Two Sides of the Same Coin? Joint Perspectives From Shared Mental Models and Interactive Team Cognition Theories on Human-AI Team Cognition

Abstract

The ubiquity of AI has made team cognitive science a foundation for understanding what increasingly complex human-AI interactions can achieve. Different models and theoretical frameworks from human teamwork literature offer varying explanations for what human-AI team (HAT) cognition entails. Predominant theoretical frameworks founded on the Shared Mental Model theory frame team cognition in terms of overlaps between team members’ knowledge structures that help them understand their task context. These techniques remain widely used to model team performance in knowledge-intensive tasks; yet, conceptual and methodological concerns about their applicability in HATs have prompted the use of other frameworks. Among these is Interactive Team Cognition theory, which posits that team interaction is team cognition: a team-level activity observable in real-time. This paper sought to identify key intersections and frictions between these two theories, motivating a panel discussion on the feasibility of a dual-perspective approach to addressing open research questions about HAT cognition.