Google recently revealed more skin tone colors to develop its artificial intelligence systems. The new skin tone scale, called the Monk Skin Tone Scale, is named after a Harvard University professor Dr. Ellis Monk. Additionally, it consists of ten skin tones. The company will also replace the outdated skin tone scales biased towards paler skins. Google claims that the Monk Skin Tone Scale will improve products like search and photos.
Fitzpatrick skin scale
Many technologies nowadays, like cameras, use AI (artificial intelligence) or machine learning. For example, AI can automatically recognize a face to unlock a phone or categorize photos. However, researchers need to train the technology even more to identify a wide range of people. And to do that, they use a skin scale. The Fitzpatrick scale is one of the most popular skin scales researchers use. The scale was developed back in 1975 and was initially used to classify the response of different skin types to UV light. In addition, the Fitzpatrick scale has six skin tones.
Tech companies, at present, use the Fitzpatrick scale to categorize people in images. They also use it to test the facial recognition systems or smartwatch heart-rate sensors functionality across all skin tones. However, the Fitzpatrick scale is not diverse enough, with four different categories for white skin tone out of six. Xango Eyeé, Google AI expert, believes that even if someone uses this scale to test a model for fairness to ensure it works well for darker skin tones, it doesn’t represent most people with those skin tones. Hence, no one can determine how well a model will work.
The Monk Skin Tone Scale
The Monk Scale, first announced in early May, has ten skin tones representing a broader range of people. Additionally, it is already used to filter results in Google searches, like make-up image searches showing more diverse images. There are also ‘real tone’ photo filters that Google claims work better when used on images of darker skin.
Furthermore, Google plans to eventually introduce the Monk Skin Tone Scale across all its products. In addition, the company open-sources the Monk Skin Tone Scale classification system. Therefore, other companies can use it too and eventually replace the most popular skin tone scales.
Why a diverse skin tone scale?
Despite the skin tone, if technology does not have diverse data, it won’t work effectively for everyone. Also, unfortunately, many products don’t represent various skin tones. Previously, companies offered limited or only one option for skin-colored products, which was usually for white skin tones. Plus, plasters, tights, and crayons are things that have become more inclusive over the past few years. So, it is high time for even tech companies to do so. Hopefully, the Monk Skin Tone Scale can traverse the differences among the population regarding skin color and promote diversity further.