I think it gets a bit more complicated than that because there are feedback loops. The problem is that an expression of the "s/he is dumb" sort is not necessarily a bona fide evaluation of someone's smarts. It may well be (and often is) just an insult -- and insults are more or less fungible.

I definitely don't discount the "sour grapes" scenario as something that probably happens a lot. In fact, I think that a lot of people's assessments of other people's intelligence involve, to put it kindly, subjective judgments along those lines, which is part of why I'm advocating trying to disrupt those.

That problem is likely to be mostly a function of two things: (1) How large a social network do you want to have (or are capable of maintaining); and (2) What's the quality of the fish in the pond in which you are fishing?

I definitely agree that those factors are pretty relevant to the aforementioned problem, but they're kind of moot. After all, (1) is equivalent to "Having a utility function that defines this as a problem", and (2) is something you can't necessarily control (If you see it as enough of a problem to move, I suppose you can, but that seems pretty expensive and it would be a shame to come to that solution without trying something like what I'm suggesting first). I'm merely suggesting that the perception of (2) may sometimes arise from an ill-formed manner of assessing "quality of fish".

How do you define and measure intelligence, then? When you say "Alice is more (or less) intelligent than Bob", what exactly do you mean?

Um, well, I guess I should quote myself here:

I think that IQ is a pretty good measurement for a lot of purposes, and that there's a tendency in lay circles to undervalue it as a measure of a person's intelligence (In the vague socially-applicable sense we're talking about. Let's say "thinking correctly and clearly" for the sake of argument)

I think that as far as things we can assign cardinal values to and compare on a continuum, IQ is our best bet, but there do seem to be some nebulous other contributing factors (Maybe the much-touted EQ, a decent education, or some other "general life experience" factors? I dunno) which can make someone at least appear more or less intelligent than their IQ might imply (Again, operationally defined as "seeming to think clearly, correctly, and quickly". If you'd like to revise this operational definition to "exactly IQ" we can do that, and I'll still argue that it's not something most people are good at detecting from a first impression). Like I actually said, I think IQ is fine, and that most people undervalue its importance. I'm not sure where you got mixed up here. We could redefine "clear, correctly, quickly" as "interesting" rather than "intelligent," although for me personally that's necessary but not sufficient

I would agree there is a lot of self-fulfilling prophesies happening here, but I think they have much more to do with things like self-confidence and much less with making correct intelligence estimates, especially ex ante.

Self-confidence may be some people's problem, but it's definitely not everyone's problem. Does it strike you as impossible or even unlikely that some people have the problem of dismissing people out of hand and thus drastically decreasing their potential social circle in undesirable ways?

These things are not exclusionary -- you start with a speed-optimization and you continue with a better scheme as you get more information. If you get stuck on your cache hit, that's a general problem not specifically tied to evaluating other people.

I agree that getting stuck on one's cache hits in social assessment is not somehow a special case rather than a specific instance of a more general phenomenon. I would argue that social situations are a great problem domain in which to apply general rationality techniques, and that the method for ameliorating a problem I perceive some (but not all) people dealing with social isolation to have can be generalized to "Tabooing concepts," something that's already gotten coverage here. I think that the domain is of enough interest to many people that this application of said technique may be worthwhile to mention, and is perhaps even a means of attacking the general "getting stuck on a cache hit" problem in a domain that might yield some immediately useful results for a non-negligible number of people. If said application is too obvious, I apologize for stating the obvious.