![\"\"](\"https:\/\/wp.technologyreview.com\/wp-content\/uploads\/2021\/04\/BeautyScoreFiltersFinal2.jpg\")
Joan Wongby<\/p>\n\n\n\nThe most widespread use of augmented reality isn\u2019t in gaming: it\u2019s the face filters on social media. The result? A mass experiment on girls and young women.Joan Wongby<\/p>\n\n\n\n
Veronica started using filters to edit pictures of herself on social media when she was 14 years old. She remembers everyone in her middle school being excited by the technology when it became available, and they had fun playing with it. \u201cIt was kind of a joke,\u201d she says. \u201cPeople weren\u2019t trying to look good when they used the filters.\u201d<\/p>\n\n\n\n
But her younger sister, Sophia, who was a fifth grader at the time, disagrees. \u201cI definitely was\u2014me and my friends definitely were,\u201d she says. \u201cTwelve-year-old girls having access to something that makes you not look like you\u2019re 12? Like, that\u2019s the coolest thing ever. You feel so pretty.\u201d<\/p>\n\n\n\n
When augmented-reality face filters first appeared on social media, they were a gimmick. They allowed users to play a kind of virtual dress-up: change your face to look like an animal, or suddenly grow a mustache, for example.<\/p>\n\n\n\n
<\/a>Podcast: In the AI of the Beholder<\/a><\/p>\n\n\n\n Computers are ranking the way people look\u2014and the results are influencing the things we do, the posts we see, and the way we think.<\/p>\n\n\n\n Today, though, more and more young people\u2014and especially teenage girls\u2014are using filters that \u201cbeautify\u201d their appearance and promise to deliver model-esque looks by sharpening, shrinking, enhancing, and recoloring their faces and bodies. Veronica and Sophia are both avid users of Snapchat, Instagram, and TikTok, where these filters are popular with millions of people.<\/p>\n\n\n\n “The beauty filter sort of changes certain things about your appearance and can fix certain parts of you.”<\/p><\/blockquote>\n\n\n\n Through swipes and clicks, the array of face filters enable them to adjust their own image, and even sift through different identities, with new ease and flexibility.<\/p>\n\n\n\n Veronica, now 19, scrolls back to check pictures from the time on her iPhone. \u201cWait,\u201d she says, stopping on one. \u201cOh yeah … I was definitely trying to look good.\u201d She shows me a picture of a glammed-up version of herself. She looks seductive. Her eyes are wide, lips slightly parted, and her skin looks tanned and airbrushed. \u201cThat\u2019s me when I\u2019m 14<\/em>,\u201d Veronica says. She seems distressed by the picture. Still, she says, she\u2019s using filters almost every day.<\/p>\n\n\n\n \u201cWhen I\u2019m going to use a face filter, it\u2019s because there are certain things that I want to look different,\u201d she explains. \u201cSo if I\u2019m not wearing makeup or if I think I don\u2019t necessarily look my best, the beauty filter sort of changes certain things about your appearance and can fix certain parts of you.\u201d<\/p>\n\n\n\n The face filters that have become commonplace across social media are perhaps the most widespread use of augmented reality. Researchers don\u2019t yet understand the impact that sustained use of augmented reality may have, but they do know there are real risks\u2014and with face filters, young girls are the ones taking that risk. They are subjects in an experiment that will show how the technology changes the way we form our identities, represent ourselves, and relate to others. And it\u2019s all happening without much oversight.<\/p>\n\n\n\n Beauty filters are essentially automated photo editing tools that use artificial intelligence and computer vision to detect facial features and change them.<\/p>\n\n\n\n They use computer vision to interpret the things the camera sees, and tweak them according to rules set by the filters\u2019 creator. A computer detects a face and then overlays an invisible facial template consisting of dozens of dots, creating a sort of topographic mesh. Once that has been built, a universe of fantastical graphics can be attached to the mesh. The result can be anything from changing eye colors to planting devil horns on a person\u2019s head.<\/p>\n\n\n\n Computers are ranking the way people look\u2014and the results are influencing the things we do, the posts we see, and the way we think.