Perceptions and Attitudes of Medical Students Toward the Integration of Large Language Models in Medical Education: Cross-Sectional Survey in China.

Background: Although artificial intelligence (AI) is being rapidly integrated into medical education, insights from medical students, particularly in the Chinese context, remain limited. Objective: This study was designed to explore Chinese medical students' perceptions of and attitudes toward the integration of AI into medical education, as well as the factors that may influence their perspectives. The findings of this research offer valuable insights to assist medical educators in the effective implementation of these innovative educational approaches. Methods: On the basis of the estimated total number of clinical medical students at the target institutions, the sample size was calculated to be 379. A web-based questionnaire survey was designed to investigate the acceptance level of medical students toward the application of AI. The questionnaire consisted of 14 questions across 4 dimensions, which included demographic characteristics, perceptions of AI application, willingness, and concerns. Each dimension contained 3 to 4 questions. Descriptive statistics were used to tabulate the frequency of each variable. Chi-square tests and multiple regression analyses were conducted to measure the influencing factors. Results: A total of 566 cross-sectional online surveys were distributed from December 2023 to January 2024 through a snowball sampling method. Finally, 490 medical students from various local tertiary medical centers were involved. Overall, a majority of the participants showed a positive attitude toward future learning and the usage of AI, manifested as totally willing to acquire relevant knowledge (222/490, 45.3%), totally willing to use AI tools (230/490, 46.9%), and totally desiring that schools would offer AI-related courses (230/490, 46.9%). However, there is still a large proportion (392/490, 80.0%) of participants who expressed concerns regarding ethical issues. The findings also indicated that gender and educational level were significantly correlated with the AI application. Specifically, regression analysis indicated that male participants were more inclined to acquire AI information through social media (odds ratio 0.458, 95% CI 0.33-0.67; P<.001) and that male or graduate-level participants were more likely to use AI for academic writing purposes (odds ratio 0.476, 95% CI 0.38-0.82; P=.001 for male; odds ratio 1.552, 95% CI 1.32-1.77; P=.009 for graduate students, respectively). Conclusions: Our findings indicate that a general awareness of AI's role in medical education is evident among students. However, subgroup-specific differences must be taken into account, particularly when designing and optimizing educational strategies integrated with AI. This consideration is critical to ensuring that such tools align with the diverse learning needs of distinct student groups.