Abstract: Research on how personality contributes to team performance has long assumed that teams benefit either from high mean levels of key traits (elevation) or from heterogeneity among their members (variability). However, empirical findings generally reveal only small effect sizes. Here we discuss potential methodological problems as well as theoretical implications not sufficiently heeded so far. Based on this, the present study examined whether team performance in a specific team task could be predicted from trait measures, repeated assessments of personality states, and situation-perception ratings. Fifty-one unacquainted teams of three to four members met online, worked on a task and completed trait, state, and situation-perception measures. Predictors were aggregated using elevation, relative variability, minimum, and maximum values. We compared a mean-prediction baseline, ridge regression, elastic net, random forests, and tuned random forests under nested fivefold cross-validation repeated 100 times and used interpretable machine learning (IML) techniques to interpret selected models. Across 17, theoretically derived feature sets, predictive accuracy was modest at best, with correlations ranging from negative to r = .25 ( SD ≈ .26 across 500 resampling iterations). The findings support an interactionist and situational account of momentary team performance and highlight the limited predictive value of personality trait compositions for momentary team performance.
Ziegler et al. (Tue,) studied this question.