Effects of multimodal explanations for autonomous driving on driving performance, cognitive load, expertise, confidence, and trust

Kaufman, Robert; Costa, Jean; Kimani, Everlyne · 2024 · Crossref

DOI: 10.1038/s41598-024-62052-9

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 whether autonomous vehicles (AVs) can serve as effective instructors for human drivers, specifically addressing the high rate of crashes caused by human error. The researchers explored if an AI coach, providing explanations modeled after expert human instructors, could improve novice drivers' performance in a challenging "performance driving" context. The study aimed to determine how different types of information (descriptive "what" vs. explanatory "why") and presentation modalities (auditory vs. visual) impact driving performance, cognitive load, confidence, expertise, and trust. The researchers conducted a mixed-methods pre-post experiment with 41 novice drivers using a full-motion driving simulator replicating the Thunderhill Raceway. Participants were randomly assigned to one of four conditions: a control group with no explanation, an auditory "what" group, an auditory "what + why" group, and a multimodal group receiving visual "what" (a projected racing line) and auditory "why" explanations. During the observation phase, participants watched an AI agent drive four laps while receiving instructions. Driving performance was measured by distance from the ideal racing line, lap time, maximum speed, and acceleration. Secondary measures included self-reported confidence, cognitive load (NASA-TLX), trust in AVs, and expertise assessments via quizzes and interviews. Results indicated that AI coaching effectively taught performance driving skills to novices. Participants in all instructional groups showed significant improvements in driving performance compared to the control group, particularly in adhering to the optimal racing line. The type and modality of information significantly influenced outcomes. The multimodal condition (visual "what" + auditory "why") generally yielded the best performance results, suggesting that visual cues for spatial positioning combined with auditory rationale for decision-making were most effective. The study found that explanations helped direct attention, mitigate uncertainty, and reduce cognitive overload compared to observation alone. Furthermore, participants reported increased confidence and trust in the AI coach after the session. The findings suggest that efficient, modality-appropriate explanations are crucial for designing Human-Machine Interfaces (HMIs) that instruct without overwhelming users. The study concludes that aligning AI communications with human learning processes—specifically by separating spatial guidance from conceptual rationale—enhances skill acquisition. The authors provide eight design implications for future AV HMI and AI coach development, emphasizing the potential of AVs not just as transportation tools, but as active agents in improving human driving safety and competence.

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.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

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

• in vehicle coaching
• trust calibration
• automation
• automation surprise
• feedback modes
• situational awareness

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).

• Empirical Findings: self report data
• Theoretical Contribution: theory or model, computational model