ABSTRACT With the rapid proliferation of generative artificial intelligence (GAI) technologies, educational institutions worldwide are actively seeking effective approaches to cultivate AI literacy. However, current practises in AI education reveal a substantial gap: whilst most students rely on AI tools in their learning processes, a considerable proportion report insufficient AI knowledge and skills. This paradox of ‘widespread use yet limited understanding’ underscores a fundamental challenge in AI education—the ‘black‐box’ problem, which prevents learners from fully grasping the mechanisms and limitations of AI systems. To address this educational dilemma, the present study integrates generative AI with explainable artificial intelligence (XAI) techniques to develop an innovative visualised instructional tool. The system employs Unity3D as the front‐end development platform and integrates ComfyUI as the back‐end generative environment. Three core features—prompt weight visualisation, stepwise generation display, and VAE decoding comparison—were designed to enhance process transparency. An empirical study was conducted using a single‐group pretest‐posttest design with participants possessing information‐related backgrounds. Results demonstrated significant improvements in three domains: understanding of AI fundamentals, mastery of system functions, and attitudes towards AI usage. The findings confirm that visualisation‐based tools can effectively enhance learners' comprehension of AI, particularly in understanding technical functionalities. By transforming the complex generative mechanisms of AI into observable and interpretable content, this study contributes practical design insights and theoretical grounding for the development of AI educational tools, ultimately advancing the promotion of AI literacy in education.
Lin et al. (2026) studied this question.