Generative artificial intelligence scaffolding for statistical competence in graduate nursing education

Graduate nursing students often experience significant anxiety related to statistics and have difficulty applying quantitative knowledge to clinical research practice. Although generative artificial intelligence may function as a learning scaffold, evidence supporting its structured integration into pedagogy remains limited.

To evaluate the impact of a structured generative artificial intelligence-supported self-directed learning intervention on statistical competence, learning capacity, and motivation among graduate nursing students.

This quasi-experimental study employed a single-group repeated-measures design and included 15 graduate nursing students enrolled in a master's program in Taiwan. The study was conducted from September 10, 2024, to January 15, 2025, with assessments performed at Weeks 1, 10, and 18. The 18-week intervention followed a five-stage scaffolding protocol: (1) instruction in fundamental statistical concepts with instructor-guided support and demonstrations using the Statistical Package for the Social Sciences, (2) independent drafting, (3) online peer collaboration, (4) generative artificial intelligence-assisted learning support, and (5) in-class metacognitive verification. Data were analyzed using descriptive statistics, generalized estimating equations, Pearson's correlation coefficients, and stepwise multiple regression analysis.

Overall learning capacity improved significantly over time (post-test estimate = 1.26, p < .001), with a very large observed effect size (Hedges' g = 3.75). Meaning construction, self-regulation, and interpersonal learning also improved significantly by the post-test (p < .001). Learning motivation increased slightly but was not statistically significant. In the stepwise multiple regression analysis, satisfaction was the only variable significantly associated with final statistics report scores (R² = 0.61; adjusted R² = 0.58, p = .001).

Participants showed improvements in learning capacity following participation in the structured generative artificial intelligence-supported protocol. Learner satisfaction may be associated with higher statistical competence. These findings support the use of generative artificial intelligence as a guided educational scaffold rather than an unregulated shortcut.