A Preliminary Exploration to Enhance Appropriate Level of Trust in Autonomous Vehicle and Pedestrian Interaction

Abstract

As autonomous vehicles (AVs) are increasingly piloted in urban environments, establishing appropriate levels of pedestrian trust is critical for ensuring both safety and traffic efficiency. This study examines the influence of external human-machine interfaces (eHMIs) on pedestrian trust during AV encounters at unsignalized crosswalks. A virtual reality (VR) experiment involving 50 participants was conducted to assess how variations in eHMI timing and vehicle behavior affect trust calibration. The experimental design included a between-subject factor comparing failure scenarios with immediate versus delayed eHMI display, and a within-subject factor encompassing seven vehicle behavior conditions. Trust ratings were collected after each scenario and analyzed using a linear mixed-effects model. Results show that the timing and consistency of eHMI signals significantly influence pedestrian trust. Notably, delayed eHMI messages led to more cautious trust responses in failure scenarios, while still maintaining high trust in successful yielding conditions. These findings suggest that strategically delaying eHMI communication may mitigate overtrust and support safer pedestrian-AV interactions. The study contributes preliminary evidence toward designing trust-calibrated AV communication systems.