Science, Expertise, Trust…and Peanuts

Yann LeCun, who is  Chief AI Scientist at Meta and an ACM Turing Award Laureate, is obviously a very smart and accomplished guy.  At X, he has been arguing for the trustworthiness of science and scientists and asserting that it makes sense for scientists to lean Left…because “they care about facts.”  (here, for example)  I don’t think many of us would disagree with the value of the scientific method and the importance of objective information; I do feel, however, that Yann’s remarks fail to address the importance of funder pressure, peer pressure, and Groupthink on scientific priorities, scientific conclusions, and especially on what gets asserted as The Science by media and politicians.

There has also been some discussion of this cartoon and various experiences that various people have had with doctors who wouldn’t listen.

Very relevantly, there is an article in today’s WSJ about peanut allergy.  The writer, Dr Marty Makary, says that concern about these allergies that began to rise in the 1990s…but that in fact, peanut allergies at that time were rare and mostly mild.  But starting in the year 2000, the prevalence began to surge:

What had changed wasn’t peanuts but the advice doctors gave to parents about them. The American Academy of Pediatrics (AAP) wanted to respond to public concern by telling parents what they should do to protect their kids from peanut allergies. There was just one problem: Doctors didn’t actually know what precautions, if any, parents should take. Rather than admit that, in the year 2000 the AAP issued a recommendation for children 0 to 3 years old and pregnant and lactating mothers to avoid all peanuts.

The AAP committee was following in the footsteps of the U.K.’s health department, which two years earlier had recommended total peanut abstinence. That recommendation was technically only for children at high risk of developing an allergy, but the AAP authors acknowledged that “the ability to determine which infants are high-risk is imperfect.” Using the strictest interpretation, a child could qualify as high-risk if any family member had any allergy or asthma.

Dr Makary notes that the peanut allergy epidemic is largely a US and UK phenomenon…his African students report no such allergies in Africa, and Makary says that there is also no peanut problem in his own origin country of Egypt.  And, interestingly, Jewish children in Israel have one-tenth the rate of peanut allergies compared with Jewish children in the UK, suggesting that genetic predisposition is not responsible for the country-to-country difference.

Many Israeli children are fed a peanut-based food called Bamba, and Dr Gideon Lack, a pediatric allergist and immunologist in London, suggested that early consumption of peanuts leads to a low prevalence of the allergy at older ages. He coauthored a paper making this point in 2008, but (this) publication was not enough to uproot groupthink. Avoiding peanuts had been the correct answer on medical school tests and board exams, which were written and administered by the American Board of Pediatrics. For nearly a decade after AAP’s peanut avoidance recommendation, neither the National Institute of Allergy and Infectious Diseases (NIAID) nor other institutions would fund a robust study to evaluate whether the policy was helping or hurting children.

From 2005 to 2014, the number of children going to the emergency department because of peanut allergies tripled in the U.S. By 2019, a report estimated that 1 in every 18 American children had a peanut allergy. Schools continued to ban peanuts, and regulators met to purge peanuts from childhood snacks as EpiPen sales soared.

and

In a second clinical trial, published in the New England Journal of Medicine in 2015, Lack compared one group of infants who were exposed to peanut butter at 4-11 months of age to another group that had no peanut exposure. He found that early exposure resulted in an 86% reduction in peanut allergies by the time the child reached age 5 compared with children who followed the AAP recommendation. 

From 2005 to 2014, the number of children going to the emergency department because of peanut allergies tripled in the U.S. By 2019, a report estimated that 1 in every 18 American children had a peanut allergy. Schools continued to ban peanuts, and regulators met to purge peanuts from childhood snacks as EpiPen sales soared.

Makary is basically asserting that peanut allergy is largely an iatrogenic disease.  One would think that if this was the case, the original recommendations on draconian peanut-avoidance might need to be modified.  But after talking with an allergist friend, Dr Makary said:

The AAP’s absolutism in 2000 had made the recommendation hard to walk back.

