What resonates with me in Jordan's interview: 1. Humanlike behavior ≠ human intelligence. 2. AI capability ≠ human judgment 3. Intelligence is social and institutional. 4. AI doomerism has limits. Where I hesitate: privileging engineering expertise over the humanities.

Interdisciplinarity. One day you’re working on a paper for a literary conference, the next it’s philosophy, then a connection to technology catches your eye and pulls you somewhere else. Living in the liminal.

Aww ❤️

@dailyneuron.bsky.social

Ian Hacking’s “Making Up People" and the ninth of his ten “engines”: Bureaucratise! What happens after a classification begins traveling across records, institutions, and time? And when does an interpretation become evidence?

Interesting finding. “Has ___” and “is ___” may refer to the same diagnosis, but they do different work. Also striking: stigma judgments varied substantially across labels and participants. The classification may remain stable while its social meaning does not.

I’m not saying the Wellcome Trust funded this particular research. I’m saying that more and more organizations are pursuing methods they hope will bring greater objectivity to mental health assessment.

Worth scrutinizing as psychiatric AI develops. An LLM’s ability to map diary language onto existing measures of psychopathology should not be confused with independently establishing psychopathology. Agreement between representations is not confirmation of the reality being represented.

In a recent referral thread, therapist describes client as “denying” their “disease.” “Denying” pathologizes disagreement; “disease” turns the contested diagnosis into an objective fact. Disagreement = pathology, but coercion is unexamined. The person most affected is absent from the discussion.