Personalizing driver safety interfaces via driver cognitive factors inference
DOI: 10.1038/s41598-024-65144-8
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 limitation of current Advanced Driver Assistance Systems (ADAS), which often employ "one-size-fits-all" interfaces that fail to account for individual cognitive differences influencing risky driving. Specifically, the authors investigate how cognitive factors such as impulsivity and inhibitory control affect driver behavior and response to safety interfaces. The research aims to develop a personalized driver safety system that infers these latent cognitive states from real-time driving behavior to dynamically trigger appropriate human-machine interfaces (HMIs), thereby enhancing safety effectiveness and user acceptance. To achieve this, the researchers conducted a behavioral experiment using a high-fidelity vehicle motion simulator with 27 participants, recruited from two age groups (18–22 and 65+) known for differing risk profiles. Participants completed baseline driving laps and trials involving two types of HMI warnings (transverse road markings and heads-up display circles) triggered either by distance or light state changes. Cognitive traits were measured via standardized assessments, including the BIS/BAS and UPPS-P scales for impulsivity, and Go/No-Go and Stop Signal tasks for inhibitory control. The computational approach involved training a Long Short-Term Memory (LSTM) recurrent neural network to encode recent driving trajectories into a low-dimensional latent space. This model utilized contrastive learning to align the latent representations with the measured cognitive factors, followed by a Support Vector Regression classifier to determine whether to activate an HMI based on the inferred cognitive profile. The results demonstrated significant correlations between cognitive factors and driving behavior; for instance, higher scores in BAS Fun Seeking and UPPS-P Positive Urgency were positively correlated with higher speeds at yellow lights, while longer reaction times in inhibitory control tasks were negatively correlated with speed. Linear mixed models revealed that HMI effectiveness varied by cognitive profile, with individuals high in impulsivity or ordinary violations driving faster when HMIs were present compared to those with lower scores. The neural network successfully inferred these cognitive factors from driving data, achieving strong clustering in the latent space. When applied to personalize HMI deployment, the system achieved a balanced accuracy of 56% and a Cohen’s Kappa of 0.145, outperforming random or always-on strategies. Crucially, the personalized interface reduced mean speed during yellow light phases by 0.59 m/s compared to random deployment, validating the hypothesis that cognitive-aware personalization improves safety outcomes. The significance of this work lies in demonstrating that latent cognitive factors can be reliably inferred from driving behavior and used to personalize ADAS interventions. By separating the inference of cognitive states from the specific HMI design, the proposed framework allows for scalable, adaptable safety systems that target the root causes of risky behavior. This approach moves beyond generic warnings toward tailored interventions that account for individual differences in impulsivity and inhibitory control, potentially reducing accident rates by improving the relevance and effectiveness of driver assistance technologies.
Key finding
Personalizing ADAS warnings using LSTM-inferred cognitive factors reduced mean yellow-light approach speed by ~0.59 m/s versus a non-adaptive baseline, providing initial evidence that driver-cognitive-profile-aware interfaces can shift safety-critical behavior in a simulator.
Methodology
simulator
Sample size: 27 (39 recruited; 7 excluded for motion sickness, 5 for technical issues)
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 direct_pdf on 2026-05-07 (2 acquisition events logged).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | — | — | — | 1 | 2026-05-07 |
| archive | success | canonical_url | — | — | 2 | 2026-06-03 |
| 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 | success | — | — | — | 1 | 2026-05-07 |
| promote | success | — | — | — | 3 | 2026-06-06 |
| 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.
- cognitive capacity variation
- risk taking
- in vehicle coaching
- mental demand
- decision making risk perception
- human error taxonomy
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).
- Methodological Resource: tool software
- Theoretical Contribution: computational model, theory or model