Enhancing decision-making in human-machine teams

One of the key advantages of groups is their ability to pool resources and share information effectively, enabling coordinated efforts that result in emergent behaviors and enhanced performance. Here we present preliminary data from two experiments that are part of a larger study aimed at developing and testing a prototype system of a 'superorganism' designed to augment decision-making within human-AI teams. One key aspect of the system, is the collaboration between humans and AI agents (AIs). In each team, humans and AIs make decisions as peers, and the optimality of each team decision depends, among other things, on mutual trust and personality, of both the human and AIs. We manipulated team composition and AIs' personality and investigated how these, and human trust in AI, affect individual and team performance. Our preliminary results suggest complex dynamics. There was a clear advantage for groups, compared to individual decision-making, as expected. However, even though humans trusted the 'humanized' AI more than the non-humanized AI, performance did not necessarily benefit from it.