Google has recently announced that it is developing a new standard method for categorizing users’ skin color to end the bias of existing methods and prevent discrimination against people of color.
to the Reuters reportThere is now a standard scale called the Fitzpatrick Scale that classifies skin color into six categories. Technology companies now use the same scale to categorize people and design their products, including face recognition systems or smartwatch heart rate sensors.
Critics say the Fitzpatrick scale, which includes four categories for white skin and two categories for black and brown, ignores the racial diversity of people of color. Now Google has announced that it is silently working on a new system for skin color classification:
“We are working on alternatives and more comprehensive methods that can be useful in developing our products. “In this way, we work with scientists, medical professionals and groups associated with people of color.”
Companies know that their products may not work well for groups about whom not enough information is available. Concerns about the Fitzpatrick scale occur when the limitation of the dark skin range leads to the production of technologies that, for example, are suitable for golden-brown skins but not for red-brown skins.
A 2018 study found that facial recognition systems often misdiagnose the gender of dark-skinned people. In April of this year, Facebook researchers tested artificial intelligence to detect diphtheria videos and found that the Fitzpatrick scale clearly could not cover the variety of brown and black skin tones. Facebook finally said it was ready to use the new methods.
Experts at the University of California, San Diego, wrote in an article last year that the Fitzpatrick scale could err in reading the heart rate on smartwatches and give people false confidence. In response to this article, Apple said that the Fitzpatrick scale is one of the methods of testing their heart rate, and the company is testing the capabilities of its clock with other methods.