Google’s new skin tone scale can help solve racial bias in AI

Doğa Yurduneri -

Google unveiled a new skin tone scale to address issues of racial bias in AI products and other related technology. Created in collaboration with Harvard professor and sociologist Ellis Monk, the 10-tone Monk Skin Tone Scale comes with a wider range of human skin tones to better reflect people of color.

  • Previously companies used to classify skin color, collect data, train machine vision software, and assess technologies like facial recognition via the 6-tone Fitzpatrick Skin Type Scale. However, AI products trained with this scale mostly represented bias toward lighter skin tones due to the scale's limited range. The tech giant has developed the new 10-tone scale to address this problem.

Zoom In: Google has already started filtering Image search results based on Monk’s scale and aims to integrate it into other products in the coming months. Google also made the Monk Skin Tone Scale data available for use by anyone under a Creative Commons license.