One of my recurring preoccupations as a teenager was something I thought of as The Strip. In my fantasies, The Strip—which I imagined would be kept somewhere secret and sacred, like maybe the Vatican—featured the face of every single woman in the world, ordered by beauty. The least attractive woman in the world (I didn’t care about the men) was at the far left; the most beautiful woman in the world at far right. If you folded The Strip in half, the crease would split the face of the most average-looking woman in the world right down the middle.
This is where I imagined my face would be, billions of women to my right, billions to my left. Perhaps I would have liked to be to the right of the crease, but even in the realm of fantasy, my priority was making sure I “passed” as pretty enough to be definitively not-ugly. The key wasn’t the not-ugly part, though; it was the definitively part of the equation. If I were at least on the crease, I wouldn’t have to worry about whether there was something more I should be doing—mysterious hair treatments, investment in acne creams—to nudge myself to the right on The Strip. I’d have passed. And there would be objective proof.
The Strip, to the best of my knowledge, does not exist. What does exist are apps like Beauty Test and Beauty Check, GRFace—that GR stands for Golden Ratio—and Face Scanner; Camera Beauty Detector, nFace, Ugly Meter, and Vanity Mirror, and that’s just for Android. Each promises the essence of The Strip: a definitive, quantified measure of your appeal, as measured by the computer, the ultimate objective eye, but without the inherent competitiveness of being measured against other people. Where The Strip would tell you how pretty you were, and how pretty your best friend was, and how pretty you were in relationship to one another, apps have you playing solitaire. Those other, prettier—or not prettier—people are still out there, of course, but with apps they appear absent. It’s you versus the computer. Rather, it’s you versus the rules—a crucial shift in mind-set away from the power that comes from being more beautiful than others.
But if “beauty is in the eye of the beholder,” as the aphorism has it, what happens when the beholder is a computer? What power does that seemingly objective beholder share with the beautiful, or the not beautiful? And what does the ability to be seen through a computer’s eyes say about the desire to be seen by our fellow humans?
Each of the dozens of apps that promise to evaluate your beauty works more or less in the same way: You take or upload a forward-facing photo of yourself, and the computer breaks down the geometry of your face, compares it with a preprogrammed ideal, and then translates that comparison into a score of how you measure up.
The usefulness of computer vision is not in its objectivity but in its ultimate absurdity, which makes it so easy to rebut.
Computers are trained to recognize a face when it detects in an image two parallel points existing above a second point, which itself hovers above a line-like series of points: that is, two eyes, a nose, and a mouth. To recognize a beautiful face as opposed to a merely human face, though, the algorithm needs refinement.
Researchers going back to 500 BCE have recognized a “golden ratio” throughout nature, with seashells, plants, and even DNA constructed so that their proportions conform to a 1.618-to-1 ratio. It’s the basis of Leonardo da Vinci’s Vitruvian Man, what with all those lines drawn over the perfect human body, and it’s the basis of most of the apps’ evaluation. If the computer sees two large points on an oval shape as eyes, and they are situated so that the distance between their outer corners divided by the length of the lips below them equals 1.618, then the computer assumes they are beautiful eyes.
The golden ratio and beauty apps seem made for each other: Computer vision of any sort must rely upon algorithms, and the golden ratio provides a logic for them. It’s in sync with the Aristotelian concept of beauty, in which beauty is presumed to be objectively definable and therefore measurable. Beauty, by this definition, is order, symmetry, and definiteness. Beauty isn’t glamour; it’s math.
Not that you need to know anything about the golden ratio to use the apps. Some of them make their geometric aesthetic evident: GRFace lays a grid over your photo with the ideal measurements listed alongside your measurements. Beauty Check and nFace do something similar, only with dots rather than a grid. Others don’t even refer to the mathematical reasoning behind their results: Vanity Mirror and Face Scanner just give you a score. But no matter the app, there is a score—which, presumably, is what the user is after.
Most of us have wondered how objectively appealing we “really” are, unbiased by intimacy or even courtesy. How do we look without the filter of politeness that prevents most people from saying things like, Yeah, you’re kinda funny looking? If beauty always depends on what other people are seeing in the moment, then we might never be able to identify our own beauty once and for all for ourselves.
Turning beauty into a math problem means that there are plenty of ways to solve beauty—and plenty of ways to get it wrong too. But either way, it gives you a right answer, and a wrong one, to the question of “Am I pretty?” It tempts us with the promise of relief, a way to stop wondering about it. Want to know how pretty you are? Here’s an answer, in hard numeric form. Take it or leave it.
In reading reviews of these apps, though, you find that people don’t want to just take it or leave it. We may be curious to see ourselves with apparently objective computer vision, but our personal experience of beauty is often more elusive than that. Reviews of beauty apps show extraordinary dissatisfaction, and not just with the results. A reviewer who noted that one app “does not work, said I was 100 percent beautiful” gave it as many stars as one who wrote, “I think I look beautiful and it’s saying I’m 20 percent?” Rather, reviewers often seemed to take issue with the very concept of a beauty app itself—even though they sought these apps out and installed them. Judging by some reviews, users were eager to test the apps in order to debunk them, as if to even the score with them for their judginess: I test my hand it’s 55 percent beautiful. I took a picture of my ugly brother and it said he was pretty. I scanned my face and got a 61 percent then scanned the wall and it got a 95, how is a wall prettier than a person. Gave Miss America a 7.8 and me a 9.2.
