Abstract: The rapid advancement of artificial intelligence (AI) is driving the transformation of medical education toward a data-driven model, creating new opportunities for scientific research training among medical students. Currently, scientific research training in medical education generally faces challenges such as high entry barriers and difficulties for students to independently complete the entire process from topic selection to research design. To address these issues, this study developed an integrated smart teaching model of ″objectives-activities-evaluation″ driven by a ″task engine″ based on the principles of ″learner-centered″ and ″constructive alignment″. Through sequenced tasks, AI tools were deeply integrated into core teaching components, including virtual topic selection, research design, and outcome evaluation. Teaching practices were conducted with 135 students enrolled in the clinical medicine (″5+3″ integration) program, class of 2023, at Tongji University School of Medicine, and the teaching effectiveness was evaluated via questionnaires. The results showed that the course significantly enhanced medical students′ AI literacy: in terms of basic AI capabilities, the proportion of medical students with no prior AI knowledge decreased from 10.4% (14/135) before the course to 0.8% (1/131) after the course; in conceptual understanding, the percentages of medical students familiar with machine learning, deep learning, and natural language processing increased from 75.6% (102/135), 68.9% (93/135), and 43.0% (58/135) to 91.6% (120/131), 88.5% (116/131) and 70.2% (92/131), respectively; in tool application, the proportion of medical students using AI tools multiple times daily rose from 3.7% (5/135) to 55.0% (72/131), and those believing AI significantly improves research efficiency increased from 20.0% (27/135) to 68.7% (90/131); in critical thinking, the percentage of meidical students actively verifying the reliability of AI-generated information increased from 36.3% (49/135) to 55.0% (72/131), and those questioning and validating AI conclusions rose from 33.3% (45/135) to 48.1% (63/131). The study demonstrates that systematically incorporating AI via a task engine can effectively enhance medical students′ conceptual understanding and tool application of AI, significantly improving the efficiency of research training. However, its effect on cultivating critical thinking skills was relatively limited, suggesting that AI-empowered teaching should guard against technology dependence during implementation and promote the simultaneous development of higher-order thinking through structured designs such as mandatory verification and reflective exercises.

Key words: Students, medical, AI empowered, Learner-centered, Constructive alignment, Task engine, Research training