According to Wired, the idea of diversifying search results came about when Pinterest's first head of diversity and inclusion Candice Morgan used the website's search engine to look for hair ideas. While the platform learned from her saved pins and showed recommendations that point to hair tutorials for black hair, searching for "hair ideas" showed her lots of beachy waves and updos that don't suit 4c (tightly coiled) curls. "We were thinking," she said, "how we can look at our database of pins and make sure we were in some way matching and labeling content across different groups of people?"
As Pinterest discovered, though, it's not easy developing a technology that can detect skin tones. The company worked with ModiFace, the L'Oreal-owned augmented reality beauty app maker, to create an algorithm that can accurately detect the skin color of people in images. The team used ModiFace's third-party Face AI library and fed it more and more examples until it's no longer getting tripped up by highlights and shadows.
Pinterest admits that the algorithm is still far from perfect, so it keeps feeding it new pins to learn from. Unfortunately, the company has only rolled it out to a select number of users as part of its beta testing period. If you happen to have access to it -- and you'll only know that you do if you try to search relevant queries -- Pinterest says you don't have to worry about your privacy. It won't save your preference and won't serve you ads based on the skin tone icon you clicked.