[Link] Study on Group Intelligence

Full dis­clo­sure: This has already been dis­cussed here, but I see util­ity in bring­ing it up again. Mostly be­cause I only heard about it offline.

The Paper:

Some re­searchers were in­ter­ested if, in the same way that there’s a gen­eral in­tel­li­gence g that seems to pre­dict com­pe­tence in a wide va­ri­ety of tasks, there is a group in­tel­li­gence c that could do the same. You can read their pa­per here.

Their ab­stract:

Psy­chol­o­gists have re­peat­edly shown that a sin­gle statis­ti­cal fac­tor—of­ten called “gen­eral in­tel­li­gence”—emerges from the cor­re­la­tions among peo­ple’s perfor­mance on a wide va­ri­ety of cog­ni­tive tasks. But no one has sys­tem­at­i­cally ex­am­ined whether a similar kind of “col­lec­tive in­tel­li­gence” ex­ists for groups of peo­ple. In two stud­ies with 699 peo­ple, work­ing in groups of two to five, we find con­verg­ing ev­i­dence of a gen­eral col­lec­tive in­tel­li­gence fac­tor that ex­plains a group’s perfor­mance on a wide va­ri­ety of tasks. This “c fac­tor” is not strongly cor­re­lated with the av­er­age or max­i­mum in­di­vi­d­ual in­tel­li­gence of group mem­bers but is cor­re­lated with the av­er­age so­cial sen­si­tivity of group mem­bers, the equal­ity in dis­tri­bu­tion of con­ver­sa­tional turn-tak­ing, and the pro­por­tion of fe­males in the group.

Ba­si­cally, groups with higher so­cial sen­si­tivity, equal­ity in con­ver­sa­tional turn-tak­ing, and pro­por­tion of fe­males are col­lec­tively more in­tel­li­gent. On top of that, those effects trump out things like av­er­age IQ or even max IQ.

I the­o­rize that pro­por­tion of fe­males mostly works as a proxy for so­cial sen­si­tivity and turn-tak­ing, and the au­thors spec­u­late the same.

Some thoughts:

What does this mean for Less Wrong?

The most im­por­tant part of the study, IMO, is that “so­cial sen­si­tivity” (mea­sured by a test where you try and dis­cern emo­tional states from some­one’s eyes) is such a stronger pre­dic­tor of group in­tel­li­gence. It prob­a­bly helps peo­ple to gauge other peo­ple’s com­pre­hen­sion, but based on the fact that peo­ple shar­ing talk­ing time more equally also helps, I would spec­u­late that an­other chunk of its use­ful­ness comes from be­ing able to tell if other peo­ple want to talk, or think that there’s some­thing rele­vant to be said.

One thing that I find in­ter­est­ing in the meatspace mee­tups is how in new groups, con­ver­sa­tion tends to be dom­i­nated by the peo­ple who talk the loud­est and most in­sis­tently. Often, those peo­ple are also fairly in­ter­est­ing. How­ever, I pre­fer the cur­rent, older DC group to the newer one, and there’s much more equal time speak­ing. Even though this means that I don’t talk as much. Most other peo­ple seem to share similar sen­ti­ments, to the point that at one early meetup it was ex­plic­itly voted to be true that most peo­ple would rather talk more.


Any­thing we should try do­ing about this? I will hold off on propos­ing solu­tions for now, but this sec­tion will get filled in some­time.