Appropriateness of Ambiguous Facial Expressions Shown by Life-like Agents in Relation to Context of Communication

Abstract

It is known that humans' ambiguous expressions often work effectively to avoid risks latent in interpersonal communication. There have been several attempts to enable agents to use ambiguous expressions as efficiently as humans. However, the ambiguous expressions will not always be appropriate to the circumstances. The agents therefore need to choose the most appropriate expressions that will suit the context. In this paper, to realize agents of this nature, we investigate the relationship between the ambiguities of the agents' facial expressions and their appropriateness to the context. We formulate a hypothesis that the agent's ambiguous facial expression works more effectively when a recipient is in a less positive or less negative context. We verify this hypothesis and show a model of contextual appropriateness of ambiguous facial expressions. In this paper, based on our previous work, we provide more detailed information and discuss it in greater depth to obtain novel insights.