Automated Failure-Mode Clustering and Labeling for Informed Car-To-Driver Handover in Autonomous Vehicles

Generic alarms are insufficient for safe handovers: Current autonomous vehicles use generic alerts to signal handovers to drivers, but this approach fails to account for drivers engaged in non-driving tasks who need time to regain situational awareness—a delay that "could potentially lead to life-threatening failures."

Explainable AI can improve handover safety: The research is grounded in evidence that "explainable AI has been shown to improve fluency and teamwork in human-robot collaboration scenarios," suggesting that providing drivers with specific information about why a handover is needed helps them refocus on the road faster.

Drivers need context, not just alerts: By automatically identifying and explaining the specific failure mode (e.g., "construction ahead," "sensor malfunction"), drivers can "more rapidly focus on the relevant parts of the driving environment" rather than spending critical seconds figuring out what went wrong.

Human testing is essential for implementation: The authors propose "human-subjects driving-simulator studies to determine the appropriate form of explanation during handovers," recognizing that the how of communication matters as much as the what—different drivers may need different types of information presented differently.

This addresses a real gap in current autonomous vehicle design: The paper identifies a concrete safety problem in partially autonomous vehicles where drivers might be distracted, then proposes a practical solution using AI transparency rather than relying solely on driver attentiveness.