Trust Dynamics in Human-AV (Automated Vehicle) Interaction

Despite many benefits of automated driving, such as reducing fuel consumption, traffic congestion and crashes, a lack of trust hinders the adoption of automated vehicles (AVs). Prior research focused on people's trust in AVs based on AVs' overall performance. The present study is focused on people's trust change in AVs over time in a sequential decision making context. We conducted a human-in-the-loop experiment with 16 participants in a virtual 3D environment wherein participants acted as passengers riding an AV. We manipulated two independent variables: level of stochasticity (high vs. low) and source of stochasticity (external vs. internal). Dependent variables included participants' moment-to-moment trust in AVs and post-experiment trust. Our results revealed that when the stochasticity was due to internal errors (e.g. AV's sensor errors) as compared to external errors (e.g. traffic jams or road blocks), participants' trust in AVs decreased more significantly. Also, the larger the cost due to an error, the larger the trust decrement.