Can Personality Scores Be Used To Optimize Team Composition? Results From A Hidden Profile Task Using Machine Learning

Teams are not merely the sum of their parts but it has proven challenging to construct teams with the optimal member composition for high performance. In this paper, we take a unique method to studying the relationship between the mix of individual personalities (measured by Big Five dimensions) in a team, and its performance at information sharing and integration (in a hidden profile task). Exploiting random assignment into teams, we use a combination of deductive (hypothesis testing) and inductive (machine learning) methods in an iterative way to understand how team personality profile may affect its performance. We find little support for a priori hypotheses derived from existing literature. In an exploratory stage, we employ machine learning to find models that improve substantially in accuracy of prediction and also produce the insight that team’s extraversion leads to higher performance when the leader is female. Our methodology offers a novel approach to composing effective teams based on information about its members.