In order to help users access a diverse array of images more easily, Pinterest is today rolling out a beta feature that makes it possible to filter results for beauty tips by skin tone. Search for a term like “red lipstick,” and you now have the option to select whether you want a lipstick that works well with a light or dark skin tone. Users will be able to select from one of four skin tone ranges.
In a blog post announcing the new feature, Pinterest engineer Laksh Bhasin said that 70 percent of users rely on Pinterest to “discover and save everyday looks and styles they want to try.”
In an interview with Wired announcing the feature, Pinterest head of diversity and inclusion Candice Morgan noted that when she entered a broad search term like “hair ideas” on Pinterest, she would be presented predominantly with images featuring white women as models.
Bhasin detailed in the blog post how the company used machine learning to bring the new feature to life. Pinterest also turned to ModiFace, a company that has created custom AR applications for beauty companies and was acquired by L’Oréal just last month.
June 5th: The AI Audit in NYC
Join us next week in NYC to engage with top executive leaders, delving into strategies for auditing AI models to ensure fairness, optimal performance, and ethical compliance across diverse organizations. Secure your attendance for this exclusive invite-only event.
Pinterest and ModiFace had to test out a number of different algorithms in order to come up with a system that wouldn’t, for example, label a photo of a model with light skin as containing a model with dark skin just because the image had a lot of shadows. Pinterest relied on Sofia, the in-house computation platform that it built in 2016, to refine the algorithms.
Another challenge Pinterest knew it would face was that not all users who classified themselves as having “medium-dark skin” would have the same skin tone. So the company decided to display the skin tones a user can select in quadrant form so that they can see which skin tones are grouped under “light-skinned” and “dark-skinned.” Presumably, the platform will add more filter options over time as its algorithms mature.
It’s also not clear how many beauty search terms this new filter will appear under. The company only said that the search feature would show up for “a predetermined list of common hair and beauty queries.”
Bhasin also noted in the blog post that Pinterest will not be storing data about which skin tone users select and that the information won’t be used for ad targeting. Facebook was the subject of a Pro Publica investigation in 2016, when journalists discovered the company was letting advertisers exclude users by race.
A spokesperson for Pinterest did not say when the inclusive discovery feature would be rolling out of beta but noted that it’s available to everyone on the web starting today.
Pinterest last reported its total number of monthly active users in September 2017, when the company announced it had surpassed 200 million MAUs. More than 100 billion images have been pinned by Pinterest users, to date.
