Bridging the supervision gap: Technology-mediated mentorship to increase accompanied driving and moderate novice driver stress

Novice drivers (NDs) face significantly elevated crash risks during their initial months of independent driving, often due to limited hazard perception and situational awareness. While Graduated Driver Licensing (GDL) systems utilize parental supervision to bridge this gap, the conventional model is constrained by parental availability and interpersonal tension. This study presents a two-stage investigation into technology-mediated mentorship as a transformative solution: remote parental accompaniment and AI-driven avatar accompaniment. Using a high-fidelity driving simulator, 43 parent-ND dyads were exposed to safety-critical “challenging events”, including sudden pedestrian crossings, lead-vehicle braking, and amber light decision points, that are too hazardous for real-world testing. The investigation assessed subjective experiences (usability, safety, tension), physiological stress indices (heart rate variability), and driving performance across three conditions: conventional in-person, remote, and avatar-based accompaniment. Results revealed that both technology-mediated conditions achieved usefulness ratings comparable to those of conventional accompaniment, while significantly reducing peak stress levels during challenging driving moments. Furthermore, at high steering intensity thresholds, technology-mediated guidance was associated with lower event probabilities compared to traditional supervision. These findings validate the potential of alternative mentorship models to enhance the GDL process. Furthermore, this study establishes the empirical foundation and a four-module architecture (integrating perception, guidance, style, and conversation) required for the development of a fully autonomous, empathetic Generative AI driving mentor.