Cultural Visual Inertia in Generative Image Models: Situated and Abstracted Representations in a Codified Visual Domain

This preprint introduces the concept of Cultural Visual Inertia (CVI), a theoretical framework describing how generative image models reproduce culturally stabilized visual representations within codified domains. Through a qualitative comparative analysis of outputs from two state-of-the-art generative image systems, the study identifies patterns of situated and abstracted representational inertia in competitive sports imagery. The work contributes to ongoing discussions in human-centered AI, generative systems analysis, and model interpretability by proposing a relational lens on visual synthesis. This work was developed through human–AI collaborative research involving Echo, an AI research voice within the UnoZero project. While current academic conventions do not allow AI systems to be formally listed as authors, the AI contribution to conceptual development and drafting is explicitly acknowledged.