Comparing conventional and generative AI-assisted task performance in physiology education

This study examined the educational impact of generative artificial intelligence (AI) on learners' performance by comparing assignment scores obtained with AI-assisted report writing with those obtained with conventional information-gathering approaches and by exploring factors associated with effective AI use. Medical students participating in a physiology laboratory course were assigned to investigate the mechanisms of action and clinical indications of four pharmacological agents. Students completed the task first with conventional resources and subsequently with generative AI tools of their choice. Reports were evaluated with a predefined keyword-based scoring system designed to capture the presence of core conceptual elements. Questionnaire data on students' prior experience with generative AI and their perceptions of AI were also collected. Thirty-four complete submissions were included in the analysis. There was no significant difference between the conventional and AI-assisted conditions (P = 0.54), whereas scores showed a moderate positive correlation (r = 0.465). Findings from exploratory multivariable analysis were consistent with these results, with performance with conventional resources emerging as the strongest associated factor among the variables examined associated with AI-assisted outcomes. These findings suggest that generative AI does not fundamentally alter task performance in this educational context but instead reflects learners' existing understanding. Effective integration of generative AI in physiology education therefore requires continued emphasis on foundational knowledge rather than reliance on AI tools alone.NEW & NOTEWORTHY In physiology education, generative AI-assisted task performance appears to reflect learners' existing foundational understanding rather than independently enhancing it. The strong association between conventional and AI-assisted performance suggests that generative AI functions as a cognitive mirror, highlighting individual differences in understanding. This insight is particularly relevant as generative AI becomes increasingly integrated into medical education.