Trust rises, attention falls: divergent effects of exposure and education in driving automation

Introduction Drivers supervising Level 2 automation must maintain situation awareness while the system controls steering and speed. Miscalibrated trust can contribute to overreliance and lapses in monitoring, whereas insufficient trust leads to disuse. Prolonged supervision is associated with increased mind-wandering, which can slow reactions to critical events. This study tested whether brief educational interventions affect trust, attention, and takeover readiness during Level 2 driving. Our focus on brief interventions reflects the short, time-constrained onboarding that drivers typically receive when adopting driving automation systems. Methods Fifty-five licensed drivers with no prior hands-on experience of Level 2 automation completed a 15-min automated highway drive. Participants received either minimal instruction (Basic), capability-focused education (Knowledge-based), or limitation-focused education (Rule-based). Trust was measured at four time points; additional measures captured self-reported mind-wandering, gaze behavior, and takeover reaction time. Results Trust increased significantly over time in all groups, and educational framing did not alter this trajectory. Capability-focused education enhanced monitoring of the human-machine interface on two false discovery rate corrected metrics and produced faster takeover reactions than limitation-focused education (no difference vs. Basic). Across participants, greater trust growth correlated with higher mind-wandering, while more structured gaze was associated with lower mind-wandering. Discussion Overall, trust formation appeared to be primarily associated with direct experience with system performance, whereas targeted education refined what drivers monitored and how quickly they responded. Together, these results clarify how experience primarily builds trust while education selectively sharpens attention and response readiness in automated driving. These findings clarify distinct roles of experience and brief education in supervising automation and have implications for driver training, human-machine interface design, and gaze-based monitoring.