Google adopts 10-tone skin scale for AI model evaluation
Google replaced the six-tone Fitzpatrick scale with a 10-tone scale developed with Harvard researchers to better capture global diversity. The new metric is now integrated into evaluations for products like the Pixel camera and image generation models.
Google has shifted its methodology for evaluating skin tone representation in AI models, moving away from the Fitzpatrick scale. The previous industry standard used six skin tones and was originally designed for healthcare purposes, which caused it to skew toward lighter skin tones and fail to capture global diversity. To address this, Google collaborated with Professor Monk at Harvard, an anthropologist and sociologist, who developed a 10-skin-tone scale based on years of user interviews and research.
Google integrated this 10-tone scale into its evaluations for products like the Pixel camera and image generation models to ensure more intentional and accurate representation. This change reflects a broader effort to improve the holistic evaluation of models, considering not just content but also factors like voice naturalness and serving conditions.