Human-agent team formation: An empirical study

This paper studies the decision-making strategies people deploy when they interact in groups comprising both human and computer agents. It focuses on settings where actions are occurring at a fast pace and decisions must be made within tightly constrained time frames, under conditions of uncertainty and partial information. Participants can form teams by negotiating over splitting rewards that they can obtain by successfully completing joint tasks. However, members may defect from their teams by accepting offers from non-team members. In this paper the authors present a behavioral study that measures the extent to which social factors such as trust and fairness affect people's commitments to their teams when these include both computer agents and humans. Results show that people offer significantly less benefit to agents than to people when forming teams. However, people are as loyal to agent-led teams as they are to human-led teams. These results have implications for the design of decision-making strategies in mixed-initiative systems for building trust and cooperative relationships between humans and agents.