Leaders or Team-Mates: Exploring the Role-Based Relationship Between Multiple Intelligent Agents in Driving Scenarios - Research on the Role-Based Relationship Between Multiple Intelligent Agents in Driving Scenarios

Shuo Li, Xiang Yuan, Xinyuan Zhao, Shirao Yang

Abstract: Intelligent agents (IAs) are increasingly used in vehicles and associated services (e.g. navigation, entertainment) to enhance user experience, as IAs were applied to the car and turned the vehicle into a service platform under the rapid development of the intellectualized and connected vehicle. However, various IAs may be employed by other services and devices. In the case of in-vehicle cross-device interaction, when users interact simultaneously with multiple services or devices, the actions and decisions of one IA may conflict with those of others. This paper presents a role-based relationship framework to resolve potential conflicts between different IAs in the driving scenarios. The article discusses four types of IA relationships: Partnership, Representative, Subordinate, and Co-embodiment. To examine people's perceptions and attitudes towards different types of relationships, we apply an evaluation system and conduct user studies (N = 30). In two scenarios (Navigation Plan & Music Switching), Participants are required to engage in conversations with IAs based on various types of relationships. Data analysis and user interviews show that Partnership is gaining popularity in leisure and entertainment settings. Moreover, Representative is more effective in efficiency-oriented use cases. In addition, the research on driver's attention behavior suggests that Representatives can convince the driver to focus on the road more efficiently in navigation scenarios than in music settings. After evaluating the different role-based relationships of IAs, design recommendations for user interactions with multiple IAs in driving scenarios are offered.