Computational gesture research

摘要

Abstract Embodied artificial agents, like humanoid robots or virtual characters, can produce a variety of co-speech gestures in interactive settings. This enables a computational branch of gesture research that offers opportunities for (1) investigating production and comprehension processes through computational cognitive modeling, and (2) studying systematically the effects and functions of gesturing in human-agent interaction. In this chapter we review current approaches to synthesize gestures, and we discuss in detail findings from experiments on the effects of gestures produced by virtual characters or robots. Current evidence suggests that synthetic gesturing has considerable effects on how an artificial agent is perceived and attributed with human-like properties. In contrast, so far, there is little evidence showing that contemporary synthetic gesturing helps human addressees to better understand or retain the information communicated by an agent. We discuss what these findings imply for the potential functions that gesture may serve in improving human-agent interaction, and more generally for our understanding of gesture through computational modeling efforts.