Abstract
An online, video-based, within-subjects experiment was conducted to investigate the effects of three different automation braking profiles and three in-vehicle display designs on drivers’ trust and reliance on automated braking systems. 36 participants watched videos of automated braking at intersections (with different braking times and displays) and rated their trust in the automation and comfort with the automation’s braking time. They then watched the same videos and were asked to press the space bar and pause the video if they believed they should take over from the automation and brake. The braking profile had a significant effect on the trust and comfort ratings, while both the braking profile and display had a significant effect on the takeover time. The findings provide insights on drivers’ trust and reliance on automation that is designed to avoid lateral collision hazards at intersections.
Intersections are a probable location of safety critical events due to lateral collision hazards (Erazo & Campbell, 2021). Vehicle automation can relieve drivers from some vehicle control tasks and provide support during such critical events. The behavior of the automation, e.g., how it brakes at an intersection, has been found to be associated with driver trust in the automation (e.g., Lee, Liu, Domeyer, & DinparastDjadid, 2021). Communicating information to drivers via a display is also crucial during such critical events, and there is a need for investigating what information should be communicated.
An online video-based study was conducted with 36 fully licensed participants aged 30-45 and residing in Canada. Participants watched first-person video clips of simulated drives which ended with an incursion scenario at an intersection with an intruder vehicle coming in from the right. Participants were told their vehicle was equipped with automation that controls the vehicle’s speed and steering, and that it would apply the brakes when the other vehicle approached from the right. Two independent within-subjects variables were manipulated in this experiment: the automation braking profile varied through braking time (early, on time, too late), and the in-vehicle display (braking alert, hazard detection alert, and combined display). The conditions were fully counterbalanced with blocking on the display. In the “early” braking condition, the time-to-impact at braking initiation was 2.5s, in the “on time” condition it was 1.5s, and in the “too late” condition it was 0.5s (collision would not be avoided). In each video, one of the three displays was projected onto the windshield. The braking alert indicated that the vehicle was about to brake and appeared 1 second before the brakes were applied. The hazard detection alert indicated that the automation detected a potentially hazardous situation and appeared 3 s before the brakes were applied. The combined display was a combination of the 2 alerts appearing at their respective times (at 1 s before braking, the hazard detection alert disappeared and was replaced by the braking alert).
In the first phase of the study, the videos all stopped as soon as the brakes were applied, which did not allow the participants to see how the braking scenario played out (i.e., if a collision with the intruder vehicle would occur). After watching each video, participants rated their comfort with the braking time and their trust in the automation using items adapted from the Situational Trust Scale for Automated Driving (Holthausen, Wintersberger, Walker, & Riener, 2020). After each display block, participants were asked to evaluate the displays using items adapted from the System Usability Scale (SUS; Brooke, 1996) and were then asked verbal follow-up questions to gather more feedback. In the second phase of the study, participants watched the same videos, but the videos were not stopped at the braking initiation point. In this phase, participants were asked to watch the full videos and press the space bar to pause the video if or when they believed they needed to take over control from the automation and start braking themselves. If they paused the video, the experimenter would collect the timestamp at which the video was paused, which allowed us to calculate the takeover time.
Repeated-measures ANOVAs were conducted on the average trust score, the comfort rating, takeover time, and SUS scores. There were significant main effects of braking time on the trust (F(2, 70) = 30.46, p < .001
Overall, participants rated the usability of all three displays highly (average SUS score around 4/5). There were no statistically significant differences between the displays; however, the braking alert display was rated the highest, while the combined display was rated the lowest. When asked about their thoughts about the display after each display block, some participants expressed that the two-stage alert presented in the combined display was unnecessary. Overall, the usability assessment suggests that participants prefer being provided with a display, but potentially due to the rapid progression of the braking scenarios at the intersection, concise communication was prioritized over the content of the alert.
