Using ChatGPT as an assessment tool for medical residents in Mexico: a descriptive experience.

Introduction: Artificial intelligence (AI) in medical education has progressed gradually, with numerous authors debating whether to prohibit, restrict, or adopt its use in academic contexts. Growing evidence exists regarding the capabilities and applications of AI in this field, particularly in supporting educational tasks such as student assessment. In this article we described our experience using ChatGPT to evaluate medical residents. Materials and methods: A descriptive cross-sectional study was conducted involving 35 medical residents from different specialty's at a secondary-level hospital. Two different exams were generated using ChatGPT in topics of Rocky Mountain Spotted Fever (RMSF) and Pertussis. Additionally, an opinion survey-previously validated was administered to assess participants' perceptions of ChatGPT ability to generate multiple-choice questions. Results: Overall average score for the Pertussis examination was 8.46, while the average for the RMSF examination was 8.29. All participants reported that the examination was well written and that the language used was coherent; 34 residents (97.14%) stated that the language was clear, concise, and easy to understand; 9 residents (25.71%) agreed that the language used was confusing; 33 residents (94.28%) rated the exams questions as difficult; 32 residents (91.42%) felt that they had adequately prepared for both examinations. Discussion: ChatGPT exhibits a promising faculty as a tool to support teaching activities in the training of medical specialists, mainly in reducing the human workload of healthcare personnel, and becoming integral to the next phase of medical education through AI-assisted content creation supervised by educators.