Enhancing self-directed learning with custom GPT AI facilitation among medical students: A randomized controlled trial

This study aims to assess the impact of LearnGuide, a specialized ChatGPT tool designed to support self-directed learning among medical students. In this 14-week randomized controlled trial (ClinicalTrials.gov NCT06276049), 103 medical students were assigned to either an intervention group, which received 12 weeks of problem-based training with LearnGuide support, or a control group, which received identical training without AI assistance. Primary and secondary outcomes, including Self-Directed Learning Scale scores at 6 and 12 weeks, Cornell Critical Thinking Test Level Z scores, and Global Flow Scores, were evaluated with a 14-week follow-up. Mann-Whitney U tests were used for statistical comparisons between the groups. At 6 weeks, the intervention group showed a marginally higher median Self-Directed Learning Scale score, which further improved by 12 weeks (4.15 [95% CI, 0.82 to 7.48]; <i>p</i> = 0.01) and was sustained at the 14-week follow-up. Additionally, this group demonstrated notable improvements in the Cornell Critical Thinking Test Score at 12 weeks (7.11 [95% CI, 4.50 to 9.72]; <i>p</i> &lt; 0.001), which persisted into the 14-week follow-up. The group also experienced enhancements in the Global Flow Score from 6 weeks, maintaining superiority over the control group through 12 weeks. LearnGuide significantly enhanced self-directed learning, critical thinking, and flow experiences in medical students, highlighting the crucial role of AI tools in advancing medical education.本研究旨在评估LearnGuide——一款专为支持医学生自主学习而开发的专用ChatGPT工具——的应用效果。本研究为一项为期14周的随机对照试验（ClinicalTrials.gov注册号：NCT06276049），共纳入103名医学生，随机分为干预组与对照组：干预组接受依托LearnGuide支持的12周基于问题的培训，对照组则接受无人工智能辅助的同等培训方案。本研究评估了主要及次要结局指标，包括第6周与第12周的自主学习量表（Self-Directed Learning Scale）得分、康奈尔批判性思维测验Z版（Cornell Critical Thinking Test Level Z）得分，以及整体心流体验得分（Global Flow Scores），并完成了14周的随访评估。组间统计学比较采用曼-惠特尼U检验（Mann-Whitney U tests）。第6周时，干预组的自主学习量表中位得分略高于对照组；至第12周时，该得分进一步提升（4.15 [95%置信区间：0.82~7.48]；*p*=0.01），并在14周随访时仍维持该优势。此外，干预组在第12周时的康奈尔批判性思维测验得分亦出现显著改善（7.11 [95%置信区间：4.50~9.72]；*p*<0.001），该优势持续至14周随访阶段。同时，干预组的整体心流体验得分自第6周起即出现提升，并在第12周时仍显著优于对照组。本研究结果表明，LearnGuide可显著提升医学生的自主学习能力、批判性思维水平及心流体验，凸显了人工智能工具在推动医学教育发展中的重要作用。