How Multi-Party Conversation Can Become an Effective Interface While Driving

Abstract

This paper proposes a unique communication approach for an in-car driving agent system by utilizing the usage of a multi-party conversation mechanism. We developed a social interface named NAMIDA that incorporates three conversational robots that can decrease the number of directed utterances towards a driver through a turn-taking process among the robots. Through this model, the driver can gain necessary location-based information without joining the conversation. The results of our pilot study revealed that the proposed multi-party conversation based interaction model is more effective in alleviating certain workload factors for drivers compared to a conventional one-to-one communication based approach that directly addresses the driver. Moreover, an analysis of the attention behaviors of drivers showed that the proposed approach could encourage drivers to focus on the road better than that of a one-to-one communication based system. Finally, the results of a subjective impression showed that the multi-party conversation based system seemed more autonomous and was more animated; it also demonstrated more natural conversation.