AgentSalon: facilitating face-to-face knowledge exchange by conversations of personal agents

摘要

This paper introduces a system called AgentSalon, intended to promote the acquaintance and dialogue of people meeting face to face at a place where people share an interest, a hobby, and a situation, such as a museum, the street, a school, an office, or a conference event. AgentSalon has a large display, which can be shared by multiple users. The personal agents belonging to each user are displayed as character animations, which start the conversation automatically. The personal agent usually operates on the portable guide system called PalmGuide, and provides a personal guide service, while managing the personal interests of the user and the behavior history. When the user connects PalmGuide to AgentSalon by infrared, the personal agent is transferred to AgentSalon together with the personal data. Then the agents start the conversation automatically. The content of the conversation is the interests and the behavior history of the users. The agents exchange views based on experience or make recommendations representing the user. The users listening to the conversation find common topics of interest, as if their minds are connected. Thus, a more valuable conversation should be realized, as if absorbed into the conversation of the agents. In this study, AgentSalon is implemented as a service to the participants of an actual conference. The principles of operation and evaluation are presented.