Generating Effective Social Behaviors for Robots

Social interaction involves a large number of patterned be-haviors that people choose and combine to achieve partic-ular communicative goals. To achieve fluent and effective human-robot communication, robots must seamlessly inte-grate various social behaviors for a given interaction context. However, very little is known about how robots might be equipped with a collection of such behaviors and how they might employ these behaviors in social interaction. I pro-pose a framework that systematically uses behavioral social-scientific findings and contextualizes these findings in an Activity-Theory-based interaction model to guide the gener-ation of social behaviors for humanlike robots. I have imple-mented the Robot Behavior Toolkit, an open-source imple-mentation of the proposed framework as a Robot Operating System (ROS) module. An evaluation of its effectiveness in generating social behaviors in two human-robot interaction scenarios showed that, compared with three different base-lines, behaviors generated by the toolkit achieve the social, cognitive, and task outcomes as predicted by the literature such as improved perceived competence, information recall, and task completion time. I will conclude with a plan of us-ing the toolkit to generate effective behaviors for educational robots