Clues and cortices: A pilot study of efficient, low-cost neuroanatomy escape room design using generative AI

Neuroanatomy is notoriously challenging due to its complexity and abstract spatial relationships, often evoking anxiety and frustration among learners. Educational escape rooms, a form of game-based learning (GBL), can promote engagement and learning but typically require significant faculty time and resources to design. This study describes an efficient, low-cost approach for creating a neuroanatomy escape room using generative artificial intelligence (AI) and evaluates its impact on learner engagement and knowledge. Existing course-aligned materials were transformed by ChatGPT 4.0 into riddles and crossword puzzles, which formed the basis of a four-station escape room for a graduate-level neuroanatomy course. Prior to and after the escape room, students completed pre- and post-questionnaires containing perception items on a 5-point Likert scale (5 = strongly agree, 1 = strongly disagree) and a 10-item neuroanatomy knowledge quiz. A total of 14 students completed the AI-generated escape room, with 11 providing matched pre- and post-questionnaire data. Participants reported high engagement (Median = 5, IQR 5-5), high satisfaction (Median = 5, IQR 5-5), and would recommend it to their peers (Median = 5, IQR 5-5), fulfilling Level 1 (Reaction) of Kirkpatrick's training evaluation framework. Although students believed the escape room improved their understanding of neuroanatomy (Median = 4, IQR 4-5), average quiz scores did not change significantly (both pre and post = 6.4 ± 2.2, p = 0.67), indicating limited knowledge gains. While cognitive gains were not demonstrated, this pilot study suggests that AI-supported escape room development is a feasible, low-burden approach to GBL that warrants further evaluation.