Robot Model in a Group with Human Participants

Communication Robots

In recent years, there has been an increasing effort to develop communication robots, for example Pepper and RoBoHoN. Their purpose is to communicate with people naturally by expressing emotions and behaving like living creatures.

Forming a Group Composed of Human and Robots

Communication robots require to not only communicate with person one on one but behave cooperatively in a group including human and robots. Today, general communication robots communicate with person one on one. They hardly take into account interactions in a group with human and robots.
In general, when a group is formed for human to communicate with each other, we regard the group as society. To take it into account, communication robots need to have a sociability to behave cooperatively in a group. Therefore, we aim to make a robot model to behave cooperatively in a group for robots in order to join human group.

“When in Rome, Do as the Romans Do”

When human forms a group, they also form group norms. Group norms are the often-unwritten rules by which groups know how to behave in a range of situations. Unlike social norms, for example moral and ethics, group norms depend on group members’ identities and way of thinking.
“When in Rome, Do as the Romans Do”, as the saying goes. When a person joins a group, s/he needs to obey group norms, for example rules and customs, in the group not completely but a little. In the same way as human, when a robot joins a human group, it needs to obey group norms in the group by learning the group norms. Furthermore, when human recognizes robots who make an effort to join a human group, it appears that human in the group accepts such robots with higher probability.
Therefore, we aim to make a robot model to recognize group norms in a group by learning human’s behaviors for robots in order to join human group.

Conclusion

We proposed a group norm model and made simulations and experiments, including human. Further studies are required to reveal whether a robot, using this model, learn a suitable behavior in a human group.