Limited-Teamwork Autonomy in Training: Transfer Effects on Performance and Coordination in Subsequent All-Human Teams

This study examines how training with autonomous teammates who lack teamwork skills influences performance and coordination when transitioning to all-human teams. Participants operated in a simulated Remotely Piloted Aircraft System environment, first training with either a human or AI pilot before completing a final mission with a human-only team. Results revealed no significant differences in final outcome performance between groups. However, AI-trained teams exhibited reduced process-based efficiency, weaker anticipation, and lower influence dynamics—indicating limited transfer of collaborative skills. These findings highlight the importance of embedding adaptive communication, trust calibration, and interactive team cognition into AI training agents. Practical implications span high-reliability domains such as healthcare, aviation, and defense, where structured AI training must be paired with strategies that foster transferable teamwork capabilities. Future research should explore longitudinal effects and domain-specific applications of HAT-to-HHT training transfer.