Abstract
Artificial intelligence (AI) drives transformation across medical specialities, requiring current and future generations of physicians to navigate ever-changing digital environments. In this context, prospective physicians will play a key role in adopting and applying AI-based health technologies, underlining the importance of understanding their knowledge, attitudes, and intentions toward AI. To dissociate corresponding profiles, we adopted a configurational perspective and conducted a two-stage survey study of 184 ($t_0$) and 138 ($t_1$) medical students at a Canadian medical school. Our principal findings corroborate the existence of distinct clusters in respondents’ AI profiles. We refer to these profiles as the AI unfamiliar, the AI educated, and the AI positive, showing that each profile is associated with different intentions towards future AI use. These exploratory insights on the variety of AI profiles in prospective physicians underline the need for targeted and adaptive measures of education and outreach.
The Adoption Gap in Medical AI
While AI is poised to revolutionize medical fields like radiology and dermatology, its actual impact depends on whether the next generation of doctors is willing to use it. Most existing research uses a "variable-based" approach, looking at factors like "familiarity" or "perceived usefulness" in isolation. However, this study argues that such an approach is too simplistic.
Instead, the authors propose a configurational perspective, which examines how different factors—knowledge, experience, beliefs, and attitudes—cluster together to form complex "profiles." By understanding these profiles, medical educators can move away from "one-size-fits-all" training and toward targeted outreach that addresses specific student concerns or knowledge gaps.
Conceptual Framework and Methodology
The researchers conducted a two-stage survey study at the University of Montréal’s medical school, collecting data in February 2020 ($t_0$, pre-pandemic) and January 2021 ($t_1$, during the pandemic). The study's framework posits that individual profiles are formed by the intersection of knowledge/experience and beliefs/attitudes, which subsequently drive the behavioral intention to integrate AI into medical practice.
Instead of standard linear regression, the researchers utilized cluster analysis to group students based on the interplay between two main dimensions:
- Knowledge and Experience: Familiarity and hands-on experimentation with AI.
- Beliefs and Attitudes: Perceived relevance of AI in the curriculum and its future role in medicine.
Identifying the Three AI Profiles
The study identified three distinct student profiles, particularly evident in the $t_1$ sample. These profiles demonstrate that increasing knowledge does not linearly correlate with increased intention to adopt; rather, the configuration of attitudes and knowledge is key.
- AI-Unfamiliar: Characterized by low knowledge and neutral attitudes, this group showed the lowest intention to adopt AI.
- AI-Educated: These students possessed high knowledge and experience but remained neutral in their attitudes. Consequently, their intention to adopt was only moderate.
- AI-Positive: While possessing only medium technical knowledge, this group held very high positive beliefs about AI's role. They were the most eager to integrate AI into their future careers.
Key Insights: Attitude Trumps Knowledge
The core finding is that a student's holistic profile—specifically their positive beliefs about AI's role in medicine—is a far more powerful predictor of their intention to use the technology than technical knowledge alone. Profile membership explained 30% or more of the variance in students' intentions. On an 8-point scale of intention to use AI, AI-Positive students scored 5.83, significantly higher than the AI-Educated (3.27) and AI-Unfamiliar (1.75) groups.
Interestingly, the COVID-19 pandemic did not drastically shift these underlying profiles. While students in 2021 reported slightly less familiarity with AI (likely due to the shift to remote learning), the fundamental clusters remained consistent. Furthermore, demographic factors like age, gender, and academic level were generally not significant predictors of intention compared to the AI profiles themselves.
Implications and Future Work
These findings suggest that medical curricula should not only focus on technical AI education but also on demonstrating the clinical value and relevance of AI to foster positive attitudes.
Limitations of the study include a generally low level of hands-on AI experimentation among students and the single-site nature of the research. Future work should incorporate emotional facets, such as the fear of professional replacement, and utilize Qualitative Comparative Analysis (QCA) to further refine how specific combinations of conditions lead to high adoption intentions.
Summary updated