Aim: In this study, we aimed to investigate pedestrians’ trust, crossing behaviour, and misuse of an automated vehicle (AV) equipped with an external human-machine interface (eHMI). We hypothesized that participants’ trust would drop to a low level when the eHMI fails (i.e., the eHMI turns on, but the vehicle does not brake), followed by a sustained low of trust in subsequent crossing trials. Furthermore, misuse of the eHMI was expected, operationalized as a situation where participants who are exposed to the eHMI signal before the AV starts to decelerate (-1 group), would start to use the information from the eHMI and cross early instead of relying on the implicit communication from the vehicle. Finally, it was expected that participants who are exposed to the eHMI signals after the AV began to decelerate (+1 group) would be run over by the AV less often when the eHMI fails as compared to participants in the -1 group.  
Literature: Prior studies showed that pedestrians tend to overtrust eHMIs, and when there is a malfunction of the eHMI, pedestrians’ behavior changed significantly in terms of showing hesitation, longer decision times, and rating an increased risk and reduced trust. However, there is no consensus on the evaluation of pedestrians’ overtrust and subsequent misuse of eHMI at different eHMI onset timings. Methods: We conducted this study at the University of Leeds CAVE simulator. Sixty participants encountered an AV in a typical Britain street virtual environment. The AV was fitted with an eHMI in the form of a 360° light band on the top and sides of the AV and additionally on the grill. The eHMI indicated that the AV has started yielding or is yet to yield for them. All participants experienced fifty trials during the entire experiment. Each trial had a variation of the independent variables: (1) behavior of the AV: Yielding with an onset of braking at 33 m and 43 m from the participant, and no yielding, (2) eHMI state: On/Off, (3) eHMI onset timing: early onset (-1 s) or late onset (+1 s). One half (N = 30) of the participants experienced the early eHMI onset timing (-1 s) where the eHMI turned on 1 s before the vehicle started braking. The other half of the participants (N = 30) experienced the late eHMI onset timing (+1 s) where the eHMI turned on 1 s after the vehicle started braking. After each trial, participants rated three subjective measures (perceived risk, comprehension of the eHMI, and trust in the eHMI) on a scale of 1 (strongly agree) to 10 (strongly disagree). The objective measure was the position of the participant in the simulator.   Results: Participants trusted the AV more with the eHMI than without the eHMI. Furthermore, participants in the -1 group crossed the road considerably earlier than participants in the +1 group. When an eHMI failure occurred, the trust levels and the comprehension of the AV in both groups dropped significantly, whereas self-reported risk rose to high levels. After the first eHMI failure experience, the -1 group trusted the AV with eHMI on significantly less compared to their trust before the failure. Additionally, the results showed that about 30% of the participants in both groups crashed with the AV when the eHMI failed for the first time. There were no significant differences in the number of crashes between the two groups.
Conclusion: Pedestrians’ trust considerably reduces when the eHMI fails. Also, there is a sustained loss of trust in the AV’s behavior after experiencing the failure in the eHMI. When the eHMI provides early information (before the implicit communication from the AV), pedestrians cross early as compared to the eHMI providing late information (after the implicit communication from the AV). Pedestrians who repeatedly encountered either the ‘early’ or ‘late’ onset eHMI crashed with the AV in similar numbers when the eHMI failed for the first time. However, the -1 group managed to avoid crashes when the eHMI failed for the second time. Application: This study is useful in considering the safety of pedestrians who interact with an eHMI-equipped AV.