Abstract
Background
The integration of AI into health care is increasingly shaping clinical practice and decision-making. Beyond technical performance, AI has implications for professional roles, clinical reasoning, and responsibility. Understanding how medical students and clinicians perceive AI, how these perceptions relate to professional identity concerns, and how clinicians evaluate AI-supported decisions in different clinical contexts is therefore important for both implementation and medical education.
Objective
This study aimed to examine (1) how medical students and clinicians perceive the impact of AI on medical practice and professional identity and (2) how clinicians evaluate explainability, trustworthiness, and responsibility in AI-supported clinical decision-making across different contexts.
Methods
A 2-phase quantitative study was conducted. In phase 1, medical students (n=93) and clinicians (n=65; N=158) completed a cross-sectional online survey assessing expectations regarding AI, perceived importance of AI across medical domains, and perceived professional identity threat. In phase 2, a separate sample of clinicians (N=68) was quasi-randomly assigned to 1 of 3 clinical vignettes (Watson, Triage, and OncoGuide) and evaluated AI-supported decisions along the dimensions of explainability, trustworthiness, and responsibility.
Results
In phase 1, participants reported generally positive expectations regarding factual-level impacts of AI on medical practice. However, these expectations were not significantly associated with perceived professional identity threat. Medical students reported significantly higher identity threat than clinicians, despite largely similar expectations regarding AI’s factual and social impact. Perceived importance of AI across medical domains was associated with more positive expectations toward AI-supported medical practice, but not with identity threat. In phase 2, clinicians’ evaluations of AI-supported decisions varied across contexts, with explainability ratings differing significantly across vignette scenarios (P=.007), with lower ratings in the Triage vignette than in the Watson and OncoGuide vignettes. Trustworthiness and responsibility did not differ significantly across scenarios.
Conclusions
The findings suggest that positive expectations regarding AI may coexist with professional identity concerns, while identity threat differed between medical students and clinicians in the present sample. At the same time, explainability evaluations differed across clinical decision contexts, whereas trustworthiness and responsibility remained comparatively stable. Together, these findings point to both professional identity-related and contextual considerations in AI implementation that may not be captured by expectations regarding technological benefits alone. Considering these dimensions may be relevant for implementation strategies and medical education.