Effects of AI- and VR-supported multicomponent learning intervention configurations on supervisor-rated feedback-literacy behaviors in medical students: a randomized controlled trial

This study aimed to investigate whether participation in two different multicomponent intervention configurations combining AI-supported reflective activities and VR-based clinical scenarios was associated with higher post-test levels of supervisor-rated feedback-literacy behaviors than usual clinical clerkship activities. Both intervention configurations comprised a preparatory workshop and mandatory AI-supported reflective activities within VR-based clinical scenarios, but differed in the inclusion of peer dialogue and in the distribution of active VR participation. A randomized controlled study was conducted among 68 fourth-year medical students during the transition to clinical clerkships. Students participated in one of two intervention configurations or a control condition. Supervisors assessed students’ feedback-literacy behaviors at pre-test and post-test using the study-specific Supervisor Observational Checklist for Feedback Literacy (SOC-FL). Quantitative data were analyzed to compare post-test scores across groups after adjustment for baseline scores, while qualitative data from 36 student reflections were analyzed thematically to provide interpretive context. After adjustment for baseline SOC-FL scores, both intervention configurations had significantly higher adjusted post-test scores than the control group (AI–VR: 21.12 points, 95% CI 18.19–24.05; AI–VR+peer: 23.45 points, 95% CI 20.59–26.31; both p < .001). The two configurations did not differ significantly (2.33 points, 95% CI − 0.53 to 5.19; p = .149). Qualitative reflections indicated that students in the AI–VR+peer configuration more frequently described collaborative sense-making, normalization of uncertainty, and reduced emotional threat, whereas those in the AI–VR configuration reported more varied experiences of independent AI use and individual reflection. Reflections from the control group tended to characterize uncertainty as evidence of incompetence and AI as a confirmation tool. The findings indicate that both multicomponent intervention configurations were associated with higher post-test levels of supervisor-rated feedback-literacy behaviors than usual clinical clerkship activities, though the two configurations did not differ significantly. Qualitative findings revealed differences in students’ reported experiences across configurations. These findings suggest that structured, multicomponent learning interventions may provide additional opportunities for the enactment of observable feedback-literacy behaviors during the transition to clinical training.