Teaching algorithms about skin colors – Harvard Gazette

The Fitzpatrick scale has been in the open domain for decades and was primarily used by the tech industry for use in machine learning and artificial intelligence. People need to annotate images or videos to train machine learning algorithms for things like face detection.

“Instead of using this Fitzpatrick scale, which isn’t designed to classify different skin tones among different groups of people, you would use the Monk Skin Tone scale, which gives you a more detailed understanding,” Monk said. He chose 10 color points to better represent the range of skin tones, opting out of a wider range as larger numbers would cause problems with annotation. Extensive fieldwork on skin tone and colorism in the US and Brazil formed the basis for the final color selection.

Monk’s partnership with Google began a few years ago when he was approached by his Research Responsible AI team, and over time the work turned into finding an alternative to the Fitzpatrick scale, the sociologist said. Then the team discovered that Monk had developed his own shell.

Google already uses the Monk Skin Tone Scale when searching for images. For example, when users search for makeup looks, they can narrow those searches by skin color to find more relevant matches. The tech giant also used the scale to update its own Real Tone filters (an initiative it had already launched to tackle the skin tone problem) on its Pixel cameras so they can capture a wider, more realistic palette of tones.

“There have also been some refinements to look for where people have complained that if you search for certain images in the past, you only get certain types of people back,” Monk said. “For example, if you search for ‘cute babies’, you can get a completely homogeneous set of babies that basically all look white. That is not an inclusive experience for people at all.”

Google says the scale will be helpful in checking the diversity of results for problems with algorithms. “This is a long-standing problem, not just at Google, but across the tech industry,” Monk says.

Monk plans to publish research and test the scale on a diverse group of people. He and the Google team did research together to validate the scale in the US, which was an extension of his own work.

Developers who don’t intentionally design their products to work well across the spectrum of skin tones will have products that don’t work as well, Monk added. The new scale is open source so all companies around the world can make their products more inclusive.

“I hope other companies, not just Google, take on the work of checking and making sure that their products are designed to work equally well for the entire range of skin tones, whether that be using the Monk Skin Tone Scale or not,” said monk. “I hope they do that job.”

Leave a Comment

Your email address will not be published.