<\/p>\n\n\n\n These real-time video filters are a recent advance, but beauty filters more broadly are an extension of the decades-old selfie phenomenon. The movement is rooted in Japanese \u201ckawaii\u201d culture, which obsesses over (typically girly) cuteness, and it developed when purikura\u2014photo booths that allowed customers to decorate self-portraits\u2014became staples in Japanese video arcades in the mid-1990s. In May of 1999, Japanese electronics manufacturer Kyocera released the first mobile phone with a front-facing camera, and selfies started to break out to the mainstream.<\/p>\n\n\n\n The rise of MySpace and Facebook internationalized selfies in the early 2000s, and the launch of Snapchat in 2011 marked the beginning of the iteration that we see today. The app offered quick messaging through pictures, and the selfie was an ideal medium for visually communicating one\u2019s reactions, feelings, and moods. In 2013, Oxford Dictionaries selected \u201cselfie\u201d as the word of the year, and by 2015 Snapchat had acquired the Ukrainian company Looksery and released the \u201cLenses\u201d feature, much to the delight of Veronica\u2019s middle school clique.<\/p>\n\n\n\n Filters are now common across social media, though they take different forms. Instagram bundles beauty filters with its other augmented-reality facial filters, like those that add a dog\u2019s ears and tongue to a person\u2019s face. Snapchat offers a gallery of filters where users can swipe through beauty-enhancing effects on their selfie camera. TikTok\u2019s beauty filter, meanwhile, is part of a setting called \u201cEnhance,\u201d where users can enable a standard beautification on any subject.<\/p>\n\n\n\n And they are incredibly popular. Facebook and Instagram alone claim that over 600 million people have used at least one of the AR effects associated with the company\u2019s products: a spokesperson said that beauty filters are a \u201cpopular category\u201d of effects but would not elaborate further. Today, according to Bloomberg, almosta fifth of Facebook\u2019s employees<\/a>\u2014about 10,000 people\u2014 are working on AR or VR products, andMark Zuckerberg recently told The Information<\/a>, \u201cI think it really makes sense for us to invest deeply to help shape what I think is going to be the next major computing platform, this combination of augmented and virtual reality.\u201d<\/p>\n\n\n\n They are subjects in an experiment that will show how the technology changes the way we form our identities, represent ourselves, and relate to others.<\/p><\/blockquote>\n\n\n\n Snapchat boasts its own stunning numbers. A spokesperson said that \u201c200 million daily active users play with or view Lenses every day to transform the way they look, augment the world around them, play games, and learn about the world,\u201d adding that more than 90% of young people in the US, France, and the UK use the company\u2019s AR products.<\/p>\n\n\n\n Another measure of popularity might be how many filters exist. The majority of filters on Facebook\u2019s various products are created by third-party users, and in the first year its tools were available, more than 400,000 creators<\/a> released a total of over 1.2 million effects. By September 2020, more than 150 creator accounts had each passed the milestone of 1 billion views.<\/p>\n\n\n\n Face filters on social media might seem technologically unimpressive compared with some other uses of AR, but Jeremy Bailenson, the founding director of Stanford University\u2019s Virtual Human Interaction Lab, says the real-time puppy filters are actually quite a technological feat.<\/p>\n\n\n\n \u201cIt\u2019s hard to do that technically,\u201d he says. But thanks to neural networks, AI can now help achieve the kind of data processing required for real-time video altering. And the way it\u2019s taken off in recent years surprises even longtime researchers like him.<\/p>\n\n\n\n Many people enjoy filters and lenses\u2014both as users and creators. Caroline Rocha, a makeup artist and photographer, says that social media filters\u2014and Instagram\u2019s in particular\u2014provided her a lifeline at a crucial moment. In 2018, she was at a personal low point: someone very dear to her had died, and then she suffered a stroke that resulted in temporary paralysis of her leg and permanent paralysis of her hand. Things got so overwhelming that she attempted suicide.<\/p>\n\n\n\n \u201cI just wanted to come out of my reality,\u201d she says. \u201cMy reality was dark. It was deep. I passed my days inside four walls.\u201d Filters felt like a breakthrough. They gave her \u201cthe chance to travel \u2026 to experiment, to try on makeup, to try a piece of jewelry,\u201d she says. \u201cIt opened a big window for me.