A robot model in limited scenarios to create a suitable decision-making criterion by interacting with people in a group

In this study, we propose a model to enable robots to create a suitable criterion for decision-making by interacting with people in a group. Using this model, a system learns a suitable criterion for the group as a group member itself through Q-learning. When people, who have different individual decision-making criteria form a group, they adjust their criteria to a common criterion for the group. This common criterion is suitable for the group. People and the system cooperatively answer easy quizzes that have vague questions. The present study investigates whether the system can make a suitable decision-making criterion in a group by learning from interactions, that is, whether its answer converges to a certain criterion. Our experiments reveal that a group consisting of the participants and the system forms a common criterion in a limited scenario. Moreover, it is likely that robots using the model will be able to adjust their decision-making criteria for human groups based on interaction with them. However, further study is required to reveal whether robot's answers affect people and whether participants create a suitable criterion taking the robot's answer into account even if the participants recognize that they are interacting in a group that includes a real robot.