Psychologist have long-recognized a general intelligence factor, known as ‘g‘, that correlates individual performance on cognitive tasks (how a person does on one test strongly predicts his/her performance on another test), and they have even studied how average intelligence of group members predicts group performance, but the idea that groups might have their own collective intelligence (‘c’) was virtually ignored.
Recently, two tandem studies conducted by Woolley et al (Science, 29 October, 2010*) have revealed strong evidence for ‘c’, a collective intelligence factor. Working with almost 700 individuals organized into groups of between 2 and 5 member each, the researchers put the groups through their cognitive paces, giving them a wide range of cognitive tasks to perform.
Tasks in the first study included: solving visual puzzles, brainstorming, making collective moral judgments, and negotiating over limited resources. As a “task criterion” (to establish a group baseline performance), each test session ended with a game of checkers between each group and a computer opponent. All individuals were given standardized intelligence test prior to conducting the study.
Various statistical methods were used to analyze the sociometrics; the results consistently yielded one factor that accounted for over 40% of the groups’ variance (the main variable being measured) in performance. This emergent, collective factor was used successfully to predict group performance on other tasks. In statistical analyses that seek to “flesh out” a causal factor, this is a big result. As a comparison, an analysis of the average and maximum intelligence scores found that these did not significantly correlate with test scores, nor were they very good at predicting “criterion task performance”.
As a control, an architectural design problem was given to 63 of the individual participants. In this case, individual intelligence was indeed a significant predictor of performance on the task. But the researchers discovered that something else is a more powerful predictor of outcome when an individual joins a task-oriented group.
A regression analysis using c and average individual intelligence confirmed this predictive correlation between c and future test performance, but not for the average intelligence score–although the researchers acknowledge a “moderate” correlation, in some tests, between average intelligence and maximum intelligence scores, and task performance.
In the second study, working with over 150 groups of 2 to 5 members each, researchers sought to replicate the results. Results once again provided confirmation of the existence of a “first factor” (the putative c factor), that accounted for over 30% of the performance score variance. The researchers concluded that “a one-factor model is the best fit for the data from both studies.”
So then, how is this mysterious “group intelligence factor” accounted for, that is, what is causing it? Woolley et al were able to isolate three key factors in emergent group intelligence: 1) social sensitivity (as measured by the “Reading the Mind in the Eyes” test), 2) the variation in speaking turns (as measured by ‘sociometric badges’ worn by a subset of the groups), and, 3) the proportion of females in the group. The second factor was found to be negatively correlated with c (the more one or two members dominated the group, the lower the group’s collective intelligence). The last factor appears to be “largely mediated by social sensitivity because (consistent with previous research) women in our sample scored better on the social sensitivity measure than men.”
The researchers state that “these results provide substantial evidence for the existence of c in groups, analogous to a well-known similar ability in individuals. Notably,this collective intelligence factor appears to depend both on the composition of the group (average member intelligence), and on factors that emerge from the way group members interact when they are assembled (e.g, their conversational turn-taking behavior).”
The research team** wonders if this information could be used to raise the intelligence of groups — a far easier (faster) task than raising individual intelligence. By measuring the effects of specific interventions on c, researchers will be able to predict the effects of those interventions on a variety of tasks. This in turn will contribute to the establishment of a new “science of collective performance”.
* The paper abstract: ‘Evidence for a Collective Intelligence Factor in the Performance of Human Groups‘ (10.1126/science.1193147)
** authors: Anita Williams Woolley, Christopher F. Chabris, Alex Pentland, Nada Hashmi, Thomas W. Malone
Top diagram (types of collective intelligence): Generozo; CC – by – sa 3.0
Data charts: from SOM, the authors, cited paper (above)