I am sure that there are many, many other examples like this, some of them with a lot worse consequences than the Peanut Panic.  It is actually dangerous to treat any individual or institution as beyond challenge.  In aviation, the reluctance to challenge authority (as in the case of a Captain and a First Officer) was identified as a significant safety problem quite a few years ago, and is explicitly addressed in the training of flight crews.  The same principle should apply more broadly.

The Platonic Form of what Science should be, is subject, when translated into operational form, to all the factors that affect other human activities and institutions: incentives, groupthink, political pressure, ambition, cliquishness, etc.

See also my related post Starvation and Centralization.

13 thoughts on “Science, Expertise, Trust…and Peanuts”

  1. Foster, ya beat me to it. Well done.

    Scientists care so much about facts that communist scientists—and I’ve met some as late the 2000s—believe that mass murder is OK because then their beautiful theory of communism would then be true, making all that blood worth it.

    Martin Gardner wrote “Fads & Fallacies in the Name of Science.” Scientists are just as stupid, willful, self-deceptive as anyone else. Yann is proof.

  2. Much of what passes for “health” information is the opposite of science. When I’m in a charitable mood, I characterize it as a plausible conjecture.

    Consider what it would have taken to confirm the “peanut guidance”. Years long studies enrolling thousands of prospective parents and then infants. Many years and many millions to produce one data point. Much easier the exercise an “abundance of caution”. Except it turned out the people being cautious didn’t understand the way the human immune system develops and their caution produced results the exact opposite of what was intended. Then since the “science” was “settled”, a quarter century of needless suffering and some significant number of deaths, still ongoing. And now, how long will various “authorities” continue to parrot the same dangerous advice because of inertia?

    As the man said, it’s not what you don’t know, it’s what you know that ain’t so that’ll get you.

  3. The problem with liberals is not that they are ignorant, it’s that they know so much that isn’t true. — G.K. Chesterton

  4. Michael Chrichton in The Andromeda Strain mentions “the law of 48”.

    Several years ago, it was asserted that the human genome had 48 chromosomes, and many pictures proved it.

    Somewhat later, someone showed that there were really on 46 chromosomes, and the pictures proved it.

    They were the same pictures.

    The Law of 48: “All scientists are blind.” He’s not my favorite author (I had a fun time fisking parts of Congo), but he had a point.

  5. Discussion, especially decisions, especially absolutism about peanuts remind me of how much I don’t know.

    As a carpenter, I casually use a tape to measure something to within 1/32”, and with real care when my eyesight was younger and when the measurement was useful or important, to 1/64”, but could not to 1/128”. Same limits when, as an engineer I spot checked quality control teams inspecting products before they were shipped. When as a mechanic or engineer I needed greater precision, I used calipers or dial indicators. With these, along with the instrument being checked against a precision-ground test piece (aka “joe block”), I could have a confidence of +/- a quarter thousandth. In a job design engineering gate valves using ceramic surfaces, I used light interference patterns to measure flatness to within millionths of an inch. As a physicist, I am familiar with the sorts of understanding that come from making measurements to far more infinitesimal precision, a quadrillion times more accurate than that millionths of an inch.

    Biological and medical science cannot reach that level of precision. And in comparison to mechanical engineering or physics, those fields involve a lot more complexity, a lot more variables that influence how one evaluates the measurement. That sort of limitation means less certainty about one’s understanding, less confidence in one’s ability to predict.

    So when decades ago I first heard about peanut allergies, first experienced, for example, not getting a bag or peanuts while flying on a business trip, I was a bit skeptical. How would researchers measure and figure out the biological mechanics that led to that practice? Was it merely ‘covering the bases’ to avoid problems—what in more recent times got the label ‘abundance of caution’? Was it avoiding insurance liabilities? Meanwhile, I realized that while I was allergic to poison ivy, I knew people whom poison ivy did not bother at all. “Maybe”, I thought, “peanut allergies are similar to that diversity.” Although I remained skeptical, my not having peanuts was very low stress inconvenience and it did enable caring for others.

