Perception of Artificial Intelligence in Medical Education and Clinical Practice Among Medical Students
TL;DR
This first Tunisian study shows that medical students have widely adopted AI, with 90.5% currently using AI in their learning activities and frequency of use is the factor most consistently associated with positive perception and motivation, highlighting the importance of structured, supervised, hands-on AI education in medical curricula.
Abstract
Introduction: Artificial intelligence (AI) is transforming healthcare systems and medical education worldwide. However, data on medical students’ usage patterns, perceptions, and motivation toward AI remain limited in Tunisia. This study aimed to assess these three dimensions among medical students at the Faculty of Medicine of Sousse. Methods: A cross-sectional study using cluster (session-based) sampling was conducted among medical students (pre-clinical, clinical/extern, and internship years) at the Faculty of Medicine of Sousse during the 2024–2025 academic year. The minimum required sample size, calculated using the Schwartz formula (expected proportion 50%, 5% margin of error, 95% confidence level), was 384 students. Data were collected using a paper-based self-administered questionnaire distributed during tutorial sessions for externs and clinical rotations for interns. Univariate analyses (Chi-square, Fisher’s exact test) and multivariate stepwise binary logistic regression were performed. Statistical significance was set at p < 0.05. Results: A total of 495 students participated, exceeding the calculated minimum sample size, with a mean age of 21.96 ± 1.77 years and 55.6% females. Most students (81.0%) were familiar with AI, and 90.5% reported using it, mainly for research (87.1%), exam preparation (57.0%), and writing tasks (52.1%). The mean perception score for AI in medical education was 5.04 ± 2.01/10, with 34.5% classified as positive. For clinical practice, the mean score was 2.40 ± 1.10/5, with 45.5% positive perception. Fear of dehumanization was reported by 72.3%, and only 15.2% believed AI would improve diagnostic accuracy. The mean motivation score was 28.53 ± 5.17/45, with 45.0% showing high motivation. In multivariate analysis, high frequency of AI use was the factor most consistently associated with these outcomes across all models (ORa: 1.73–6.47, p < 0.05). Female gender was associated with positive perception in education (ORa = 1.73, p = 0.030) and higher motivation (ORa = 3.35, p < 0.001). Conclusions: This first Tunisian study shows that medical students have widely adopted AI, with 90.5% currently using AI in their learning activities. Perceptions remain mixed, with low trust in AI-generated content despite high usage and strong demand for training. Frequency of use is the factor most consistently associated with positive perception and motivation, highlighting the importance of structured, supervised, hands-on AI education in medical curricula.