\u201d<\/p>\n\n\n\n She had studied art history in school, and Instagram filters felt like a deeply human and artistic world, full of opportunity and connection. She became friends with AR creators whose aesthetic spoke to her. Through that, she became a \u201cfilters influencer,\u201d though she says she hates that term: she would try different filters and critique them for a growing audience of followers. Eventually, she started creating filters herself.<\/p>\n\n\n\n Rocha became connected with creators likeMarc Wakefield<\/a>, an artist and AR designer who specializes in dark, fantastical effects. (One of his hits is \u201cHole in the Head<\/a>,\u201d in which a see-through hole replaces the subject\u2019s face.) The community was \u201cso close and so helpful,\u201d she says\u2014\u201cbeautiful,\u201d even. She had no technical expertise when she started creating AR effects, and spent hours poring over help documents with help from others.<\/p>\n\n\n\n Her first viral filter was called \u201cAlive\u201d: it overlaid the electrical pulse of a heartbeat right across the face of its subject. After a moment, the line distorts into a heart that encircles one eye before flashes of colored light illuminate the screen. Rocha says Alive was an homage to her own story of mental illness. <\/p>\n\n\n\n Rocha\u2019s experience is not unusual: many people enjoy the playfulness of the technology. Facebook describes AR effects as a way to \u201cmake any moment more fun to share,\u201d while Snapchat says the goal of Lens \u201cis to provide fun and playful creative effects that allow our community to express themselves freely.\u201d<\/p>\n\n\n\n But Rocha has changed her view. This artistic conception of filters now seems idealistic to her, not least because it is not necessarily representative of how the majority of people use filters. Artistic or funny filters may be popular, but they are dwarfed by beauty filters.<\/p>\n\n\n\n Life online for women is toxic and filled with hate and sexism. Some activists say it\u2019s time to reimagine how the whole thing works.<\/p>\n\n\n\n Facebook and Snapchat were both hesitant to provide any data breaking out filters that are solely appearance enhancing from those that are more novel. Facebook\u2019s creators categorize their own filters into 17 ambiguous buckets, whose names include \u201cAppearance,\u201d \u201cSelfies,\u201d \u201cMoods,\u201d and \u201cCamera styles.\u201d \u201cAppearance\u201d is in the top 10 most popular categories, said the Facebook spokesperson, but refused to elaborate further.<\/p>\n\n\n\n Rocha says she sees many women on social media using filters nonstop. \u201cThey refuse to be seen without these filters, because in their mind they think that they look like that,\u201d she says. \u201cIt became, for me, a bit sick.\u201d<\/p>\n\n\n\n In fact, she struggled with it herself. \u201cI\u2019ve always fought against this kind of fakeness,\u201d she says, but \u201cI\u2019d say, \u2018Okay, I have to change my picture. I have to make my nose thinner and give myself a big lip because I feel ugly.\u2019 And I was like, \u2018Whoa, Whoa, no, I\u2019m not like that. I want to feel beautiful without changing these things.\u2019\u201d<\/p>\n\n\n\n She says the beauty-obsessed culture of AR filters has become increasingly disappointing: \u201cIt has changed because, in my point of view … the new generation of creators just want money and fame.\u201d<\/p>\n\n\n\n \u201cThere is a bad mood in the community,\u201d she says. \u201cIt\u2019s all about fame and number of followers, and I think it\u2019s sad, because we are making art, and it\u2019s about our emotions \u2026 It\u2019s very sad what\u2019s happening right now.\u201d<\/p>\n\n\n\n \u201cI don\u2019t think it\u2019s just filtering your actual image. It\u2019s filtering your whole life.\u201d<\/p><\/blockquote>\n\n\n\n Veronica, the teenager, sees the same patterns. \u201cIf someone is completely portraying themselves in one filter and has only posted photos in a filter meeting all of the beauty standards and gaining followers and making money off of the beauty standard that we have right now\u2014I don\u2019t know if that\u2019s, like, genius or if that\u2019s terrible,\u201d she says.<\/p>\n\n\n\n Claire Pescott is a researcher at the University of South Wales who studies the behavior of preteens on social media. In focus groups, she\u2019s observed a gender difference when it comes to filters. \u201cAll of the boys said, \u2018These are really fun. I like to put on these funny ears, I like to share them with my friends and we have a laugh,\u2019\u201d she says. Young girls, however, see AR filters primarily as a tool for beautification: \u201c[The girls] were all saying things like, \u2018I put this filter on because I have flawless skin. It takes away my scars and spots.