Impact of Generative Artificial Intelligence on Clinical Reasoning Skills among Medical Students

Abstract

Background: Generative artificial intelligence may support medical students through interactive case-based learning and feedback, but its effect on independent clinical reasoning remains uncertain.

Objective: To compare clinical reasoning performance between medical students receiving structured generative AI-assisted learning and those receiving conventional instruction.

Methods: This proposed prospective, parallel-group randomized educational trial included 60 undergraduate medical students allocated equally to AI-assisted learning and conventional learning groups. Both groups completed baseline and postintervention clinical reasoning assessments scored on a 100-point scale. The primary comparison used baseline-adjusted analysis of covariance, with exploratory comparisons of mean improvement and achievement thresholds. The reported dataset and analyses are synthetic and intended for methodological illustration.

Results: Baseline scores were comparable between the AI-assisted and conventional groups (55.63 ± 8.19 vs. 56.70 ± 8.82; p=0.629). Postintervention scores were higher in the AI-assisted group (77.73 ± 8.73 vs. 69.80 ± 8.08; p<0.001). Mean improvement was 22.10 versus 13.10 points, with a between-group difference of 9.00 points (95% CI 4.25–13.75; p<0.001). The baseline-adjusted mean difference was 8.37 points (95% CI 4.36–12.37; p<0.001), and Cohen’s d for the postintervention difference was 0.94.

Conclusion: In this synthetic analysis, structured AI-assisted learning was associated with higher postintervention clinical reasoning scores than conventional instruction. Prospective trials using authentic student data and validated assessments are required before concluding that generative AI improves independent clinical reasoning competence.