Blending AI with Expertise: A Comparative Study of Concept Maps in Anatomy Education

OBJECTIVE: Concept maps are valuable tools in anatomy education, helping students organize complex information visually. With the rise of generative artificial intelligence (AI), tools like large language models (LLMs) and large action models (LAMs) offer new opportunities to automate concept map creation. However, their effectiveness compared with human-generated maps remains unclear. METHODS: In this cross-sectional study, 74 medical students at Qatar University evaluated three concept maps-one generated by a large language model, one by a large action model, and one by a human expert. Students rated each map on clarity, structure, and scientific accuracy using Likert-scale items and selected their preferred map based on explanation, engagement, and future use. Data were analyzed using non-parametric and categorical tests. RESULTS: Human-generated maps demonstrated the highest ratings for contribution to understanding, while clarity ratings were comparable between the human-generated and large language model maps. Maps generated by the action-based model received lower ratings overall. Students consistently preferred the human-generated map across all categories, including explanation, engagement, and future use. While overall differences in categorical preferences were not statistically significant consistently, the selected pairwise comparisons showed significant differences favoring the human-generated map. CONCLUSIONS: Human-generated concept maps remain the most effective and preferred learning tool in anatomy education. Artificial intelligence models, particularly large language models, show promising potential as supportive tools but do not yet surpass expert-designed materials.