To Explain or Not to Explain: A Study on the Necessity of Explanations for Autonomous Vehicles

Not all autonomous vehicle actions require explanations: The research found that the necessity for explanations depends heavily on context—people consistently agree explanations are needed for near-crash events, but hold divergent views on whether ordinary or anomalous driving situations warrant explanation, suggesting resources could be allocated more efficiently to high-stakes scenarios.

Driver characteristics influence explanation preferences: The study reveals that "driver types" affect whether people feel explanations are necessary, indicating that a one-size-fits-all approach to AV transparency may not be optimal and personalization based on user profiles could improve both trust and user experience.

Context-dependent trust building: While previous research emphasized that explanations generally increase trust and acceptance of autonomous vehicles, this work demonstrates that explanation necessity "shifts with situations," suggesting that strategically providing explanations in critical moments may be more effective than constant narration of vehicle decisions.

New benchmark for responsible AI development: The authors created a dataset of 1,103 video clips with human-annotated explanations and necessity ratings, providing researchers with empirical data to develop smarter explanation systems that know when to communicate with users rather than explaining every action.

Designing for human expectations: The findings suggest that responsible AV design should account for when humans naturally expect explanations (particularly in safety-critical moments) rather than imposing explanations based solely on technical capabilities, aligning AI systems with human mental models.