Abstract
In this study, we verify the effectiveness of a sympathy expression model for a bystander robot that we have proposed. Because the social nature of a robot is important for communicating with multiple people, we focus on a robot that is in a bystander position, which is unrelated to the communication between participants. In our proposed model, the bystander can express emotions cooperatively in a group by learning the expressiveness of others. To test this, we assume an interaction of three robots using a neural network-based emotion generation model. The bystander learns moral emotions on the basis of its emotions from the communication between a speaker and an addressee. We verify the effectiveness of the model using a psychological model to evaluate the bystander from the position of the speaker.
