When in Doubt, Agree with the Robot? Effects of Team Size and Agent Teammate Influence on Team Decision-Making in a Gambling Task

Abstract

Human-machine teaming is expected to provide substantive benefits to team performance; however, introduction of machine agents will also impact teamwork. Agents are likely to exert substantial influence on team dynamics, even if they possess only limited abilities to engage in teamwork processes. This influence may be mitigated by team size and experience with the agent. The purpose of this experiment was to investigate the influence of an agent on team processes in a team consensus gambling task. Teams were either two or three humans and a machine agent. Participants completed fifty rounds of a gambling task, similar to the game roulette. In each round, team members entered their belief about what the next round outcome would be, a proposed wager, and how confident they were. The machine agent also made a suggestion regarding outcome and wager, but its accuracy was fairly low. The human team members then had to come to a consensus regarding outcome and wager. Overall, the agent exerted significant influence on team decision making, wagering, and confidence. Contrary to initial predictions, team size had only a modest effect, mostly on confidence ratings. Experience with the agent also did not have much effect on the agent’s influence, even as the team was able to observe that the agent’s accuracy was low. These results suggest that machine agents are likely to exert significant influence on team processes, even when they possess limited abilities to engage in teamwork.