DOI: 10.1177/20552076261504553 ISSN: 2055-2076

Design and evaluation of a Socratic GPT to support clinical reasoning in nursing simulation debriefing

Fu-Yu Yang, Li-Chun Chang, Li-Fen Chao
Show PDF Cite

Objective

This study developed and evaluated a Socratic GPT, a three-phase generative artificial intelligence (GAI) agent that integrates Socratic questioning, Identification, Situation, Background, Assessment, and Recommendation (ISBAR)-guided prompts, iterative cognitive scaffolding, and metacognitive feedback to support clinical reasoning during nursing simulation debriefing.

Methods

A nonrandomized quasi-experimental study was conducted with 52 second-year nursing students at a single university. Students were assigned by class to either Socratic GPT-supported debriefing, or teacher-led debriefing. Outcomes were assessed using a clinical reasoning scale (CRS), an objective structured clinical examination (OSCE), and a Likert-type questionnaire evaluating students’ perceptions.

Results

Analysis of covariance, with baseline clinical reasoning score entered as a covariate, showed no significant between-group difference in posttest self-reported clinical reasoning, F (1, 49) = 0.146, p = .704, partial η 2 = .003, adjusted mean difference = −0.045, 95% CI [−0.278, 0.189]. The Socratic GPT-supported group (M = 26.96, SD = 2.31) had significantly higher posttest OSCE scores than the teacher-led group (M = 25.29, SD = 3.00), U = 226, Z = −2.04, p = .04, r = .28.

Conclusion

These findings provide preliminary evidence that pedagogically structured Socratic GPT-supported debriefing may support nursing students’ OSCE performance in a related clinical context, although no significant between-group difference was observed in self-reported clinical reasoning. Further research using larger samples, multiple clusters, and more direct measures of clinical reasoning is needed to determine the educational effects of GAI-supported simulation debriefing.

More from our Archive