Measuring Collective Intelligence in Groups: A Reply to Credé and Howardson

Abstract

Recent work by Woolley, Chabris, Pentland, Hashmi, & Malone and their colleagues finds evidence for a general collective intelligence factor that predicts a group’s performance on a wide variety of tasks, like the general intelligence factor does for individuals. Credé and Howardson argue that there is not yet sufficient evidence to conclude such a collective intelligence factor exists. Specifically, C&H suggest that the general factor is not strongly enough correlated with all the tasks examined, but we point out problems with their interpretation of the evidence presented. C&H also suggest that the effort and abilities of the group members are “statistical artifacts” that may have inflated the correlations observed, but we demonstrate that the data do not support that interpretation. We concur with C&H about the importance of improving measures of collective intelligence, and we elaborate upon their suggestions for future research to better understand the boundaries and the causal mechanisms of this phenomenon.

Keywords: collective intelligence, group performance, construct validity