Man, this sure is a dicy topic, but I do think it’s pretty likely that Torres has a personality disorder, and that modeling these kinds of things is often important.
A while ago we had a conversation on the forum on whether Elon Musk might be (at least somewhat) autistic. A number of people pushed back on this as ungrounded speculation and as irrelevant in a way that seemed highly confused to me, since like, being autistic has huge effects on how you make decisions and how you relate to the world, and Musk has been a relevant player in many EA-adjacent cause areas for quite a while.
I do think there is some trickiness in talking about this kind of stuff, but talking about someone’s internal mental makeup can often be really important. Indeed, lots of people were saying to me in-person that they were modeling SBF as a sociopath, and implying that they would not feel comfortable giving that description in public, since that’s rude. I think in this case that diagnosis sure would have been really helpful and I think our norms against bringing up this kind of stuff harmed us quite a bit.
To be clear I am not advocating for a culture of psychologizing everyone. I think that’s terrible, and a lot of the worse interactions I’ve had with people external to the community have been people who have tried to dismiss various risks from artificial intelligence through various psychologizing lenses like “these people are power-obsessed, which is why they think an AI will want to dominate everyone”, which… are really not helpful and seem just straightforwardly very wrong to me, while also being very hard to respond to.
I don’t currently have a great proposal for norms for discussing this kind of stuff, especially as an attack (I feel less bad about the Elon autism discussion, since like, Elon identifies at least partially as autistic and I don’t think he would see it as an insult). Seems hard. My current guess is that it must be OK to at some point, after engaging extensively with someone’s object-level arguments, to bring up more psychologizing explanations and intuitions, but that it currently should come pretty late, after the object-level has been responded to and relatively thoroughly explored. I think this is the case with Torres, but not the case with many other people.
Yeah, I don’t want to make a claim that reference to an individual’s mental condition would be categorically inappropriate. However, I think at a minimum there needs to be a reason for making the assertion that furthers an important interest, that the assertion is tailored to that interest, and that there isn’t a clear yet less inflammatory & invasive way to get the information across.
I think there are few cases in which this test would be met as applied to a critic. Saying that the critic has a long history of dishonest, volatile, paranoid, or whatever kind of behavior (and showing the receipts where appropriate) is more convincing to explaining why people shouldn’t engage than a thinly-supported armchair diagnosis.
While I don’t agree with everything in that quote, I do see some points of convergence—there is awareness of downsides, consideration of what is potentially to be gained from the discussion, a suggestion that this should not occur without significant object-level engagement first, and some sense of narrow tailoring (insofar as discussion of “psychologizing explanations and intuitions” is more narrowly tailored than trotting out a stigmatized DSM/ICD diagnosis).
For what it’s worth, I endorse @Habryka’s old comment on this issue:
Yeah, I don’t want to make a claim that reference to an individual’s mental condition would be categorically inappropriate. However, I think at a minimum there needs to be a reason for making the assertion that furthers an important interest, that the assertion is tailored to that interest, and that there isn’t a clear yet less inflammatory & invasive way to get the information across.
I think there are few cases in which this test would be met as applied to a critic. Saying that the critic has a long history of dishonest, volatile, paranoid, or whatever kind of behavior (and showing the receipts where appropriate) is more convincing to explaining why people shouldn’t engage than a thinly-supported armchair diagnosis.
While I don’t agree with everything in that quote, I do see some points of convergence—there is awareness of downsides, consideration of what is potentially to be gained from the discussion, a suggestion that this should not occur without significant object-level engagement first, and some sense of narrow tailoring (insofar as discussion of “psychologizing explanations and intuitions” is more narrowly tailored than trotting out a stigmatized DSM/ICD diagnosis).