> Haidt is not the world's most careful data analyst
We can, and probably should, just end the discussion there. Haidt is really good at finding data to support his claims, but then failing to mention that the correlation he's describing as "definitive" is, actually, really weak. This happens throughout "The Anxious Generation," at least.
Calling him "directionally correct" when he's pretty bad at actually showing the work as to why he is correct is just saying that you think he has a good point because his vibes match your vibes.
I don't think I'm just saying that. I'd say instead:
1) evidence in favor of reasonable, unsurprising priors does not need to be held to the same standards of rigor as it would for less likely hypotheses. Put differently, extraordinary claims require extraordinary evidence. You can call my agreement with Haidt on the big picture "vibes" but I'd say instead that I just judge the likelihood that the underlying claims are true to be high.
2) the "Haidt production function" faces tradeoffs between making big points, writing books, and attending to every detail. When I read people's critique of his meta-analytic techniques (the first link I posted), I saw a lot of folks saying, he's not even doing meta-analysis because he's not weighting by precision! Reading that, I thought, he very much is doing meta-analysis: even if he's not doing "random effects meta-analysis" that you'd learn in a textbook, he's synthesizing many quantitative results, which is the core of it. (I have written three meta-analyses and RA'd for a fourth.) And when the 'proper' technique was applied, it shrunk the effect size estimate from like 0.2 to 0.15, which, like, if whatever hypothesis was true at 0.2, it's probably also true at 0.15. Social science theories don't generally stand or fall on differences like that. So I thought he came out looking like the wiser person there. Academics have a tendency to get bogged down in implementation details. Haidt doesn't.
(I don't expect this to be persuasive, just explaining why I don't find his data 'errors' to be a nonstarter.)
IIRC the effect size at 0.15 was narrowly for pre-teen girls on social media. Every other age, and all boys, were below 0.1 when looking at total screen time (i.e. games, youtube). Parents should check up on young girls, but most kids will be fine.
Does it seem plausible to that a system that is intentionally, systematically, algorithmically optimized to keep your attention and drive engagement really has so little affect on us?
We can, and probably should, just end the discussion there. Haidt is really good at finding data to support his claims, but then failing to mention that the correlation he's describing as "definitive" is, actually, really weak. This happens throughout "The Anxious Generation," at least.
Calling him "directionally correct" when he's pretty bad at actually showing the work as to why he is correct is just saying that you think he has a good point because his vibes match your vibes.