Effects of Multi-IAs on Subjective Perceptions in Autonomous Driving with Critical Events

In-vehicle intelligent agents (IVIAs) are increasingly being integrated into autonomous vehicles (AVs) to support natural and context-aware interactions. However, little is known about how the configuration of IVIAs—such as using multiple specialized agents versus a single unified one—affects subjective perceptions in complex driving contexts. This study investigated how different configurations of IVIAs affect subjective perceptions in AVs during critical driving events. Two IVIA systems were designed: a single-agent system and a multi-agent system comprising a Daily IVA (DIA) and an Autonomous Driving IVA (ADIA). Sixty participants experienced one of the agent types under either normal or critical event conditions in a Level 5 driving simulator. Subjective perceptions were assessed through subjective ratings (e.g., trust, safety, workload), gaze behavior, and post-experiment interviews. Results showed that the multi-agent system enhanced subjective perceptions compared to the single-agent system, particularly under critical events. These findings provide practical implications for the design of context-aware IVIA systems in AVs.