PNNL

Abstract

This paper discusses our recent studies on Embodied Conversational Agent (ECA) design strategies. We attack the problem from two specific directions, that is, via the embodiment that the character ‘wears’ during its relations with the user, and the method of interaction used by the ECA to engage the user in a dialogue. A rigorous literature review lead to a set of hypotheses that were empirically validated. Our results indicate that while users generally prefer to interact with a young character matching their ethnicity, no significant preference were indicated for character gender. For interaction, our results indicated that there were no significant differences in perceived trust between the facial and facial/body treatments (suggesting there is little utility in adding gesture and posture behaviors to an interaction scheme) – although both were rated much higher than the non-trusting treatment. Other interesting results (e.g. people with a high propensity to trust were more trustworthy across treatments, users that interacted with the non-trusting character made more mistakes, disliked the interaction and felt less in control than users in the other treatments) from this work are also discussed.