Toward Adaptive Trust Calibration for Level 2 Driving Automation
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the challenge of calibrating human trust in Level 2 driving automation, specifically within "action-automation" contexts where the system performs actions unless the human intervenes. The authors identify a critical tradeoff: while increasing automation transparency can improve trust calibration, excessive information often increases cognitive workload and distracts the driver. Existing frameworks largely fail to explicitly model the coupled dynamics of trust and workload in real-time, particularly for novice users in complex urban environments. To resolve this, the study proposes a probabilistic framework using a Partially Observable Markov Decision Process (POMDP) to model human trust and workload dynamics, aiming to optimize automation transparency based on real-time estimates of the driver’s state. The researchers conducted a user study using a medium-fidelity driving simulator to collect data from ten participants (after excluding six due to data quality issues). The experiment employed a within-subject design with eight conditions varying traffic density, intersection complexity (presence of pedestrians), and the presence of augmented reality (AR) cues. Automation reliability was manipulated by varying the distance the automated vehicle stopped before a stop line. Human trust and workload were inferred indirectly through behavioral metrics: reliance on the automation (taking over vs. staying hands-off) and eye-gaze position (road, vehicle, pedestrian, etc.), tracked via eye-tracking glasses. The authors used a modified Baum-Welch algorithm to estimate the POMDP parameters, selecting the model structure that minimized the Akaike Information Criterion through cross-validation. The results demonstrate that the proposed POMDP framework can effectively model the coupled dynamics of trust and workload. The optimized model structure identified automation reliability and transparency as key actions influencing trust, while reliability, transparency, and intersection complexity influenced workload; notably, traffic density was found insignificant. The emission probabilities confirmed that high trust correlates with higher reliance on automation, while high workload correlates with specific gaze patterns. The framework successfully varied automation transparency to balance trust calibration against workload, showing that dynamic adjustment of AR cues can align human trust with system reliability without overwhelming the driver. The significance of this work lies in its contribution to human-machine interaction in safety-critical domains. It is the first study to optimize system behavior policy design for real-time trust calibration in Level 2 driving automation by explicitly modeling the coupling between trust and workload. By using observable behavioral metrics rather than intrusive surveys, the approach offers a practical method for adaptive interfaces that maintain safety and performance. This framework provides a foundation for developing automation systems that dynamically adjust their transparency to support novice users, ensuring trust remains calibrated to actual system capabilities rather than being maximized indiscriminately.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-08-09 |
| archive | success | unpaywall | — | — | 2 | 2026-08-09 |
| extract | success | pdftotext | — | — | 4 | 2026-08-10 |
| clean | success | clean | — | — | 2 | 2026-08-10 |
| chunk | success | chunk | — | — | 2 | 2026-08-10 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 2 | 2026-08-10 |
| promote | success | — | — | — | 1 | 2026-08-09 |
| summarize | success | llm | qwen3.6-27b-nvidia | summ-v5 | 2 | 2026-08-10 |
| tag | success | vector_similarity | — | — | 17 | 2026-08-11 |
| verify | partial | — | — | — | 2 | 2026-08-10 |
Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified_with_issues.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
- trust calibration
- automation
- trust in automation foundations
- automation surprise
- situational awareness
- mode 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: computational model, conceptual framework