Investigating Explanations in Conditional and Highly Automated Driving: The Effects of Situation Awareness and Modality

Zhou, Feng · 2022 · Crossref

DOI: 10.2139/ssrn.4017144

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 addresses the challenge of maintaining driver situation awareness (SA) and trust in conditional and highly automated vehicles (SAE Levels 3 and 4), where drivers are often out of the control loop. The authors propose an SA-based explanation framework grounded in Endsley’s three-level model of SA: perception (Level 1), comprehension (Level 2), and projection (Level 3). The research investigates how explanations mapped to these specific SA levels, delivered via different modalities, affect drivers’ situational trust, cognitive workload, and satisfaction with the explanations. The researchers conducted a between-subjects experiment with 340 participants recruited from Amazon Mechanical Turk. The study utilized a 3 (SA levels: L1, L2, L3) by 2 (modality: visual-only, visual + auditory) factorial design. Participants viewed six simulated driving scenarios involving unexpected events, such as abrupt stops or lane changes. Explanations were provided before the automated vehicle’s actions to align with the respective SA level: L1 provided environmental perception data, L2 added comprehension of the vehicle’s understanding, and L3 included projection of future states. A control group received no explanations. Dependent variables were measured using the Situational Trust Scale for Automated Driving (STS-AD), an explanation satisfaction scale, and the Driving Activity Load Index (DALI) for mental workload. The results indicated that SA level significantly influenced situational trust, with Level 2 explanations yielding the highest trust scores compared to Level 1, Level 3, and the control condition. This suggests that providing information about the vehicle’s comprehension of the situation is most effective for building trust, likely because it addresses the specific SA deficit caused by automation. Regarding explanation satisfaction, a significant interaction effect was found: participants preferred visual-only explanations for Levels 1 and 2 but were more satisfied with combined visual and auditory explanations for Level 3, where the information load was higher. Cognitive workload was significantly higher in the Level 2 condition compared to Level 1 and the control, particularly for visual-only explanations. This increased workload is attributed to the active interpretation required to comprehend the vehicle’s reasoning, which correlates with the observed increase in trust. The study concludes that explanations designed according to a structured SA framework can effectively mitigate the "out-of-the-loop" problem by restoring driver understanding and trust. Specifically, Level 2 explanations offer the optimal balance for fostering trust, despite a moderate increase in cognitive load. The findings also highlight the importance of modality selection, suggesting that multimodal explanations are beneficial when conveying complex, projected future states (Level 3), while simpler perceptual or comprehension-based explanations are best delivered visually. These insights provide actionable guidelines for designing explainable AI interfaces in automated driving systems to enhance human-AV interaction.

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.

• situational awareness
• mode awareness
• automation
• automation surprise
• trust calibration
• situation awareness theory

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, conceptual framework