\u2019 And these were children of 10 and 11.\u201d<\/p>\n\n\n\n \u201cI don\u2019t think it\u2019s just filtering your actual image,\u201d she says. \u201cIt\u2019s filtering your whole life.\u201d<\/p>\n\n\n\n And this change is only just beginning. AR filters on social media are part of a rapidly growing suite of automated digital beauty technologies<\/a>. The app Facetune has been downloaded over 60 million times and exists simply for easy video and photo editing.Presets are a recent phenomenon<\/a> in which creators\u2014and established influencers in particular\u2014create and sell custom filters in Adobe Lightroom. Even Zoom has a \u201ctouch up my appearance\u201d feature that gives the appearance of smoother skin in video calls. Many have heralded<\/a> the option to buff your appearance as a low-effort savior during the pandemic.<\/p>\n\n\n\n During our conversations, I asked Veronica to define what an \u201cInstagram Face\u201d looks like. She replied quickly and confidently: \u201cSmall nose, big eyes, clear skin, big lips.\u201d<\/p>\n\n\n\n This aesthetic relies on categories of AR effects called \u201cdeformation\u201d and \u201cface distortion.\u201d As opposed to the Zoom-like touch-up that simply blends skin tones or saturates eye color, distortion effects allow creators to easily change the shape and size of certain facial features, creating things like a \u201cbigger lip,\u201d a \u201clifted eyebrow,\u201d or a \u201cnarrower jaw,\u201d according to Rocha.<\/p>\n\n\n\n Teenagers Sophia and Veronica say they prefer distortion filters. One of Sophia\u2019s favorites makes her look like singer and influencer Madison Beer<\/a>. \u201cIt has these massive lashes that make my eyes look beautiful. My lips triple in size and my nose is tinier,\u201d she says. But she\u2019s cautious: \u201cNobody looks like that unless you are Madison Beer or someone who has a really, really good nose job.\u201d<\/p>\n\n\n\n Veronica\u2019s \u201cideal\u201d filter, meanwhile, is a distortion filter called Naomi Beauty on Snapchat, which she says all her friends use. \u201cIt is one of the top filters for two reasons,\u201d she says. \u201cIt clears your skin and it makes your eyes huge.\u201d<\/p>\n\n\n\n There are thousands of distortion filters available on major social platforms, with names like La Belle, Natural Beauty, and Boss Babe. Even the goofy Big Mouth on Snapchat, one of social media\u2019s most popular filters, is made with distortion effects.<\/p>\n\n\n\n In October 2019, Facebook banned distortion effects because of \u201cpublic debate about potential negative impact.\u201d Awareness of body dysmorphia was rising, and a filter calledFixMe, which allowed users to mark up their faces as a cosmetic surgeon might, had sparked a surge of criticism<\/a> for encouraging plastic surgery. But in August 2020, the effects were re-released with a new policy banning filters that explicitly promoted surgery. Effects that resize facial features, however, are still allowed. (When asked about the decision, a spokesperson directed me toFacebook\u2019s press release from that time<\/a>.)<\/p>\n\n\n\n When the effects were re-released, Rocha decided to take a stand and began posting condemnations of body shaming online. She committed to stop using deformation effects herself unless they are clearly humorous or dramatic rather than beautifying and says she didn\u2019t want to \u201cbe responsible\u201d for the harmful effects some filters were having on women: some, she says, have looked into getting plastic surgery that makes them look like their filtered self.<\/p>\n\n\n\n Krista Crotty is a clinical education specialist at the Emily Program, a leading center on eating disorders and mental health based in St. Paul, Minnesota. Much of her job over the past five years has focused on educating patients about how to consume media in a healthier way. She says that when patients present themselves differently online and in person, she sees an increase in anxiety. \u201cPeople are putting up information about themselves\u2014whether it\u2019s size, shape, weight, whatever\u2014that isn\u2019t anything like what they actually look like,\u201d she says. \u201cIn between that authentic self and digital self lives a lot of anxiety, because it\u2019s not who you really are. You don\u2019t look like the photos that have been filtered.