- 認知科学 (ISSN:13417924)
- vol.11, no.3, pp.214-227, 2004 (Released:2007-04-13)
A series of analyses on the dynamics of participation in face-to-face multi-party conversations were conducted to demonstrate that the centrality of participation in a conversation is correlated with participants' interest in objects discussed as the topics of the conversation. The correlation can be exploited for Social Navigation in real world human interactive situations.Exploitation of behavioral histories of others is an important strategy for effective problem solving. By accumulating a volume of ‘word of mouth’ information, a number of recommendation systems have been developed and deployed on the basis of this idea to facilitate searching on the Web for objects that match with people's personal tastes. Furthermore, recent advance in ubiquitous computing technologies is making it possible to collect the same types of information in real world settings.This paper focuses its attention on the possible cues for Social Navigation which are available from human interactive conversational behaviors. Based on a dynamic transition model of participation roles in conversations, a hypothesis was presented that the more central a role one plays in a conversation, the more highly interested she is in the topic of conversation. The hypothesis was then tested and confirmed through analyses of human interaction behaviors in experimental poster presentation sessions. The analyses revealed that (a) the frequency of interaction with exhibitors was a good indicator of visitors' interest in the corresponding posters, (b) interest in the posters was proportionate with the time visitors spent being directly addressed by exhibitors rather than as side-participants, and (c) addressee-hood of visitors manifested itself in the coordinated production of verbal backchannel responses.These results suggest that interactional cues for conversational participation structures can effectively be utilized as a measure of human interest. Since collection of conversational behavior cues does not incur any additional burden on users, it has a wide application possibilities for Social Navigation in the ubiquitous society.