Modeling Adaptive Autonomy at the Team Level: Understanding Team-Wide Autonomy and its Impact on Situation Awareness in Human-Autonomy Teams

Abstract

Human-autonomy teams (HATs) are increasingly used in high-stakes environments and depend on maintaining strong situational awareness (SA) across human and AI agents to operate effectively under changing conditions. However, current systems often use fixed levels of autonomy, which can either overwhelm human operators or reduce their engagement, degrade SA and overall team performance. In response, we propose a conceptual model for adaptive autonomy at the team level that dynamically adjusts control distribution between human and AI agents. Using collective team autonomy, the model shifts control based on shared SA and environmental complexity, ensuring the right agent takes the lead when conditions change. It requires humans and AI to contribute to team SA and continuously recalibrates autonomy to balance human adaptability with AI efficiency. By explicitly applying adaptive automation work to agent teams, the model contributes a new way of viewing autonomy as a dynamic, team-shared resource rather than a fixed setting. This approach supports the development of safer and more resilient HATs that can perform reliably in unpredictable, high-risk environments.