Psychological associations between perceived generative AI-supported learning and creative design performance in vocational design education: a within-course correlational study.

Generative artificial intelligence (GenAI) is increasingly used in design courses, but evidence remains limited on how students' appraisal of such support co-varies with domain cognition, creative confidence, engagement, and independently assessed creative work. This within-course correlational study examined within-course change and theory-specified contemporaneous associations in a 72-h Furniture Appreciation and Selection course involving 150 vocational design students in China. Sustainable design cognition and creative self-efficacy were measured before and after the course; perceived GenAI-supported learning and learning engagement were measured after it; and final creative design performance was rated independently by three experts. Paired-samples t tests described within-course change. Item-level CFA preceded the score-level association analyses, while standardized composite-score ordinary least squares models and 10,000-resample percentile bootstrap confidence intervals summarized theory-specified post-course associations and coefficient products. Sustainable design cognition increased from 3.326 to 3.792, t (149) = 10.362, p < 0.001, dz = 0.846, and creative self-efficacy increased from 3.509 to 3.877, t (149) = 7.424, p < 0.001, dz = 0.606; these values indicate change during the course as a whole rather than a GenAI-specific intervention effect. Among the direct post-course associations, learning engagement showed the largest coefficient with expert-rated creative design performance (β = 0.515, p < 0.001; f2 = 0.361). The theory-specified coefficient products were 0.076 for SDC-CSE-LE, 0.204 for CSE-LE-CP, and 0.0080 for AISL-SDC-CSE-LE-CP. The last interval narrowly excluded zero [0.0002, 0.0234], but the product was statistically detectable and negligible in magnitude. Because the study followed one cohort without a comparison condition and most model variables were measured at the same post-course occasion, the findings support within-course change and contemporaneous association, not causal or temporally ordered effects.