A Foundational Pedagogical Approach For AI Education In Medicine: A Prospective Pilot Study (Preprint)

Abstract

<sec> <title>BACKGROUND</title> Despite the rapid integration of artificial intelligence (AI) into healthcare delivery, a significant gap persists in AI-focused curricula within undergraduate medical education. Most existing programs focused on raising awareness rather than building technical competency, and in-depth educational options remained prohibitively expensive or extracurricular, leaving medical students ill-equipped to critically evaluate AI tools they would encounter in practice. </sec> <sec> <title>OBJECTIVE</title> The objective of this study was to design, implement, and evaluate a foundational, open-source AI elective course for medical students that bridges foundational mathematics with clinical AI applications using a novel pedagogical framework. </sec> <sec> <title>METHODS</title> The authors designed and administered a 20-lecture open-source elective course for fourth-year medical students at the University of North Carolina at Chapel Hill School of Medicine during the Fall 2025 semester (September to November 2025). The curriculum introduced a novel “Four Pillars of AI” pedagogical model covering Data, Loss, Training, and Network, bridging foundational mathematics with clinical AI applications. The course included guest lectures, an AI rollout simulation, and capstone projects. All course materials were open sourced on GitHub. Pre-course and post-course Post-course surveys demonstrated marked improvements compared to pre-course surveys: 90.9% of students agreed or strongly agreed they could answer basic AI questions from colleagues, compared to low baseline confidence. Additionally, 81% reported comfort critically reading medical AI literature, and 80% reported increased comfort investigating bias in AI systems. Final capstone projects exhibited a high degree of technical sophistication. Surveys were administered to assess changes in student confidence, knowledge, and attitudes toward AI in medicine. </sec> <sec> <title>RESULTS</title> Post-course surveys demonstrated marked improvements compared to pre-course surveys: 90.9% of students agreed or strongly agreed they could answer basic AI questions from colleagues, compared to low baseline confidence. Additionally, 81% reported comfort critically reading medical AI literature, and 80% reported increased comfort investigating bias in AI systems. Final capstone projects exhibited a high degree of technical sophistication. </sec> <sec> <title>CONCLUSIONS</title> A structured, open-source elective course grounded in the “Four Pillars of AI” pedagogical model can effectively build foundational AI competency among medical students. The course demonstrated meaningful improvements in student confidence and technical skills, suggesting that integrating accessible, technically rigorous AI education into medical school curricula is both feasible and effective. </sec> <sec> <title>CLINICALTRIAL</title> N/A </sec>

Ähnliche Arbeiten