressoR or sk.learn.linear_model.LinearRegression or regress in stata…

  46. Daniel Lakeland on Forking paths in LLMs for data analysisJuly 20, 2026 3:44 PM

    Tris says: > This is exactly why I keep hammering on the need for architectural boundaries rather than just better…

  47. Andrew on Rebecca Makkai points out: Fancy three-dimensional sets are easier to construct in books than in movies, but harder to explainJuly 20, 2026 3:42 PM

    Chris: These are some good examples. Also there've been lots of great movies based on OK but not great novels,…

  48. Chris on Rebecca Makkai points out: Fancy three-dimensional sets are easier to construct in books than in movies, but harder to explainJuly 20, 2026 3:18 PM

    In my opinion there are quite a few great books that have made great movies - these are often oldish…

  49. Anoneuoid on Forking paths in LLMs for data analysisJuly 20, 2026 3:17 PM

    @Daniel You'll need known good examples of the in/output to test against. You can bootstrap by giving multiple bots the…

  50. Andrew on Rebecca Makkai points out: Fancy three-dimensional sets are easier to construct in books than in movies, but harder to explainJuly 20, 2026 3:15 PM

    Jonathan: By "New Haven," yes, I was referring to that city's former role as a location for Broadway tryouts, not…

Proudly powered by WordPress