\u201d<\/p>\n\n\n\n \u201cThere’s just somewhat of a validation when you’re meeting that standard, even if it’s only for a picture.”<\/p><\/blockquote>\n\n\n\n For young people, who are still working out who they are, navigating between a digital and authentic self can be particularly complicated, and it\u2019s not clear what the long-term consequences will be.<\/p>\n\n\n\n \u201cIdentity online is kind of like an artifact, almost,\u201d says Claire Pescott, the researcher from the University of South Wales. \u201cIt\u2019s a kind of projected image of yourself.\u201d<\/p>\n\n\n\n Pescott\u2019s observations of children have led her to conclude that filters can have a positive impact on them. \u201cThey can kind of try out different personas,\u201d she explains. \u201cThey have these \u2018of the moment\u2019 identities that they could change, and they can evolve with different groups.\u201d<\/p>\n\n\n\n But she doubts that all young people are able to understand how filters affect their sense of self. And she\u2019s concerned about the way social media platforms grant immediate validation and feedback in the form of likes and comments. Young girls, she says, have particular difficulty differentiating between filtered photos and ordinary ones.<\/p>\n\n\n\n Pescott\u2019s research<\/a> also revealed that while children are now often taught about online behavior, they receive \u201cvery little education\u201d about filters. Their safety training \u201cwas linked to overt physical dangers of social media, not the emotional, more nuanced side of social media,\u201d she says, \u201cwhich I think is more dangerous.\u201d<\/p>\n\n\n\n Bailenson expects that we can learn about some of these emotional unknowns from established VR research. In virtual environments, people\u2019s behavior changes with the physical characteristics of their avatar, a phenomenon called the Proteus effect<\/a>. Bailenson found, for example, that people who had taller avatars were more likely to behave confidently than those with shorter avatars. \u201cWe know that visual representations of the self, when used in a meaningful way during social interactions, do change our attitudes and behaviors,\u201d he says.<\/p>\n\n\n\n But sometimes those actions can play on stereotypes. A well-known study from 1988<\/a> found that athletes who wore black uniforms were more aggressive and violent while playing sports than those wearing white uniforms. And this translates to the digital world: one recent study<\/a> showed that video game players who used avatars of the opposite sex actually behaved in a way that was gender stereotypical.<\/p>\n\n\n\n Bailenson says we should expect to see similar behavior on social media as people adopt masks based on filtered versions of their own faces, rather than entirely different characters. \u201cThe world of filtered video, in my opinion\u2014and we haven\u2019t tested this yet\u2014is going to behave very similarly to the world of filtered avatars,\u201d he says.<\/p>\n\n\n\n Considering the power and pervasiveness of filters, there is very little hard research about their impact\u2014and even fewer guardrails around their use.<\/p>\n\n\n\n I asked Bailenson, who is the father of two young girls, how he thinks about his daughters\u2019 use of AR filters. \u201cIt\u2019s a real tough one,\u201d he says, \u201cbecause it goes against everything that we\u2019re taught in all of our basic cartoons, which is \u2018Be yourself.\u2019\u201d<\/p>\n\n\n\n Bailenson also says that playful use is different from real-time, constant augmentation of ourselves, and understanding what these different contexts mean for kids is important.<\/p>\n\n\n\n \u201cEven though we know it\u2019s not real\u2026 We still have that aspiration to look that way.\u201d<\/p><\/blockquote>\n\n\n\n What few regulations and restrictions there are on filter use rely on companies to police themselves. Facebook\u2019s filters, for example, have to go through an approval process that, according to the spokesperson, uses \u201ca combination of human and automated systems to review effects as they are submitted for publishing.\u201d They are reviewed for certain issues, such as hate speech or nudity, and users are also able to report filters, which then get manually reviewed.<\/p>\n\n\n\n The company says it consults regularly with expert groups, such as the National Eating Disorders Association and the JED Foundation, a mental-health nonprofit.