This will be a hybrid in-person and remote talk.
Prior literature indicates that the performance of human groups across a variety of tasks can be explained by one common statistical factor termed “collective intelligence” (CI). However, there is an ongoing discussion about the validity of the one-factor-model and prior studies indicate that CI might be better explained by a more complex latent structure. We propose a hierarchical model of collective intelligence introducing collective reasoning, memory and attention as sub-factors of an overarching CI-factor. We test our model on a sample of 22 studies, including 5,279 individuals in 1,356 groups using meta-analysis. The hierarchical model shows a superior fit in comparison to the one-factor model in the complete sample as well as in a variety of subsamples. Furthermore, our results indicate that CI functions differently for established groups vs. newly formed groups as the hierarchical model displays a better fit for established groups. Our subsequent subgroup analysis suggests a higher specialization of CI in established groups as they show higher performances across tasks, higher heterogeneity in performance and lower intercorrelations between tasks compared to newly formed groups. Furthermore, we show that sub-factors of CI are predicted by individual skill and social perceptiveness and discuss implications of these findings. Our results contribute to a nuanced understanding of CI and its predictors across a variety of groups and settings.