Trust is crucial for ensuring the safety, security, and widespread adoption of automated vehicles (AVs), and if trust is lacking, drivers and the public may not be willing to use them. This research seeks to investigate trust profiles in order to create personalized experiences for drivers in AVs. This technique helps in better understanding drivers' dynamic trust from a persona's perspective. The study was conducted in a driving simulator where participants were requested to take over control from automated driving in three conditions that included a control condition, a false alarm condition, and a miss condition with eight takeover requests (TORs) in different scenarios. Drivers' dispositional trust, initial learned trust, dynamic trust, personality, and emotions were measured. We identified three trust profiles (i.e., believers, oscillators, and disbelievers) using a K-means clustering model. In order to validate this model, we built a multinomial logistic regression model based on SHAP explainer that selected the most important features to predict the trust profiles with an F1-score of 0.90 and accuracy of 0.89. We also discussed how different individual factors influenced trust profiles which helped us understand trust dynamics better from a persona's perspective. Our findings have important implications for designing a personalized in-vehicle trust monitoring and calibrating system to adjust drivers' trust levels in order to improve safety and experience in automated driving.

翻译：信任对于确保自动驾驶汽车的安全性、可靠性和广泛采用至关重要，如果缺乏信任，驾驶员和公众可能不愿使用它们。本研究旨在通过调查信任画像，为自动驾驶汽车中的驾驶员创建个性化体验。该技术有助于从角色视角更好地理解驾驶员的动态信任。实验在驾驶模拟器中进行，参与者在三种条件下被要求接管自动驾驶控制权，包括控制条件、误报条件和漏报条件，并在不同场景中设置了八次接管请求。我们测量了驾驶员的倾向性信任、初始习得信任、动态信任、个性特征和情绪。通过K-means聚类模型识别出三种信任画像（即信任者、摇摆者和不信任者）。为验证该模型，我们基于SHAP解释器构建了多项逻辑回归模型，选取了最重要的特征来预测信任画像，F1分数达到0.90，准确率达到0.89。我们还讨论了不同个体因素如何影响信任画像，这有助于从角色视角更好地理解信任动态。本研究结果对于设计个性化的车内信任监控与校准系统、调节驾驶员信任水平以提升自动驾驶安全性和体验具有重要意义。