There’s a well-known 2010 finding that collective intelligence barely depends on how smart the individual members are. Instead, the original study pointed to things like the number of women in the group and how evenly people took turns speaking, with the strongest predictor being something called “Reading the Mind in the Eyes” (RME), which is basically a measure of social/emotional perceptiveness. This paper is a direct attempt to replicate that model across three new studies (312 people total), and it comes to the opposite conclusion: individual IQ explained around 80% of the differences in group performance, and when the data from two studies were combined into a more rigorous statistical model, that number effectively became 100%. So smart groups aren’t smart because of gender balance or turn-taking or empathy; they’re smart because they’re made of smart people.
The paper is also careful about why it gets such a different answer than the original study. The authors argue it’s partly a methods issue: earlier work relied on regression, which can miss how multiple variables move together, while this paper uses structural equation modeling (SEM) to separate a general “group-IQ” factor from effects tied to just one specific task. Under that lens, the modest link between empathizing and group performance in the original study shrinks to a narrow effect on a single task (a word-completion exercise), not on group intelligence as a whole. The authors also push back directly on peer-review objections raised about the work — including claims of low statistical power and requests to strike parts of the study — laying out why they think their design was adequately powered to detect the very effects it was testing against.
It’s a good reminder that a single high-profile finding can take years and several failed replications to properly stress-test.
Link to full article: http://dx.doi.org/10.1016/j.intell.2016.11.004