<\/p>\n\n\n\n “We know people may feel pressure to look a certain way on social media, and we’re taking steps to address this across Instagram and Facebook,” said a statement from Instagram. “We know effects can play a role, so we ban ones that clearly promote eating disorders or that encourage potentially dangerous cosmetic surgery procedures\u2026 And we’re working on more products to help reduce the pressure people may feel on our platforms, like the option to hide like counts.” https:\/\/playlist.megaphone.fm\/?e=MIT3184912056<\/p>\n\n\n\n Facebook and Snapchat also label filtered photos to show that they\u2019ve been transformed\u2014but it\u2019s easy to get around the labels by simply applying the edits outside of the apps, or by downloading and reuploading a filtered photo.<\/p>\n\n\n\n Labeling might be important, but Pescott says she doesn\u2019t think it will dramatically improve an unhealthy beauty culture online.<\/p>\n\n\n\n \u201cI don\u2019t know whether it would make a huge amount of difference, because I think it\u2019s the fact we\u2019re seeing it, even though we know it\u2019s not real. We still have that aspiration to look that way,\u201d she says. Instead, she believes that the images children are exposed to should be more diverse, more authentic, and less filtered.<\/p>\n\n\n\n There\u2019s another concern, too, especially since the majority of users are very young: the amount of biometric data that TikTok, Snapchat and Facebook have collected through these filters. Though both Facebook and Snapchat say they do not use filter technology to collect personally identifiable data, a review of their privacy policies shows that they do indeed have the right to store data from the photographs and videos on the platforms. Snapchat\u2019s policy says that snaps and chats are deleted from its servers once the message is opened or expires, but stories are stored longer. Instagram stores photo and video data as long as it wants or until the account is deleted; Instagram also collects data on what users see through its camera.<\/p>\n\n\n\n Meanwhile, these companies continue to concentrate on AR. In a speech made to investors in February 2021, Snapchat co-founder Evan Spiegel said \u201cour camera is already capable of extraordinary things. But it is augmented reality that\u2019s driving our future\u201d, and the company is \u201cdoubling down\u201d on augmented reality in 2021, calling the technology \u201ca utility\u201d.<\/p>\n\n\n\n And while both Facebook and Snapchat say that the facial detection systems behind filters don\u2019t connect back to the identity of users, it\u2019s worth remembering that Facebook\u2019s smart photo tagging feature\u2014which looks at your pictures and tries to identify people who might be in them\u2014was one of the earliest large-scale commercial uses of facial recognition. And TikTok recently settled for $92 million<\/a> in a lawsuit that alleged the company was misusing facial recognition for ad targeting. A spokesperson from Snapchat said “Snap’s Lens product does not collect any identifiable information about a user and we can’t use it to tie back to, or identify, individuals.”<\/p>\n\n\n\n And Facebook in particular sees facial recognition as part of it\u2019s AR strategy. In a January 2021blog post<\/a> titled \u201cNo Looking Back,\u201d Andrew Bosworth, the head of Facebook Reality Labs, wrote: \u201cIt\u2019s early days, but we\u2019re intent on giving creators more to do in AR and with greater capabilities.\u201d The company\u2019s planned release of AR glasses is highly anticipated, and it has already teased<\/a> the possible use of facial recognition as part of the product.<\/p>\n\n\n\n In light of all the effort it takes to navigate this complex world, Sophia and Veronica say they just wish they were better educated about beauty filters. Besides their parents, no one ever helped them make sense of it all. \u201cYou shouldn\u2019t have to get a specific college degree to figure out that something could be unhealthy for you,\u201d Veronica says.<\/p>\n","protected":false},"excerpt":{"rendered":" The most widespread use of augmented reality isn\u2019t in gaming: it\u2019s the face filters on social media. The result? A mass experiment on girls and young women.Joan Wongby Tate Ryan-Mosley archive page Veronica started using filters to edit pictures of herself on social media when she was 14 years old. She remembers everyone in her<\/p>\nThe rise of selfie culture<\/strong><\/h2>\n\n\n\n
A “beautiful” community<\/strong><\/h2>\n\n\n\n
Reality distortion field<\/strong><\/h2>\n\n\n\n
\u201cI wish I was wearing a filter right now”<\/strong><\/h2>\n\n\n\n
![\"\"](\"https:\/\/wp.technologyreview.com\/wp-content\/uploads\/2021\/04\/IMG_0890.jpg?w=960\")
Selfie regulation<\/strong><\/h2>\n\n\n\n