Visual Node Programming and Bias Awareness in AI: Teaching Ethical AI Through ComfyUI

ComfyUI, an open-source visual node-based interface for AI image generation, offers educators a practical platform for teaching both the technical and the ethical dimensions of artificial intelligence. This paper argues that its modular, transparent workflow design can make complex AI concepts accessible to students without programming experience while serving as a hands-on laboratory for bias exploration. The argument is grounded in a theoretical framework that draws together critical AI literacy, constructionism, critical technical practice, and seamful design, and it is tested against its strongest published counter-argument that pedagogy should teach students to work with, rather than see inside, the black box (Bearman and Ajjawi, 2023). Reviewing the empirical literature on visual programming, including its sceptical strands, the paper reframes the claimed benefits of node-based learning as testable hypotheses and sets out the design of a workshop-based study through which they will be examined with art and fashion students in the 2026–27 academic year. Drawing on recent research on bias in machine learning and text-to-image generation, it outlines pedagogical strategies such as comparative prompting, custom bias-exploration nodes, and cross-cultural output analysis through which students can observe, analyse, and mitigate bias in generative models. It concludes that visual node programming can move AI education beyond technical proficiency and platform prompting toward critical, ethically informed practice.