关键词: 自动驾驶, 人机信任, 人与AI信任

Abstract: This study investigates the role of professional background in shaping initial trust in autonomous vehicles (AVs), with a particular focus on how privacy risk perception influences trust differences between professionals and non-professionals. Previous research has primarily concentrated on the issue of insufficient trust in AVs and sought to enhance trust through improved design and communication. However, excessive trust among ordinary consumers may equally lead to greater risks and hazards. Therefore, how to calibrate consumers’ trust to be closer to that of industry professionals—who are also the system designers—has become a pressing issue. Although existing studies have examined various perceptual factors affecting trust in AVs, there remains a lack of systematic evidence on how prior experiential differences, represented by professional background, shape initial trust. Drawing on three empirical studies with a total of 1,027 participants, this research systematically examines the mechanisms through which privacy risk perception influences trust formation. Study 1 employed an online survey to compare professionals and non-professionals. Results showed that non-professionals reported significantly higher initial trust in AVs, were more likely to overestimate system performance, underestimate potential risks, and be more susceptible to social influence. Regression analyses further revealed an interaction effect between privacy risk perception and professional background: privacy risk perception significantly predicted professionals’ trust levels but had no significant effect on non-professionals. In addition, social influence, perceived usefulness, and safety risk perception jointly predicted initial trust, indicating that trust formation is a complex process shaped by multiple interacting factors. Study 2 experimentally manipulated privacy risk levels to further explore the interaction between privacy risk perception and professional background. Results demonstrated that increased privacy risk significantly reduced professionals’ trust, whereas non-professionals’ trust fluctuated more dramatically. A moderated mediation analysis showed that privacy risk level predicted non-professionals’ trust through privacy risk perception, but the effect was nonsignificant for professionals. This suggests that professionals’ trust is relatively stable, while non-professionals—lacking accurate recognition of privacy risks—are more sensitive to contextual changes. Study 3 examined the impact of enhancing privacy risk perception. Results indicated that increasing non-professionals’ privacy risk awareness heightened their sensitivity to risks, thereby significantly reducing their initial trust. This finding suggests that enhancing privacy risk perception among non-professionals can effectively mitigate excessive trust and narrow the trust gap with professionals. Taken together, this research reveals that professionals and non-professionals rely on different cognitive pathways in forming initial trust in AVs, and clarifies the interactive mechanism between privacy risk perception and professional background. The findings provide important implications for strategies to strengthen consumer trust in AVs. Specifically, enhancing non-professionals’ privacy risk perception helps calibrate excessive trust and narrow the trust gap, while social influence and perceived usefulness also play critical roles in shaping trust. Leveraging social influence and emphasizing the usefulness of AVs may be effective approaches to promoting rational trust. Overall, this research deepens the understanding of trust formation mechanisms in human–machine interaction and offers practical insights for fostering more rational and well-calibrated public trust in autonomous vehicles.

Key words: Autonomous vehicle, human-machine trust, human-AI trust