I read Robert as accusing you of attempting a rhetorical trick in which, by making a disjunctive accusation where one of the disjuncts is shocking[1] and grave, you algorithmically intend to intimidate people into accepting the other disjunct, which they would be less likely to do if you argued for it on its own merits rather than pairing it with the shocking disjunct. I don’t think you would be getting this pushback if you had said, “Maybe she’s a good judge of character; after all, every time she judges a man to be safe and is correct, that’s some amount of probabilistic evidence that she’s capable of making such distinctions rather than being tied to base rates.”
(I feel bad about being in the “tone police” role here, and anticipate that you have reasons why the “Maybe she’s a good judge of character” alternative in fact omits important substantive criticisms you mean to make, but this comment still seems good to post, because I think your diagnosis of the relevant defect in site culture being a denial of “true accusations are better than false ones” is off-base.)
I feel bad about being in the “tone police” role here,
I’m more objected to Benquo’s comment on grounds of it being false, or at least not engaging with what Sinclair obviously meant, than about tone. [edit: er, I guess I also think that giving a false dichotomy to make a point seem persuasive seems to fall under “deception” rather than tone and I don’t think you need to bring tone into the question to object to the comment].
I do think a pretty valuable I’ve gotten a lot from Benquo over the years has been reframings of things in ways that make me engage with something that my natural frame glossed over. But, in this case I think his implied point just… doesn’t seem logically valid and is kinda offtopic? Or at least he hasn’t made the case for it. (I think “maybe she’s a good judge of character?” is still missing the point of what Sinclair pretty obviously meant. If the base rates are low, the dice coming up negative isn’t a very interesting outcome in the first place and you shouldn’t be updating [edit] much [/edit] from a single instance)
(I separately think Robert’s comment wasn’t very good either, doesn’t quite check out even as a clever quip, and he probably should have resisted the urge to get partial credit for “not being clever”)
I am, to be clear, actively interested in Benquo engaging with the base rates question and explaining why his frame here is useful in spite of that background fact.
Yeah true, but I’d reword as “shouldn’t be updating much from a single instance”, and I think concretely it shouldn’t be enough of an update to substantially privilege hypotheses like “Sinclair is making some kind of cognitive error here”.
(My actual guess is that Benquo has a background frame/hypothesis like “People frequently underweight their own personal intuitions over statistics [or, vaguely assumed ‘statistics’ that they probably don’t even have a citation for and if they looked up the study it might not even say what they thought it said].” And, like, I totally think this might be true and relevant and worth having privileged anyway, but I don’t think Sinclair’s anecdote is evidence for or particularly illustrative of it)
Yes, that’s what the first half of my comment was intended to convey. I disendorse the way I communicated that (since it was both unclear and provocative).
I’m complying with Sinclair’s explicit preference to be treated as someone who might possibly do crimes, by not censoring the flow of credence from “people who don’t expect me to do crimes to them are making a mistake” to “I have done crimes to such people.” You are asking me to do exactly what Sinclair complained about and assume that they’re necessarily harmless, or to pretend to do this.
I think this is a class of situation where people prefer obfuscation: the preference is not for you to assume that the subject is harmless, but to take actions that merely imply that the subject might not be harmless without explicitly spelling out the “subject is harmful” disjunct, such that the message slips past a System 1 scapegoating circuit while System 2 can piece together what it needs to know. Implying something without stating it isn’t the same thing as pretending the opposite.
I read Robert as accusing you of attempting a rhetorical trick in which, by making a disjunctive accusation where one of the disjuncts is shocking[1] and grave, you algorithmically intend to intimidate people into accepting the other disjunct, which they would be less likely to do if you argued for it on its own merits rather than pairing it with the shocking disjunct. I don’t think you would be getting this pushback if you had said, “Maybe she’s a good judge of character; after all, every time she judges a man to be safe and is correct, that’s some amount of probabilistic evidence that she’s capable of making such distinctions rather than being tied to base rates.”
(I feel bad about being in the “tone police” role here, and anticipate that you have reasons why the “Maybe she’s a good judge of character” alternative in fact omits important substantive criticisms you mean to make, but this comment still seems good to post, because I think your diagnosis of the relevant defect in site culture being a denial of “true accusations are better than false ones” is off-base.)
Anecdotally, I felt a jolt of fear when I first saw your comment without knowing to whom it was addressed.
I’m more objected to Benquo’s comment on grounds of it being false, or at least not engaging with what Sinclair obviously meant, than about tone. [edit: er, I guess I also think that giving a false dichotomy to make a point seem persuasive seems to fall under “deception” rather than tone and I don’t think you need to bring tone into the question to object to the comment].
I do think a pretty valuable I’ve gotten a lot from Benquo over the years has been reframings of things in ways that make me engage with something that my natural frame glossed over. But, in this case I think his implied point just… doesn’t seem logically valid and is kinda offtopic? Or at least he hasn’t made the case for it. (I think “maybe she’s a good judge of character?” is still missing the point of what Sinclair pretty obviously meant. If the base rates are low, the dice coming up negative isn’t a very interesting outcome in the first place and you shouldn’t be updating [edit] much [/edit] from a single instance)
(I separately think Robert’s comment wasn’t very good either, doesn’t quite check out even as a clever quip, and he probably should have resisted the urge to get partial credit for “not being clever”)
I am, to be clear, actively interested in Benquo engaging with the base rates question and explaining why his frame here is useful in spite of that background fact.
Quantitatively small updates are still updates!
Yeah true, but I’d reword as “shouldn’t be updating much from a single instance”, and I think concretely it shouldn’t be enough of an update to substantially privilege hypotheses like “Sinclair is making some kind of cognitive error here”.
(My actual guess is that Benquo has a background frame/hypothesis like “People frequently underweight their own personal intuitions over statistics [or, vaguely assumed ‘statistics’ that they probably don’t even have a citation for and if they looked up the study it might not even say what they thought it said].” And, like, I totally think this might be true and relevant and worth having privileged anyway, but I don’t think Sinclair’s anecdote is evidence for or particularly illustrative of it)
Yes, that’s what the first half of my comment was intended to convey. I disendorse the way I communicated that (since it was both unclear and provocative).
I’m complying with Sinclair’s explicit preference to be treated as someone who might possibly do crimes, by not censoring the flow of credence from “people who don’t expect me to do crimes to them are making a mistake” to “I have done crimes to such people.” You are asking me to do exactly what Sinclair complained about and assume that they’re necessarily harmless, or to pretend to do this.
I think this is a class of situation where people prefer obfuscation: the preference is not for you to assume that the subject is harmless, but to take actions that merely imply that the subject might not be harmless without explicitly spelling out the “subject is harmful” disjunct, such that the message slips past a System 1 scapegoating circuit while System 2 can piece together what it needs to know. Implying something without stating it isn’t the same thing as pretending the opposite.