Structured Text-to-Image Artificial Intelligence–Assisted Instruction for Miao Embroidery–Based Guofeng Character Design: A Randomized Trial
Journal of Visualized Experiments · 2026-11-10 · (237)
DOI: 10.3791/74929 ↗At a glance
Randomisierte kontrollierte Studie in Journal of Visualized Experiments, published online on 2026-11-10. Written by Mingyu Deng.
Abstract (original language)
Traditional costume education requires students to translate cultural knowledge and craftsmanship concepts into design. This single-center randomized trial compared a structured text-to-image artificial intelligence (AI)-assisted instructional package with conventional instruction among 120 undergraduate design students at Jiangxi Institute of Fashion Technology, China. Participants were randomized 1:1 to AI-assisted or conventional instruction comprising four 90 min sessions delivered in four continuing instructional groups. The AI-assisted package integrated structured prompting, constrained generation, cultural verification, and manual revision. The primary outcome was the expert-rated Guofeng character image-board score (0-100); cultural cognition and stimulus-evoked aesthetic experience were secondary outcomes. Three allocation-masked raters assessed 112 valid artworks, and 113 participants provided valid follow-up questionnaire administrations. Intention-to-treat analyses included all participants across 50 imputed datasets. Individual-level analysis of covariance (ANCOVA) estimated an adjusted artwork difference of 4.56 points favoring AI-assisted instruction (95% confidence interval [CI], 2.58-6.54; p < 0.001), with a residual-standardized contrast of 0.84 (95% CI, 0.48-1.20). Adjusted differences were 1.67 points for cultural cognition (95% CI, 0.79-2.55) and 9.54 points for aesthetic experience (95% CI, 4.46-14.62), both with Holm-adjusted p < 0.001. The intraclass correlation coefficient (ICC) for averaged artwork ratings was ICC(2,k) = 0.889. Complete-case and per-protocol estimates favored the AI-assisted package; however, mixed-effects and cluster-robust confidence intervals for all three outcomes included zero. The individual-level estimates indicated higher immediate scores, with uncertainty after accounting for instructional-group dependence. The findings concern performance with the assigned tools and do not establish the isolated effect of image generation, durable learning, or improved unaided design ability.
Cite
APA
Deng, M. (2026). Structured Text-to-Image Artificial Intelligence–Assisted Instruction for Miao Embroidery–Based Guofeng Character Design: A Randomized Trial. Journal of Visualized Experiments. https://doi.org/10.3791/74929
Vancouver
Deng M. Structured Text-to-Image Artificial Intelligence–Assisted Instruction for Miao Embroidery–Based Guofeng Character Design: A Randomized Trial. Journal of Visualized Experiments. 2026. doi:10.3791/74929
BibTeX
@article{Deng2026,
author = {Deng, Mingyu},
title = {Structured Text-to-Image Artificial Intelligence–Assisted Instruction for Miao Embroidery–Based Guofeng Character Design: A Randomized Trial},
journal = {Journal of Visualized Experiments},
year = {2026},
number = {237},
doi = {10.3791/74929}
}
EUVIMED