Ethical concerns toward medical artificial intelligence and acceptance intentions: a structural equation modeling analysis of the risk perception–trust pathway

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Objective With the deep integration of artificial intelligence (AI) into medical imaging, clinical decision-making, and health management, ethical concerns regarding medical AI among surveyed individuals have become increasingly prominent. This study examines the hierarchical structure of ethical concerns among surveyed participants and their associations with perceived risk, trust, attitude, and acceptance of medical AI. Methods A questionnaire survey was conducted with 697 valid responses. SPSS and AMOS were used for reliability and validity assessment, confirmatory factor analysis, structural equation modeling, bootstrap analysis of indirect pathways, and robustness checks. Results The results support a hierarchical multidimensional structure of ethical concern, with six first-order dimensions—privacy and data protection, responsibility and accountability, fairness and accessibility, safety and reliability, human–machine collaboration and humanistic care, and technical interpretability and transparency—collectively representing a higher-order ethical concern construct. All six dimensions are significantly associated with higher overall perceived risk of medical AI. Among them, human–machine collaboration and humanistic care ( β = 0.238), fairness and accessibility ( β = 0.196), and technical interpretability and transparency ( β = 0.187) show relatively stronger associations. Overall perceived risk is negatively associated with overall trust ( β = −0.240), while overall trust is positively associated with overall attitude ( β = 0.608) and overall acceptance ( β = 0.395). Overall attitude is positively associated with overall acceptance ( β = 0.397). Personal trust propensity is positively associated with overall trust ( β = 0.555). Bootstrap analysis indicates that overall perceived risk shows an indirect statistical association with overall acceptance involving overall trust within the structural model. Sensitivity analysis incorporating demographic control variables (gender, age, education level, and prior experience with medical AI) indicates that the structural relationships remain consistent. Conclusion Ethical concerns among surveyed participants constitute a hierarchical construct that is systematically associated with perceived risk of medical AI. Perceived risk, trust, attitude, and acceptance are interconnected through multiple statistical pathways, highlighting the complexity of medical AI evaluation processes among surveyed participants.