Collective Intelligence

Collective Intelligence

Collective intelligence is a general factor in group performance, analogous to general intelligence in individuals: a group that does well on one kind of task tends to do well on others, and that tendency is measurable. The factor was reported by Anita Woolley and colleagues in 2010. The contested part is not whether it exists but what predicts it, and the mechanism the trade literature leans on rests on an instrument that is itself under serious dispute.

What it is

The original finding is Anita Woolley, Christopher Chabris, Alex Pentland, Nada Hashmi and Thomas Malone's, published in Science in 2010. Across a range of group tasks they found a general collective intelligence factor. It was not strongly correlated with the average or the maximum individual intelligence of members. It was correlated with the average social sensitivity of members, with the equality of conversational turn-taking, and with the proportion of women in the group. Social sensitivity was measured with the Reading the Mind in the Eyes test.

That last detail is where most of the subsequent trouble lives, and it is the part popular retellings reproduce most confidently.

In effect

The replication record has to travel with the finding, and it runs in both directions.

The strongest challenge is a direct one by an independent team. Timothy Bates and Shivani Gupta published three studies in Intelligence in 2017 in which individual intelligence accounted for around eighty per cent of the differences between groups. The hypotheses that group intelligence rises with the number of women and with equality of turn-taking were not supported. Eyes Test performance tracked individual intelligence, and in a combined model it exerted no influence on the group-level factor at all. An earlier comment by Marcus Crede and Garett Howardson in 2011 had argued that the factor might substantially be an artefact of a general factor of personality.

The strongest defence is real and it comes from the original team. Christoph Riedl, Young Ji Kim, Pranav Gupta, Thomas Malone and Anita Woolley published a meta-analytic pooling in PNAS in 2021 covering twenty-two studies and more than five thousand people. The factor held up, predicting performance on out-of-sample criterion tasks, and was most strongly predicted by group collaboration process, then individual skill, then composition. The proportion of women remained a significant predictor, mediated by social perceptiveness. A correction to the statistical reporting was published the following year.

Our verdict: the 2010 finding is real and the general factor survives a large pooled analysis, while the specific mechanism the trade literature depends on is not secure, and the direct independent replication found individual intelligence doing most of the work. Cite the finding only with the replication record attached.

What it does not say

It does not say group intelligence is independent of individual intelligence. That was the headline of the original paper and it is the claim the independent replication contradicted most sharply.

It does not establish that social sensitivity is the key predictor, because the instrument used to measure it is disputed. The Reading the Mind in the Eyes test, from Simon Baron-Cohen and colleagues, has a live construct-validity controversy. A systematic scoping review in Clinical Psychology Review in 2024 found that most studies using it reported no validity evidence on key dimensions, and that where evidence was reported it frequently failed accepted standards. A 2023 exchange in PNAS argues that validating it would require an interpretable factor model it does not have. The live question is whether it measures theory of mind at all or something nearer emotion recognition and vocabulary. Never cite it as a measure of theory of mind without the measurement dispute in the same sentence.

It does not fold those two disputes into one. The question of what predicts the factor and the question of whether the instrument measures what it says are separate arguments with separate evidence.

It is not presented with any of this in the trade account this vault entered it through. Colin Fisher states the 2010 result at pp. 57 and 58 with no caveat, names the Eyes Test without naming its originator, and gives none of the replication record. It is the riskiest evidential bet in an otherwise unusually well-sourced book.

Sources

  1. Woolley, A. W., Chabris, C. F., Pentland, A., Hashmi, N., & Malone, T. W. (2010). Science, 330, 686-688. The original finding.
  2. Bates, T. C., & Gupta, S. (2017). Intelligence. Three studies, 312 people. Individual intelligence accounted for around 80 per cent of between-group differences; the gender composition and turn-taking hypotheses were not supported.
  3. Riedl, C., Kim, Y. J., Gupta, P., Malone, T. W., & Woolley, A. W. (2021). PNAS, 118(21). Twenty-two studies, 5,279 individuals in 1,356 groups. Correction published 2022.
  4. Crede, M., & Howardson, G. (2011). Intelligence. The general-factor-of-personality artefact argument.
  5. Eyes Test construct validity: scoping review, Clinical Psychology Review, 2024, reporting that 63 per cent of studies gave no validity evidence on key dimensions. A 2023 exchange in PNAS on the factor model.
  6. Fisher, C. M. The Collective Edge, pp. 57 and 58. Ingested 2026-09-23. Stated without caveat.