Construction and application of a problem-based learning teaching model based on the DeepSeek-R1 model

Abstract

Objective To leverage the chain-of-reasoning capability of DeepSeek-R1 (DS) to construct an AI patient capable of demonstrating the clinical reasoning process,and integrate it into PBL teaching for respiratory diseases,evaluating its effectiveness in improving students′ clinical reasoning skills,pathophysiological understanding,and learning motivation. Methods The study participants were eight-year program medical students enrolled in this course at Fudan University. A total of 127 students,divided into five teaching groups,were participated in DS-PBL teaching practice. Post-class questionnaires were used to evaluate their feedbacks. A 5-point Likert scale and multiple-choice questions were employed to evaluate teaching effectiveness,and open-ended questions were analyzed thematically. Results Students reported a high overall satisfaction with the DS-PBL teaching model(79.5% selected “satisfied” or “very satisfied”). 92.1% of the students agreed that “AI intervention helped broaden my diagnostic thinking,” and 90.6% agreed that “by comparing the AI′s debriefing with that of the instructor,I gained a clearer understanding of the limitations of AI in clinical decision-making.” Regarding multidimensional competency improvement,students showed notable gains in understanding acute respiratory distress syndrome (ARDS) in terms of “linking micro-structural damage to macro-functional failure” (83.5%) and “explaining clinical manifestations through pathophysiological mechanisms” (81.9%). Open-ended feedback revealed issues with the AI tool,including “information contradiction”,“role confusion”,and “insufficient depth of reasoning”. Conclusions The capability of DeepSeek-R1 can effectively simulate the clinical reasoning process in respiratory medicine,making it a useful auxiliary tool in PBL teaching for basic medical courses.