Linking Behaviour and Perception to Evaluate Meaningful Human Control over Partially Automated Driving
archive: archived pipeline: cataloged verified
Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)
Summary
This study addresses the tension in partially automated driving where drivers retain legal responsibility but experience reduced control, potentially undermining their sense of agency and ability to intervene safely. The authors investigate whether current interaction strategies provide "Meaningful Human Control" (MHC), a normative framework requiring that automation tracks human reasons and allows responsibility to be traced to humans. To evaluate this, the researchers developed a methodology integrating objective behavioural metrics with subjective perception scores, comparing two common interaction strategies: Haptic Shared Control (HSC), where human and automation apply torque simultaneously, and Traded Control (TC), which relies on explicit handovers based on torque thresholds. The experimental design involved 24 participants in a driving simulator study using virtual reality and a haptic steering wheel. Participants performed overtaking manoeuvres on a rural road under three conditions: manual baseline, HSC, and TC. To test intervention capabilities, the study simulated "silent automation failures" where the automated system failed to detect a motorcyclist, creating a collision risk that required driver intervention. Data collection included telemetry-derived behavioural metrics (e.g., reaction time, steering torque conflict, trajectory jerk) and post-trial surveys assessing MHC-related perceptions such as sufficient control, shared understanding, and responsibility. Linear mixed-effect models were used to test hypothesized correlations between behavioural metrics and subjective scores, while qualitative analysis of post-experiment questionnaires provided contextual insights. The results revealed significant relationships between driver behaviour and perception. Confirmatory analysis showed a significant negative correlation between the perception that the automated vehicle understood the driver and the level of conflict in steering torques, indicating that higher disagreement reduces perceived alignment. Surprisingly, exploratory analysis found a positive correlation between reaction times and the perception of sufficient control, contradicting the hypothesis that faster reactions indicate better control. Qualitative feedback identified mismatches in intentions, lack of safety, and resistance to driver inputs as key factors reducing perceived MHC. Conversely, subtle haptic guidance aligned with driver intent positively influenced perceptions. The study found that HSC generally resulted in lower steering torque conflict and smoother trajectories compared to TC, though both modes presented challenges in maintaining high perceived control during failures. These findings suggest that evaluating MHC requires combining behavioural telemetry with subjective assessment, as neither alone captures the full interaction dynamic. The results imply that future partially automated systems should prioritize effortless driver interventions, transparent communication of automation intent, and context-sensitive authority allocation. By reducing conflict and aligning automation with driver expectations, designers can better support the sense of agency and responsibility necessary for safe human-automation collaboration, addressing the current gap between legal accountability and operational control in automated vehicles.
Key finding
Confirmatory analysis showed a significant negative correlation between drivers' perception of the automated vehicle understanding them and conflict in steering torques. Exploratory analysis revealed a positive correlation between reaction times and the perception of sufficient control. Qualitative feedback indicated that mismatched intentions, perceived lack of safety, and resistance to driver inputs reduced perceived MHC, while subtle haptic guidance aligned with driver intent improved it. Authors recommend prioritising effortless interventions, transparent automation-intent communication, and context-sensitive authority allocation.
Methodology
simulator
Sample size: N=24 (13 male, 11 female), aged 23-36 (M=29.2, SD=3.75); recruited from TU Delft, Dec 2023
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 discover_arxiv_cs.HC on 2026-05-04 (4 acquisition events logged).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 2 | 2026-05-28 |
| archive | success | — | — | — | 1 | 2026-05-04 |
| extract | success | cached | — | — | 2 | 2026-06-10 |
| clean | success | — | — | — | 1 | 2026-06-01 |
| chunk | success | — | — | — | 1 | 2026-06-01 |
| embed | success | — | — | — | 1 | 2026-06-02 |
| enrich | skipped | — | — | — | 3 | 2026-07-02 |
| promote | success | — | — | — | 1 | 2026-05-04 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 16 | 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).
- Empirical Findings: self report data
- Theoretical Contribution: conceptual framework, computational model