Explainable Artificial Intelligence (XAI) in Medical Education: A Multi- Modal Framework for Enhancing Human-AI Collaboration

<title>Abstract</title> <p>Background While Artificial Intelligence (AI) rapidly advances clinical diagnostic accuracy, the black-box nature of complex algorithms presents critical barriers to pedagogical integration in medical curricula. This study evaluated the efficacy of Explainable AI (XAI) as an interactive pedagogical intervention designed to enhance clinical reasoning and model interpretability during AI-assisted diagnostic training. Methods A prospective, mixed-methods randomized controlled trial was conducted with 120 third-year medical students. Participants were randomized to either a standard AI instruction group (n = 60) or an XAI-enhanced group (n = 60) utilizing the CerViD-MultiModal framework, which integrated dynamic SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) visualizations into neuroimaging case analyses. Standardized psychometric instruments evaluated AI Literacy (0-100), System Usability Scale (SUS, 0-100), NASA Task Load Index (NASA-TLX, 0-100), and Confidence in AI Interpretation (1–5). Results The XAI-enhanced group demonstrated statistically significant improvements across all primary outcomes compared to the control group. AI Literacy scores increased markedly by 34.1% (85.0 ± 5.6 vs. 63.4 ± 6.8, p &lt; 0.001, Cohen’s d = 3.44). Perceived system usability (SUS) improved by 66.5% (36.3 ± 4.3 vs. 21.8 ± 4.8, p &lt; 0.001, d = 3.15). The XAI intervention significantly attenuated perceived cognitive workload on the NASA-TLX scale by 37.4% (26.5 ± 6.5 vs. 42.3 ± 7.1, p &lt; 0.001, d = -2.31). Furthermore, learner confidence in AI interpretation grew by 30.8% (3.4 ± 0.4 vs. 2.6 ± 0.4, p &lt; 0.001, d = 1.89). Conclusions Beyond serving as an algorithmic transparency mechanism, XAI operates as a transformative pedagogical instrument in medical training. By demystifying model decision-making processes, the CerViD-MultiModal framework optimizes human-AI collaboration, significantly bolstering diagnostic comprehension and user trust while mitigating cognitive load in clinical learners. Trial Registration ClinicalTrials.gov NCT07743658 Registered 29 July 2026 (Retrospectively registered). https//clinicaltrials.gov/study/NCT07743658</p>