Building Trust Profiles in Conditionally Automated Driving

Avetisyan, Lilit; Ayoub, Jackie; Yang, X. Jessie; Zhou, Feng · 2023 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2306.16567

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 challenge of establishing user trust in conditionally automated vehicles (AVs), a prerequisite for widespread adoption and safety. While previous research has examined factors influencing trust, such as system reliability and transparency, there is a lack of understanding regarding how individual differences—such as personality, emotions, and dispositional trust—interact to shape dynamic trust profiles during driving. The authors aim to identify distinct trust profiles to enable personalized in-vehicle monitoring and calibration systems that adjust to drivers’ specific trust dynamics. The researchers conducted a between-subjects experiment using a desktop driving simulator with 70 university students. Participants were assigned to one of three conditions representing different levels of system reliability: a control condition with valid takeover requests (TORs), a false alarm condition, and a miss condition where the AV failed to detect obstacles. The study measured dispositional trust, initial learned trust, personality traits (using the Big Five model), and emotional responses. Crucially, dynamic trust was recorded continuously via self-reported ratings every 25 seconds during approximately 30 minutes of simulated driving. The authors employed a data-driven methodology, using K-means clustering on 48 normalized features to identify trust profiles, and validated these clusters using a multinomial logistic regression model interpreted with SHAP (SHapley Additive exPlanations). The analysis identified three distinct trust profiles: "oscillators," "believers," and "disbelievers." Oscillators (23 participants) exhibited high dispositional and initial learned trust but showed rapid fluctuations in dynamic trust based on immediate experiences. Believers (31 participants) maintained relatively high and increasing dynamic trust, characterized by lower levels of negative emotions and higher confidence. Disbelievers (16 participants) displayed consistently low dynamic trust and were more critical of the AV’s performance. Statistical analysis revealed significant differences in personality and emotions across profiles; notably, oscillators scored significantly higher in Agreeableness (specifically lower on "carper" traits) than disbelievers. Believers reported significantly lower levels of resentful aversion and nervous fear, while oscillators reported higher happiness than disbelievers. The validation model achieved an F1-score of 0.90 and an accuracy of 0.89, confirming that personality, dispositional trust, and emotional states are strong predictors of trust profiles. The findings demonstrate that trust in AVs is not uniform but varies significantly based on individual psychological and emotional factors. By identifying these profiles, the study provides actionable insights for designing personalized trust-calibration systems. For instance, understanding that oscillators are highly sensitive to immediate system performance allows designers to implement interventions that stabilize trust during errors. The results underscore the importance of integrating emotional and personality data into AV design to enhance safety and user experience, moving beyond generic trust-building strategies toward tailored approaches that account for individual driver personas.

Key finding

Three distinct trust profiles (oscillators, believers, and disbelievers) were identified based on personality, emotions, and dynamic trust, with believers showing higher confidence and disbelievers exhibiting lower trust and more negative emotions.

Methodology

simulator

Sample size: 70

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 author_sweep_intake on 2026-05-29 (2 acquisition events logged).

StageOutcomeToolModelPromptAttemptsCompleted
discover success author_sweep 3 2026-05-29
archive success canonical_url 1 2026-06-04
extract success cached 3 2026-06-10
clean success clean 1 2026-06-04
chunk success chunk 1 2026-06-04
embed success embed Qwen/Qwen3-Embedding-8B 1 2026-06-04
enrich success 1 2026-05-29
promote success 1 2026-06-04
summarize success llm qwen3.6-27b-prismaquant summ-v5 2 2026-06-10
tag success vector_similarity 15 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).