Comparing individual and peer-supported ChatGPT use in nursing case analysis: Effects on nursing diagnostic skills and cognitive load in a randomized controlled trial.

AIM: To investigate the effects of three instructional conditions for human-AI collaboration on students' nursing diagnostic skills and cognitive load. BACKGROUND: AI tools are increasingly used to support nursing education, including analyzing clinical scenarios, generating cases and formulating nursing diagnoses. However, the actual educational benefits and efficacy of AI in these specific domains remain unclear. DESIGN: The study was a pretest-posttest, parallel-group, randomized controlled design. METHODS: A randomized controlled trial was conducted with 108 eligible students who were randomly assigned to one of the three conditions: the Individual group, the Individual + ChatGPT group and the Peer + ChatGPT group. Within the three weeks of case-based learning sessions, students formulated nursing diagnoses for surgical cases. Data were collected using the Nursing Diagnosis Evaluation Form, the Cognitive Load Measurement Scale and the Agent Persona Instrument. RESULTS: The final analyses included data from 91 students. Although students' nursing diagnostic skills improved after the intervention, we did not observe a significant interaction effect among the groups (p = .936). A decrease in total cognitive load was observed in all three groups, while germane cognitive load showed a descriptive increase in the Peer + ChatGPT group, though this pattern did not reach statistical significance. CONCLUSION: Compared with the traditional method, ChatGPT-supported conditions showed no statistically significant advantage in diagnostic skills. However, when students worked with both peers and ChatGPT, they may have demonstrated greater cognitive engagement through increased germane cognitive load, but it requires replication with larger samples and more rigorous assessments.