Soldier–AI Integration: AI Trust and Teaming Metrics

Author(s):

Kimberly Pollard, Shan G. Lakhmani, Murat Kucukosmanoglu, Cheryl Giammanco, Samantha K. Berg, and Andrea Krausman

Abstract:

In support of the Army Command and Control Modernization Priority, ARL is developing automated and AI-enabled technologies, including large language models and adaptive machine learning algorithms to enable faster and more informed decision-making across echelons. Soldiers work in teams with other humans and with automated systems and intelligent agents. It is critical, therefore, to understand how different AI-enabled systems affect team processes and team and user states, such as trust, cohesion, workload, and team adaptation. Advanced AI systems differ in meaningful ways from previous forms of autonomy, raising the question of whether existing methods of measuring team processes and user states are sufficient for use with these new AI-enabled teams. To address this question, we performed a non-exhaustive literature review to uncover what metrics are successfully being used to measure team processes and user states in AI-enabled teams. We found similar methods being used in the literature as have been used with less advanced autonomy, as well as some opportunities to effectively harness human–human measures of team states and processes with AI’s new capabilities.