Study finds CG skin and hair algorithms favor white features

A new paper from Theodore Kim, Alexa Schor, Julian Posada, and Alka V. Menon presents the first systematic review of how human depiction is handled in the top computer graphics conference (SIGGRAPH) and journal (ACM Transactions on Graphics). The study confirms that algorithms widely perceived as universal for rendering photorealistic humans are actually biased toward historically hegemonic characteristics. For skin, algorithms described as generic for 'human skin' are in fact designed for translucent, high-albedo materials—i.e., white skin. This creates a conceptual binarization where white skin becomes the computational substrate for all skin, imposing a hierarchical assumption that all skin descends from the math and physics of white skin. Hair algorithms follow a similar pattern, treating 'human hair' as rods, wires, and threads—analogous to straight hair. The first examples of computer-generated Type 4 hair (tightly coiled) only appeared after the murder of George Floyd in 2020.

The authors introduce two new conceptual labels: 'McDaniels Methods' for algorithms that reinforce racial hierarchy under a false cover of diversity, and 'Durald Methods' for algorithms co-designed with the people being depicted. The paper argues that these biases are not just social oversights but stem from foundational mathematical choices in computer graphics. The findings point to neglected avenues for future research, including developing algorithms that explicitly model diverse skin and hair types from the ground up. For professionals in graphics, gaming, and VFX, this study is a wake-up call that 'generic' rendering pipelines may perpetuate exclusion, and that true diversity requires rethinking core algorithmic assumptions.