Thank you. Will probably take some time to understand it thoroughly.
It looks like the researches task was to prove that if a team does well on one task, there is a high chance that it will do well on another (no shit Sherlock : ) in order to define a parameter they call “group intelligence” (as different from “individual intelligence”).
To do this they needed to prove that the correlation between group’ score in a test compared with score in other tests is higher then correlation between group’s score and intelligence.
They measured intelligence using Raven’s Progressive Matrices, a test that is proven to be done on average better by people with Asperger’s syndrome (i.e. people who are not functioning in groups well). The tests they conducted did not necessary require the same skills as in what they used to measure intelligence.
The first study they conducted proves their hypothesis quite will. Numbers show exactly what they wanted. Unfortunately I could not find any data on percentage of female for the first tests (they have not measure it?).
The second study does not look as good in terms of proving their hypothesis (see table S3b). But the data for this study shows data for percentage of females. During this study it seems that this percentage worked positively on some tests and was negative on other tests, getting correlation with the so-called “collective intelligence” only as a result of different weighting for different tests (i.e. looks a bit questionable). Interestingly it seems “percentage of females” have less correlation with the “collective intelligence” value then “highest intelligence in a group”.
Interestingly when I was reading your original link on this story, the sentence “Before we did the research, we were afraid that collective intelligence would be just the average of all the individual IQs in a group” jarred with me, I immediately thought (based on my experience of teams) that the correlation will be stronger not with “the average of al the individual IQs”, but with the “IQ of a team member with the highest IQ”. Looking at their data it seems I was correct — seems like I had better original hypothesis on the matter then the authors of the study :)
PS Thank you for the link — I enjoyed looking though this paper, this is not a topic I usually read on.