A New Theory of Robot Conflict and Compromise

Conflicts during collaborative decision-making in human-robot teams can arise over both tasking, for example, problem analysis, and teaming issues, for example, dividing responsibilities. This new theory proposes that conflict is shaped by whether the human operator adopts a tool or teammate mental model of the robot. An advanced tool model promotes technical disagreements over task execution algorithms, requiring performance trust in the robot to facilitate compromise. By contrast, a humanlike teammate model attributing Theory of Mind to the robot frames disagreements as personalized narratives; conflict resolution requires relationship trust. Methods for investigating conflict were explored in a project on human-robot teaming for threat assessment. Where the human and the robot disagree on the quantitative weightings of cues to threat, the Brunswik Lens Model identifies sources of disagreement. Alternatively, using immersive simulation and naturalistic dialogue elicits narrative conflicts over interpretation of threat scenarios. The theory can guide system design to mitigate conflict and optimize collaboration.