What Did My Car Say? Impact of Autonomous Vehicle Explanation Errors and Driving Context On Comfort, Reliance, Satisfaction, and Driving Confidence

Kaufman, Robert; Broukhim, Aaron; Kirsh, David; Weibel, Nadir · 2025 · openalex

DOI: 10.1145/3706598.3713088

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates how errors in autonomous vehicle (AV) explanations impact passenger trust, reliance, satisfaction, and confidence, while examining the moderating roles of driving context and personal traits. Motivated by the critical need for trustworthy AV deployment, the authors address a gap in existing research: while the negative effects of driving errors are well-documented, the consequences of explanation errors—where the AV miscommunicates its actions or rationale—remain largely unexplored. The study hypothesizes that explanation errors will negatively impact user perceptions and that contextual factors like perceived harm and driving difficulty will influence these effects. The researchers conducted an online experiment with 232 participants using simulated driving scenarios generated via the CARLA simulator. Participants viewed 24 realistic driving videos, each presented under three explanation conditions: accurate (correct action and rationale), low error (correct action, incorrect rationale), and high error (incorrect action and rationale). Crucially, the AV’s driving behavior remained identical across all conditions, allowing the isolation of explanation accuracy effects. Participants rated their comfort relying on the AV, preference for control, explanation satisfaction, and confidence in the AV’s ability. They also rated the perceived harm and difficulty of each scenario. Additionally, the study measured individual differences, including prior trust in AVs and domain expertise. Results demonstrated that explanation errors negatively affected all outcome measures, with impacts proportional to the severity of the error. High errors (incorrect action and rationale) caused greater declines in trust and satisfaction than low errors (incorrect rationale only). Surprisingly, explanation errors reduced ratings of the AV’s driving ability despite identical, lawful driving performance. Contextual factors significantly moderated these effects; perceived harm amplified the negative impact of errors more than driving difficulty did. Furthermore, participants with higher prior trust and greater AV domain expertise reported more positive outcome ratings overall, suggesting that personal traits buffer against some negative perceptions. The findings underscore the necessity for accurate, contextually adaptive, and personalized AV explanations to foster safe reliance. The study concludes that explanation errors can severely undermine trust even when vehicle performance is flawless, highlighting a critical design challenge for explainable AI in safety-critical domains. The authors recommend that future AV systems prioritize explanation accuracy and adapt communication based on situational risk and user characteristics to mitigate trust deterioration and ensure ethical deployment.

Key finding

Explanation errors in autonomous vehicles negatively impact user comfort, reliance, satisfaction, and confidence, with these effects being amplified by the severity of the error and the perceived harm of the driving context.

Methodology

simulator

Sample size: 232

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via openalex_oa_fetch on 2026-05-08 (4 acquisition events logged).

StageOutcomeToolModelPromptAttemptsCompleted
discover success 1 2026-05-07
archive success openalex 15 2026-06-10
extract success cached 3 2026-06-10
clean success 1 2026-06-01
chunk success 1 2026-06-01
embed success 1 2026-06-02
enrich success 1 2026-05-07
promote success 2 2026-06-10
summarize success llm qwen3.6-27b-prismaquant summ-v5 3 2026-06-10
tag success vector_similarity 17 2026-06-11
verify success 2 2026-06-10

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).