An Empirical Study of AI Agent–Enhanced Instruction in Vocational Graphic Design Courses

Abstract

Amid the rapid development of generative artificial intelligence, design-oriented courses in vocational education face the dilemma that students possess strong technical skills but weak design thinking. Existing AI-assisted teaching approaches primarily focus on outcome generation or post hoc evaluation, lacking precise support for cognitive processes in the early stages of creation. To address this issue, this study constructed an AI Agent teaching assistant for pre-creative cognitive diagnosis based on the Coze platform and conducted a three-week quasi-experimental study to examine the impact of AI Agent intervention on students’ comprehensive design performance, perceived design competence, and learning engagement. Using a mixed-methods approach that integrated interviews, questionnaires, and instructional experiments, this study empirically analyzed the teaching effectiveness of the AI assistant and students’ technology acceptance mechanisms. The results indicate that the experimental group significantly outperformed the control group in the final comprehensive design project (t(64) = 8.54, p < 0.001, Cohen’s d ≈ 2.13). The AI assistant significantly enhanced students’ design competence, academic performance, and learning engagement, while demonstrating strong perceived usefulness and affective empowerment value.