Building of turn-taking avatars that express utterance attitudes: A social scientific approach to behavioral design of conversational agents

Abstract

In everyday communication, humans comprehend the attitudes of others conveyed via nonverbal behavior, such as facial expression, body posture and gaze behavior. In this paper, we describe a model for comprehending participants' desire to start to speak or to listen based on nonverbal behavior during conversation. We use a social scientific approach that is based on both an analysis of a video observation and an experiment using avatars. We explain the building of the model. We discuss detecting participant's attitudes using computer vision and the expression of their attitudes using their avatars' facial expressions and body postures. © 2011 Springer-Verlag.