    Then, in the early 2000s, I first hand witnessed a peanut allergy reaction. At our kitchen table, my wife had our 4-month-old granddaughter on her lap. My wife was making “ants on a log”, peanut butter filled celery sticks with raisins on top. She offered the granddaughter some peanut butter. It only touched our granddaughter’s lips. Immediately the granddaughter’s lips began to swell. The response was rapid enough that not only could we see that the swelling had happened, we could see the swelling happening. Gave us both an “Oh, no! What have I done?” scare.

    Over the next couple decades that granddaughter has had a couple hospital several day oxygen tent stays triggered by accidental exposure to peanuts, eg, peanut oil in a product that she did not eat but someone else in the restaurant did. That, despite her care to recognize and avoid such exposure. She has weighed and tried several different suggested response paradigms, with some small degree of success. But she remains allergic…and very careful.

    I doubt the correctness of the claims of Yann LaCunn about ‘facts’…and remain more than a little skeptical about absolutized assertions regarding peanut allergies. Furthermore, one cannot miss the obvious waffling, political correctness, karen confrontations. These confuse making balanced evaluations much less caring choices that have concern for others while not actually just following a fad.

    Nonetheless, I simultaneously acknowledge that sometimes we have to make decisions about medical subjects that even in the 21st C we have not mastered. I’m glad that discussion and research about peanut allergies continues. Awareness of the history of medicine (ought) makes one humble.

  6. A question asked to a sample of physicians in Boston: “If a test to detect a disease whose prevalence is 1/1000 has a false positive rate of 5%, what is the chance that a person found to have a positive result actually has the disease, assuming you know nothing about the person’s symptoms or signs?”

    Very few respondents gave the correct answer, which is 2%. The most common answer was 95%.

    https://x.com/jeremykauffman/status/1837835646146118100

    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4955674/

    Should note that this survey was from 1978, which was a LONG time ago…I wonder if anyone has tried to measure this again more recently?

  7. Roy,

    Your comment is spot on. There is a great deal of complexity when it comes to determining causality, especially in the sciences and then in biology. Of course that tends to get forgotten in the rush for grants and publication

    As the saying goes, you can get the data to say anything if you torture it enough

    I used to be on the social science side of the fence until I got tired of poking holes in causal explanations… one day I had to sit on a panel where everyone was bowled over by a paper and its conclusions. I asked the panel about how confident they were , they said extremely, and I asked whether they were confident enough to bet their life on it. Of course none were but on the other hand you have a hard time falsifying anyone’s theory

    Things do move forward over time and our understanding increases, but anyone who insists that we should take a scientific finding at a given point of time, at face value as the basis for concrete action is an idiot.

  8. “assuming you know nothing about the person’s symptoms or signs”

    The trouble with that assumption, and with similar ones in other such thought experiments and quizzes is that it’s a bad one. You do know something about the person’s symptoms or signs – you know that they were significant enough for the person to be given the test.

  9. “you know that they were significant enough for the person to be given the test.”

    This is specifically wrong in the case of “screening” tests and is where even small false positive rates cause the most trouble. 1% of 10 million is 100,000 needless follow up procedures, each of which will have their own error rates.

    The fight for numerical literacy in the medical profession has been going on for a long time and is nowhere close to the end. Numerical literacy among politicians is scarcer than hen’s teeth.

  10. In aviation, the reluctance to challenge authority (as in the case of a Captain and a First Officer) was identified as a significant safety problem quite a few years ago, and is explicitly addressed in the training of flight crews.

    I remember hearing about Korean Flight 801 and how that crash happened because the crew couldn’t challenge the pilot enough.

  11. Nate Winchester – September 24, 2024 at 8:45 am:
    I remember hearing about Korean Flight 801 and how that crash happened because the crew couldn’t challenge the pilot enough.

    Because of some peculiarities of the Korean language. There are different grammatical forms for addressing equals, inferiors, and superiors. It makes contradicting a superior extra difficult. This effect was apparent in the recorded cockpit talk.

    After investigating the Flight 801 crash, KAL adopted English as the language of all flight operations.

  12. Peanut allergies are an anaphylactic reaction to peanut oils used as adjuvants in vaccines. But no doctor or medical scientist will tell you that

Comments are closed.