Modeling dispositional and initial learned trust in automated vehicles with predictability and explainability
DOI: 10.1016/j.trf.2020.12.015
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 critical barrier to automated vehicle (AV) adoption: public trust. While AVs offer safety and efficiency benefits, psychological factors, particularly trust, hinder acceptance. The authors focus on modeling "dispositional trust" (enduring tendency to trust) and "initial learned trust" (trust shaped by past experiences) prior to direct interaction with AVs. The research aims to overcome the limitations of traditional regression models, which lack predictive power, and black-box machine learning models, which lack explainability. To achieve this, the study proposes a method combining eXtreme Gradient Boosting (XGBoost) for high-accuracy prediction and SHapley Additive exPlanations (SHAP) for interpreting feature contributions. The researchers conducted an online survey via Amazon Mechanical Turk, collecting data from 1,175 participants, of whom 1,054 were retained after screening for invalid responses. The survey captured 23 features across five categories: knowledge (eagerness to adopt technology, AV knowledge), experience (ADAS usage, prior AV exposure), risk and benefit perceptions, behavioral assessments (willingness to let children ride alone), and feelings during manual driving (control, excitement, stress, fear). Highly correlated variables were removed, and the response variable was converted into a binary classification of trust versus distrust. The XGBoost classifier was trained using 10-fold cross-validation, and its performance was compared against logistic regression, decision trees, naive Bayes, linear SVM, and random forest. SHAP was then applied to analyze global feature importance, main effects, interaction effects, and local individual predictions. The XGBoost model demonstrated superior performance, achieving 85.5% accuracy, a 0.92 ROC-AUC, and an 86.8% F1 measure, outperforming all other tested models. SHAP analysis identified perceived benefits, perceived risks, excitement about driving, AV knowledge, and eagerness to adopt new technology as the most significant predictors. The results revealed complex interaction effects; for instance, the impact of perceived benefits on trust depended on prior AV experience. Participants with low benefit perceptions trusted AVs more if they had prior experience, whereas those with high benefit perceptions trusted AVs less if they had prior experience. Similarly, higher perceived risks generally decreased trust, but this effect was moderated by experience and knowledge levels. Additionally, greater years of driving experience correlated with lower trust in AVs, though this was influenced by perceived benefits. The study concludes that combining XGBoost with SHAP provides a robust framework for predicting and explaining trust in AVs, offering both high predictability and interpretability. This approach allows researchers and developers to identify not only which factors drive trust but also how these factors interact. The findings highlight that trust is not determined by single variables but by the interplay of cognitive perceptions (risk/benefit), emotional states (excitement/fear), and experiential history. This methodology offers a pathway to better understand public acceptance barriers and design interventions that calibrate trust effectively in the early stages of AV deployment.
Key finding
XGBoost with 10-fold cross-validation predicted trust vs. distrust better than logistic regression, decision trees, naive Bayes, linear SVM, and random forest (accuracy 85.5%, ROC AUC 0.92, recall 91.6%); SHAP ranked perceived benefit, risk, excitement, knowledge about AVs, and eagerness to adopt technology as the top predictors among 1054 screened U.S. survey respondents.
Methodology
survey
Sample size: n=1054
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 openalex_abstract on 2026-05-08 (12 acquisition events logged).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 3 | 2026-05-29 |
| archive | success | — | — | — | 1 | 2026-05-08 |
| 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 | semantic_scholar | — | — | 6 | 2026-05-08 |
| promote | success | — | — | — | 1 | 2026-05-05 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 18 | 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: computational model