Physiological Synchrony as a Predictor of Team Performance: A Machine Learning Approach

Abstract

As human-AI collaboration becomes more common, robust metrics for evaluating team performance are essential. This study examines whether physiological synchrony can predict team effectiveness using machine learning (ML) models. Dyads completed a mission-planning task while multimodal physiological data (e.g., HRV, respiration, fNIRS) and communication were collected. We extracted both individual and interactional-level features (e.g., synchrony, coherence) and trained ML models to classify team performance. Logistic Regression and SVM models achieved up to 96% accuracy when including interactional synchrony features, outperforming models using individual data alone. Key predictors included breathing synchrony and oxygenated hemoglobin coherence. These findings demonstrate the added value of modeling physiological coupling to understand emergent team dynamics. Future work will expand the sample size and incorporate team-level recurrence and entropy metrics to better capture collective coordination. This approach offers a pathway toward real-time performance monitoring and adaptive interventions in high-stakes collaborative environments.