Analysis of bidirectional information transfer on drumming ensemble using robotic session system

Abstract

For the design of artificial systems with the ability to co-create with humans, it is necessary to consider how robots interpret and apply its human partner's actions in its own actions. Nowadays many robots with the ability to cocreate with human appeared but few research succeeded in quantifying the interaction between humans and robots, making it difficult to assess the importance of the robot's contribution to the interaction. This paper proposes one way to quantify a robot's contribution to a drumming session with a human partner. In this research, we focused on imitation rate (how much the robot applied its human partner's drumming patterns) during drumming session and evaluated how the person's devotion to the interaction changed according to imitation rate. For our experiment, we set up a simple drumming session with a robot and human. Our robotic system could control imitation rate and actually play with human. We conducted a session experiment and analyzed drumming data from the viewpoint of an information transfer. In order to access the person's reaction to having his or her past actions imitated by the robot, we introduced an index named Devotion Rate (DR), which is defined by dividing 'agent to human' information flow by 'human to agent' flow in each phrase.By analyzing information transfer, we founded that the flow from human to agent and agent to human both increased with imitation rate, and that the relationship of the flows was proportional. But DR showed non-linear relationships and specific score in 60% to 80% imitation. Our findings indicate that cognitive function of interaction can be analyzed not only by one-way correlation but by bidirectional information transfer including interaction history.