Designing for acceptance and circularity: An explainable machine-learning framework for human–AI co-created sustainable fashion

Sustainable fashion design requires early-stage decisions that jointly consider consumer acceptance and material efficiency. This study develops an explainable machine-learning framework for evaluating human–AI co-created sustainable fashion through four linked modules: garment-design records, expert evaluations, a factorial consumer experiment, and production waste data. Using 4200 design records, 350 expert-rated samples, 864 consumers, and 426 production patterns, the study separates displayed source attribution from actual design provenance and evaluates prediction under grouped validation. Results show that human–AI co-creation attribution and sustainability information increase purchase intention, whereas a 35% green premium reduces acceptance; however, sustainability information does not significantly offset the premium penalty. Logistic regression is the most consistent in its purchase-intention screening performance, and the ridge regression is strong in predicting waste-rate. The results situate explainable machine learning as a tool that can be used to assist designers in their decision-making process related to sustainable fashion design, rather than a designer replacement.