6 Comments
User's avatar
Kevin Munger's avatar

The initial scope conditions are begging the question. From yesterday: "For open-ended questions, Meehl thought clinical expertise was indispensable. It was only for problems with simple multiple-choice answers where he thought statistical decision-making could play a role."

Once we subset to the questions that machines can analyze, then the machines can outperform humans. The myth of John Henry. There's no meaningful way to compare machine prediction against human prediction in cases where machines can't function. But we don't have any way to measure the value of human prediction in these contexts -- if we could measure it, machines could do it better.

Quantitative science requires translating the world into machine language. Progress in statistical prediction comes from expanding the scope of the world that is machine-readable. This is the imperative of the "bureaucratic mindset" of high modernism developed by the state -- and more recently, the inductive "high-tech modernism" developed by tech companies: to simplify humanity in order to govern it.

^ https://direct.mit.edu/daed/article/152/1/225/115009/The-Moral-Economy-of-High-Tech-Modernism

Expand full comment
Ben Recht's avatar

Exactly this! I 100% agree that all statistical prediction is question begging.

One minor caveat: I am not ruling out the possibility there are quantities out there that are measurable, predictable, and yet not statistical. But I need to think a bit to formulate some good examples.

Expand full comment
ramjet_oddity's avatar

What interests me about this series is that Meehl isn't just an ordinary psychologist, he's a Freudian psychoanalyst (it's of interest of mine—I went from rationalism to psychoanalysis too.) I wonder how he squares it with his analytic work, because psychoanalysis is *the* exemplar of "clinical-first", the analysand has to be treated in her "singularity", her particular history that has to be considered, which is why psychoanalyst sorts make the claim (and rightly so imo) that it doesn't make sense to average the results of psychoanalytic analysands to compare with eg. CBT.

I wonder of course if Meehl's response is that as an analyst he's not in the business of making standardized predictions (and neither can any psychoanalysts I can think of, really)

Expand full comment
Ben Recht's avatar

Stay tuned, Lecture 11 is all about psychoanalysis and serves as a dramatic counterpoint to Lecture 10.

Expand full comment
Static Void's avatar

Why shouldn't we "subscribe to the bureaucratic utilitarian mindset"?

Expand full comment
Mohan's avatar

The use of Cohen’s d here is really peculiar. Since they are comparing two percentages, an odds ratio would make much more sense.

Expand full comment