Predicting Trust In Autonomous Vehicles: Modeling Young Adult Psychosocial Traits, Risk-Benefit Attitudes, And Driving Factors With Machine Learning

TL;DR

This study investigates trust in autonomous vehicles among young adults (n=1457) using machine learning and SHAP explanations to identify key predictors. It finds that perceived AV risks and benefits, feasibility and usability attitudes, institutional trust, prior AV experience, and mental models are the strongest predictors of trust, while psychosocial and many driving/tech factors are less influential. Random Forests achieve about 85.8% balanced accuracy in predicting high vs. low trust, with SHAP revealing that risk-benefit factors account for most predictive power. The work highlights the importance of explainable AI, risk-focused design, and subpopulation tailoring to enhance AV trust and adoption. Practical implications address UI/UX, education, and regulatory transparency to support trustworthy human-AV interaction across diverse user groups.

Abstract

Low trust remains a significant barrier to Autonomous Vehicle (AV) adoption. To design trustworthy AVs, we need to better understand the individual traits, attitudes, and experiences that impact people's trust judgements. We use machine learning to understand the most important factors that contribute to young adult trust based on a comprehensive set of personal factors gathered via survey (n = 1457). Factors ranged from psychosocial and cognitive attributes to driving style, experiences, and perceived AV risks and benefits. Using the explainable AI technique SHAP, we found that perceptions of AV risks and benefits, attitudes toward feasibility and usability, institutional trust, prior experience, and a person's mental model are the most important predictors. Surprisingly, psychosocial and many technology- and driving-specific factors were not strong predictors. Results highlight the importance of individual differences for designing trustworthy AVs for diverse groups and lead to key implications for future design and research.

Figures (7)

• Figure 1: Overall process architecture, from data collection and processing to model development and explanations using SHAP.
• Figure 2: Full Feature Model: mean absolute SHAP scores, averaged across 5 folds. This shows the most important features for predicting both high and low trust (shown in combination, as they are mirrors of each other). Risk and benefit factors were the most important.
• Figure 3: Full Feature Model SHAP summary plot showing feature value impact on trust. This plot shows how the value of a feature impacts the model's output. The more extreme the SHAP value, the more indicative that value was of being high trust (positive SHAP values) or low trust (negative SHAP values). Features are organized by overall importance.
• Figure 4: Feature Subset (risk and benefit only): mean absolute SHAP scores, averaged across 5 folds. This shows the most important features for predicting both high and low trust (shown in combination, as they are mirrors of each other). Results are similar to the full feature set model, supporting risk and benefit features as the most important predictors of trust.
• Figure 5: Feature Subset (risk and benefit only) SHAP summary plot showing feature value impact on trust. This plot shows how the value of a feature impacts the model's output. The more extreme the SHAP value, the more indicative that value was of being high trust (positive SHAP values) or low trust (negative SHAP values). Features are organized by overall importance.