Will AI make the beauty industry less racist?

The beauty industry is rarely recognised for its racial inclusivity. Primarily, because it’s lacking.

Ill-informed in-store consultants ​and ​brands that perpetuate insensitive racial stereotypes​ ​only scrape the surface of the problems that lie within an industry that has never succeeded in catering to non-white customers. But Artificial Intelligence experts and entrepreneurs believe that with recent technological developments, a less racist beauty industry is in the near future.

The merger of AI and cosmetics has given rise to beauty tech companies that are specifically programmed to address race in their functions.

“We’ve trained our computer-vision analysis to work with different ethnicities,” explains Sampo Parkkinen, CEO and co-founder of ​Revieve​, an AI-powered digital beauty advisor. Launched in 2016, Revieve combines selfie analysis with user data to recommend personalised product selections, taking into account skin tone, but also less obvious factors like the proportions of facial features. “The way this becomes evident to the end-consumer is through certain ethnicities having differences in the types of beauty considerations we discover on their selfies, leading to perhaps different types of products being recommended.” Revieve’s approach seeks to fill the gap between customers of colour and the too common cases of white cosmetic consultants that despite certifications, licenses and on-the-job training, still don’t know how to assist them due to an unfamiliarity with the (typically lesser-known) beauty brands aimed at people of colour.

Feeling uncomfortable and underserved at the make-up counter is typical for people of colour, and something the industry has long shrugged off as just the way it is. Brands like Fenty have begun to fill the gaping hole, but the industry is still a long way off from having as many options for customers of colour as they do for white customers and stores seldom have stock of the already limited make-up offerings POC have to choose from.

Some cosmetics stores have finally taken note of the issue and employed their own AI software similar to that of Revieve in the past few years. Sephora, which has been called out for racism in the past ​l​ ast year ​a viral video​ showed two black shoppers tell a white Sephora salesperson they felt they were being racially profiled, to which the employee ignorantly justified herself with “I have tons of black friends”​ w​ as the surprising pioneer of this trend. In 2012, the chain debuted ​Color IQ​, in collaboration with Pantone, an in-store handheld device that scans a customer’s skin to assign them a number corresponding to their shade. As most AI beauty technologies are just emerging, Color IQ is the extent of many customers’ experiences with AI.

Color IQ’s success can be attributed to Sephora’s racially diverse customer base being factored into the equation with the end goal of simplifying the shopping journey for POC, a model also practised by ecommerce site My Beauty Matches. CEO Nidhima Kohli founded My Beauty Matches in 2012, having experienced the struggle to find make-up as an Indian woman living in London. “AI in beauty will finally help POC with ‘choice paralysis,’” Kohli says. The site uses AI solution ​Beauty Matching Engine​ to select products from a catalogue of over 300,000, going above and beyond skin colour matching by taking into account other characteristics of POC’s skin, like skin type and necessary SPF.

AI can ease the frustrating task of shopping for make-up as a POC, it can streamline the process and take over for the ignorant salesperson entirely. But one must ask, is that really what customers want? To be surveyed and scanned by a grey box? Ask any make-up junkie; human interaction is a valuable component of shopping and a face-to-face conversation is more personal than any AI-generated “personalized routine” could ever be. AI designed to recognise race should exist in the beauty industry, no doubt. But the question of ​how​ it should exist and function within that space is one that should be left up to the POC it plans to target. Beauty brands need to facilitate dialogues with customers of colour to determine if their use of AI is actually helpful, or just invasive.

Speaking to women of colour that shop for make-up regularly has shown me the importance of such conversations. Each has unique ideas on how AI can work towards inclusivity without being a cop-out of addressing the industry’s racism. One approach is to utilize AI as a means of researching products for POC. “AI can help companies collect data on the kinds of concerns their customers have and whether certain races are more likely to be concerned with certain things,” Tamara Jones, a 25-year-old Afro-Caribbean-Latina, says. “I think that could actually help them to offer more relevant products and maybe take chances on lesser-known brands that happen to cater to the things their customers are looking for.”

This strategy was taken up by Proven, a custom skincare startup using AI to analyze​ a lot​ of data. “[Our AI] analyzes more than 100,000 products, 8 million testimonials and reviews from over 4,000 scientific journal articles,” CEO Ming Zhao says. She co-founded Proven last year with CTO Dr Amy Yuan, a computational physicist. Zhao feels that this mass of data “eliminates any ambiguity or ‘trial and error.’”

Still, going down the route of personal data collection institutes a slew of ethical implications. Storing data is dangerous territory and exacerbates discomforts in an already problematic situation, especially with chief tech bro ​Elon Musk inciting AI paranoia​. The issue at hand becomes increasingly relevant as 31% of enterprises plan to begin using AI in the next 12 months, ​according to Adobe​. AI has a history of voiding users’ privacy to collect data and while

the new EU-wide General Data Protection Regulation is requiring companies to ask for your consent on their sites before recording your info, it’s unclear if and how this law would be applied to AI that exists in physical settings, or what protections the UK will implement post-Brexit.

Research firm Gartner ​predicts that from 2020 onwards​, AI will create 2.3 million jobs globally and eliminate 1.8. It’s presumable that most these new jobs will be in tech, designing, developing and implementing AI, whereas the jobs eradicated will be those consisting of tasks that can be performed up to standard or even better by a machine than a human, like colour matching a customer’s skin. But even if the machine’s work is superior to that of the person, it doesn’t excuse the racism it intends to avoid, so instead of looking at AI machines as the solution to the racist beauty industry, we should regard it them a temporary cushion for the transitional period to the diverse industry we hope to achieve by hiring more workers of colour for the jobs created by AI. As we integrate the two industries, one dominated by white women and the other by white tech bros, it’s critical that the emerging beauty tech industry has more than the intention to implement diversity, but is diverse in and of itself.

The need for POC workers in beauty tech was mirrored by the majority of the women I spoke to, who voiced different variations of the sentiment that AI’s potential replacement of beauty professionals feels like a high-tech band-aid for the industry’s racism. “[AI’s] place in the movement is limited and shouldn’t be seen in any way as the beauty industry’s answer to dealing with racial discrimination and the lack of representation of people of colour,” Hannah Fatima, a 21-year-old half-Filipino half-Sri Lankan, said. “Particularly because the issue of race in the beauty industry is deep-rooted in the marginalisation of minorities. The change must come in employing more people of colour within instrumental roles of beauty [businesses] everywhere.”

Hiring POC, especially women, has the added benefit of subverting discriminatory AI, whose faults can be attributed to those of their human programmers. While conducting research at Massachusetts Institute of Technology, of AI that white developers may overlook. Society is at odds over AI—we embrace it, fear it, don’t know anything about it. I fall somewhere between the first two. But regardless of stance, it’s become increasingly clear that

​Joy Buolamwini discovered that the AI facial recognition systems of Microsoft, Face++ and IBM couldn’t recognize her face due to her dark skin, but worked for her lighter-skinned friends. Buolamwini studied the systems using 1000 images, the results of which proved that all three systems lacked accuracy in identifying dark-skinned females. IBM replicated Buolamwini’s research and adjusted their AI so that it recognizes dark-skinned females with a success rate of 96.5%. POC are able to approach the relationship between AI and race differently​—they bring a hyperawareness to ​noticing the racist tendencies while AI carries the opportunity to have a positive impact, the weight of it lies in more than just scanning skin and collecting data. POC need to be employed in beauty tech in order to make any major difference.

AI isn’t the be-all and end-all to dismantling the beauty industry’s racism, but if it puts POC both behind the scenes and at the forefront, it can make real progress.

Words by Sarah Kearns