Integrating Large Language Models into Calligraphy and Painting Education: Effects on Learning Outcomes

Abstract

The use of LLM is rapid and we are not yet clear how much they can add to the education of art. This review explores how these systems can affect learning in calligraphy and painting, two studio arts that are deceptively embodied, based on cultural knowledge, visual judgment, and authorship. Recent peer-reviewed literature was synthesized using a structured integrative approach, concerning the use of artificial intelligence in education, the co-creation between human and artificial intelligence, creative agency, feedback, and visual-art practice. In the literature, large language models are reported to open up access to explanation, questioning, comparison and reflective feedback. These are reported or expected benefits; only wider access to explanation and feedback has consistent empirical support, and that support comes mainly from adjacent educational settings rather than from studio classes. Their benefits seem to be most likely when students make an initial intention, experiment with ideas in terms of material practice, and are confirmed by the teacher. There is less evidence of independent mastery of the skills, originality, delayed retention, or transfer once the support is removed. The main risks are: cognitive offloading, culturally generic critique, stylistic convergence, not being aware of who the author is, lack of equitable access and general confidence in fluent responses. The review suggests the use of the framework of the CRAFT-LM: cultural grounding, retained agency, adaptive feedback, formative process assessment, teacher moderation, and learning monitoring. It sets priorities for empirical testing of the discipline without offering an untested evaluation model as a result. Current evidence suggests that LLM could be used as a scaffold rather than a source for creating, assessing, or replacing studio practice, provided that they are used with caution, under the guidance of the teacher, and within explicit and culturally-aware approaches to process-based studio learning.