Conversation

Identity communities often form based on outsider understandings of fields on the boundary of science and pseudoscience. When I asserted recently that fMRI stuff is largely nonsense, I got many outraged responses, apparently mostly from psychiatrists.
Quote Tweet
Obviously, few disciplines in the last 2 decades have made as much progress as human neuroimaging. But I remain worried about common inferences drawn, especially in the psychiatric literature. Hope to get a paper out on this topic by the end of the year.
Show this thread
4
43
The inferential path between the fMRI instrument and anything meaningful is extraordinarily long, complex, and tenuous. Just at the front end it involves several stages of statistically torturing the distorted and noisy data to get some stable signal out of it.
1
20
I would guess few psychiatrists can follow details of fMRI data processing methods, so their faith in it has another basis. Confronted with the extreme nebulosity of human mental dysfunction, having SOME authoritative knowledge source must be reassuring?
1
11
This tweet thread prompted by the analogy with upset responses whenever I say “deep learning stuff is mostly nonsense.” Those seem to come mostly not from actual AI researchers, but AI fans. Personal and non-professional community identities depend on belief in AI progress.
3
18
“Quasiscience” may be a useful term for a field somewhere between science and pseudoscience. The foundations are dubious and the whole thing may be nonsense. There well may be some real stuff there, but it’s impossible to sort out which, given the methods used.
1
22
Quasisciences address problems everyone wants answers for, but for which no good methods are available. Practitioners collude to obscure the foundational problems. Funders want answers and choose to overlook doubts. Identity communities form around the field’s outputs.
5
35
Replying to
Hmm. Seems like you’re conflating application métis and fan theories a bit? Most programmers don’t understand semiconductor manufacturing but build their theories atop abstractions that they trust up to a point. Programming lore does not appear to be a quasiscience in your sense.
Replying to and
There is a spectrum going from programming lore on one end (based on strong, nearly leakproof abstractions) through fMRIology to social psych and say startupology. The abstractions they depend on get progressively leakier, and conclusions shakier.
2
4
Replying to and
Fandom seems almost like an epiphenomenon of shaky foundations but not in itself an indictment of the foundations. You look for more social reassurance when your truths have a half life of 5 years rather than 50. But that’s degree, not kind.
1
Show replies