i mean i guess there’s a whole spectrum from sexual assault to harrassment to plain old social awkwardness but “how safe i would feel” is a pretty good proxy both for how common I believe these things are and how likely I am to feel creeped out/feel uncomfortable
like “being creeped out/uncomfortable” is for most people at least something of a truth-tracking thing and we should optimise for the thing it’s trying to track
[quite importantly “be legibly safe to my system 1″ is a very different thing to “try to Goodhart on my system 1′s sense of safety”!]
i mean i guess there’s a whole spectrum from sexual assault to harrassment to plain old social awkwardness
but “how safe i would feel” is a pretty good proxy both for how common I believe these things are and how likely I am to feel creeped out/feel uncomfortable
like “being creeped out/uncomfortable” is for most people at least something of a truth-tracking thing and we should optimise for the thing it’s trying to track
[quite importantly “be legibly safe to my system 1″ is a very different thing to “try to Goodhart on my system 1′s sense of safety”!]