When is Human–Robot Joint Agency Effective? The Case of Cooperative Reaction Games

Abstract

Here, using a cooperative reaction game, we compared human interaction with an anthropomorphic robot and a computer program. The assistant (either the robot or the program) took over half of the tasks that the participant faced in the game. In half of the game sessions, the assistant reacted slowly, and in the other half, it reacted faster than naive participants. Together with the fast-paced robot, participants scored significantly more points, than with the fast-paced program. This difference was made by the players, as there was no difference in performance between the robot and the program. In addition, human reaction time varied with the pace of the robot, but not the program. We believe that the participants played better with the robot and adopted its pace because they perceived it as a real co-agent. According to the survey, the participants preferred to play with the robot. Together, these results suggest promising prospects for joint human–robot agency and the use of anthropomorphic robots.