Work

AI-enabled digital resources for biochemistry and molecular biology education: development and empirical assessment via teacher-student surveys

Year: 2026

Type: article

Abstract: Biochemistry and Molecular Biology presents substantial learning challenges for medical students. While artificial intelligence (AI)-enabled digital resources hold considerable promise for enhancing teaching and learning, empirical evaluations of integrated resource suites in preclinical courses remain limited, particularly from a multi-stakeholder perspective. To address this gap, this study applies the Technological Pedagogical Content Knowledge framework to develop and evaluate AI-driven teaching resources for this subject. We developed three categories of AI-enabled resources: an intelligent curriculum platform, an AI learning agent, and a preclinical-to-clinical case library. Using convenience sampling, four self-developed 5-point Likert questionnaires were distributed to medical students, preclinical teachers, and clinical practitioners. Reliability was assessed via Cronbach’s α coefficient, and content validity was evaluated by five subject experts who rated each item to compute item-level and scale-level content validity index values. Production time was recorded for three teaching outputs under both manual and AI-supported workflows, and open-ended feedback was analyzed using descriptive thematic summarization. Notably, this was a cross-sectional study that relied entirely on self-reported perceptual data. Students whose ratings ranged from 4.35 to 4.60 (out of 5) responded favorably to basic functions but identified limitations in complex knowledge interpretation and personalized learning plan generation. Teacher scores varied substantially (2.82–4.50 out of 5). The question bank module received the most positive feedback, whereas the AI micro-lesson production function received the lowest ratings. AI-generated clinical cases were evaluated as acceptable, though users perceived gaps in atypical scenario simulation and frontier knowledge updates. Monthly valid AI agent accesses totalled 1,575. AI-supported workflows demonstrated shorter observed creation cycles; however, equivalent output quality was not formally verified. User feedback highlighted tool adaptability limitations, slow content updates, and insufficient cross-disciplinary technical support. Participants reported satisfactory subjective experiences with routine knowledge organization and auxiliary teaching tasks. However, notable deficiencies were identified in high-level personalised services, teacher-end operability, clinical scenario authenticity, and technical support. Given the study design and the absence of objective learning outcome measures, the observed time-saving potential cannot be interpreted as proven objective benefits. Four tentative optimization strategies are proposed to inform the iterative improvement of intelligent teaching practice.

Source: BMC Medical Education

DOI: 10.1186/s12909-026-10743-4