Understanding Social Agents in Virtual Space

Abstract. Daily conversation plays an important role in maintaining human relations. But although many researchers have proposed ways to support having daily conversations when in remote sites, none have achieved this successfully yet. We believe that the reason for these failures is that previous studies focused on designing a communication environment that would simulate face-to-face situations (e.g. life-size video projection and multi-user virtual spaces.) We propose an additional approach which consists in placing social software agents as facilitators of human communication. In our virtual space, the social agent approaches people and provides them a topic to talk about when their conversation falters. To understand the effects of using the agent we conducted a series of social psychology experiments which include an intercultural communication experiment between Kyoto University and Stanford University. Our conclusion was that: 1) the agent should have a social identity that is shared by the people, and 2) the agent should be as fair and impartial as possible to prevent people from disuniting. These results show that it is important to include these social aspects when designing software that interacts with human societies