Generative AI in creative education: baseline-linked evidence on learning processes, artistic performance, and authorship

Abstract

Generative artificial intelligence is increasingly shaping creative learning in higher education art courses, yet its educational value remains contested when creative support, output generation, authorship, and skill development become intertwined. This study investigates AI-supported creative learning in an undergraduate art course. The study used linked data from 133 students, including prior course performance, final creative-performance assessment, and post-project questionnaire responses. AI use was operationalized through two dimensions: assistive use for ideation, refinement, process management, and technique learning, and content-generative use involving AI contributions to creative output. The analysis combined baseline-adjusted outcome models, performance-change models, propensity-score adjustment, AI-use intensity comparisons, and robustness checks across alternative AI-use definitions. The findings show a clear process-outcome decoupling pattern. AI use was associated with stronger perceived support and skill transfer, while final artistic performance and baseline-to-final performance change showed no corresponding improvement. Content-generative AI use was associated with lower evaluated authorship, suggesting that AI-shaped artifacts may reduce the visibility of students' creative decision-making in final outcomes. The study contributes a differentiated account of AI in creative education and argues that responsible AI integration should focus on how students use, document, justify, and reflect on AI within the creative process.