Interaction with an Autonomous Team Member Determines the Relationship between Team Trust and Team Performance

Abstract

This study examines the effects of individual team members’ trust on Human-Autonomy Team (HAT) and all-human team performance in a Remotely Piloted Aircraft System-Synthetic Task Environment (RPAS-STE). Twenty-one three-member teams consisting of two participants – in the roles of navigator and photographer – teaming with either an autonomous agent or a trained human experimenter – a pilot – flew an RPA with the goal of photographing targets. We regressed a measure of team performance, Target Processing Efficiency (TPE), on each team member’s trust in their team. We found that both team members’ (navigator and photographer) trust in the team predicted TPE, where increases in trust in team predicted increases in team performance. Importantly, we found that the relationship between trust in team and TPE was dependent on the role of the human operator, and team composition, where team composition mediated the relationship for the photographer and moderated the relationship for the navigator. We suggest that heterogeneous interactions with the agent may explain these role-specific differences, which is in line with interactive team cognition’s (ITC) proposition that interaction is a dominant factor in team performance. In sum, roles that have more interaction with an autonomous agent will have their team performance affected more so by the presence of an agent than their trust in the team. To alleviate these differing effects, agents should have communication and coordination capabilities comparable to human teammates. Alternatively, forms of interaction that do not depend on natural language could be pursued.