Imitation Mechanisms of Social Resonance for Embodied Agents

Humans rely on imitation in social interaction, overtly for learning novel behavior and coordinating interactional characteristics with interlocutors, and internally for grounding the understanding of other‘s behaviors in own experiences. This project develops a probabilistic model of how perception-behavior links create resonances across hierarchical structures, from motor levels to higher (semantic) levels, and how these structure and links are acquired through imitation learning. These mechanisms are applied to the recognition, understanding, and generation of hand-arm gestures for embodied agents.