Scrolling through these reviews from disappointed users—and from the occasional pleased user too—you begin to see a pattern: People want to believe there’s an objective measure of beauty, but they don’t like to think that the measure applies to them. Users are eager to believe that objective assessment is possible, but when they are being assessed, the algorithm must be off, the facial recognition not good enough, the code not quite finished. The app is wrong.
Of course, these apps are self-evidently terrible. Some of them flat-out don’t work, or give differing results to the same picture. Camera Beauty Detector kept telling me I don’t have a face. Beauty Check kept wanting to rate not my face but my armpit (a 53/100, incidentally). And according to Vanity Mirror, my cat is more attractive than I am (which is probably true). Yet we still use them, and review them, and download others just like it. These reviews offer a clue as to why.
Given that we tend to frame women’s relationship with beauty as being self-flagellating, it might be hard to believe that anyone would turn to these apps for joy, and at first glance the displeasure in the reviews seems to reflect this. But pleasure is the essence of these apps, even for users who rated it poorly. The pleasure they provide doesn’t come from their giving an objective answer to how beautiful we are, but in letting us pose the question and then reject the answer if we want to. They give us an opportunity to know for certain that beauty is real and that it matters—while at the same time letting us regard ourselves as an exception. The usefulness of computer vision is not in its objectivity but in its ultimate absurdity, which makes it so easy to rebut.
If you’ve ever been tempted by curiosity or vanity to download one of these math-driven beauty apps, you might think you’re casting a ballot for the Aristotelian concept of beauty. But if the joy of these apps lies not in their objectivity but in the tension between our own vision and the computer’s, maybe we should look to a different philosopher to explain their appeal: Immanuel Kant.
The Kantian concept of beauty rests not on measurable parameters and harmonious ratios but on one’s aesthetic judgment, which struggles to free itself from merely subjective taste. We may be swayed to find something beautiful because it moves us emotionally and personally, but this is always in tension with the detachment required to proclaim something beautiful and believe others will share our sentiment. Some of the beauty of a thing rests not in the thing itself but in the process by which it is assessed. From this perspective, a subjective response might be pleasing but it reveals no true beauty, which suggests why your average user might download a beauty app. If subjective responses were enough, anyone’s words of assurance would suffice. But there is a pleasure that beauty can give us beyond what the object in question provides directly, the pleasure in objectivity that we can take only in acts of judging themselves.
People use beauty apps not to actually quantify their appeal but to anticipate that quantification. We look forward to the moment of judging as much as to the judgment itself. It’s a less loaded, mechanized version of what you might feel before a door opens on a blind date—not the curiosity about what the other person looks like but that of seeing how the other person’s face will react to your own. The apps’ designers understand this, framing and accentuating the moment just before judgment is passed. Some apps make sounds as they scan your image: Vanity Mirror bathes your photo a halo of light while an electronic choir holds a suspended note until your results are revealed. Others generate a shuffling crescendo while calculating results, enriching the sensory joy of judgment. The app’s assessment is made to feel consequential enough that it feels even better to dismiss it with our own judgment.
From lived experience, we trust that beauty goes beyond what can be measured. But we don’t trust that fact so much that we ignore external, objective ideas of beauty that often circumscribe our lives. Women’s lives in particular have been defined by the beauty standard, whether any given woman has tried to meet it, forget it, change it, or all of the above. And that beauty standard derives not from math but from power—a fact made plain by the way it sorts us hierarchically for ends not our own, with women who are attractive earning more in the workplace (unless they’re too attractive, which carries penalties of its own), being likelier to win court cases, and finding similarly attractive partners.
With an app, we’re competing not against one another but against a phantom ideal that describes exactly no one.
Math is merely an arbitrary tool in that rating process that seems to dictate our real-life chances of “winning.” But in the world of beauty apps, it also gives the illusion of flattening out power. The Strip I imagined had winners and losers, because we were being ranked against one another, much as we are in life. With an app, potentially we can all “win,” because we’re competing not against one another but against a phantom ideal that describes exactly no one. I uploaded the face of Florence Colgate—a British woman proclaimed to have a face with perfect adherence to the golden mean—into one of these apps, and she still had a ways to go before reaching algorithmic perfection.
By their mere existence, beauty apps might look like they’re reinforcing the power behind the beauty standard. But when we use them, the power temporarily shifts into our hands, letting us seize the power of the beauty standard by rejecting what the app might tell us about ourselves. A user might still have to walk out the door and be subject to human commentary on her appearance. But when she’s playing with evaluation apps, she’s able to substitute the weak machine judgment for the human eye, leaving her able to dismiss the judgment as being inconsequential if she wishes: how is a wall prettier than a person. We can feed the fantasy that beauty standards are silly, regardless of whether they appear to momentarily favor us: Gave Miss America a 7.8 and me a 9.2. We seek objective truth about beauty not only to measure our own beauty against it but also to measure our own highly subjective concept of beauty against it as well, and even gather a sense that it might triumph. Kant again: The pleasure of judgment lives not just in the theater of judgment but in getting to preside over the adjudication.
Once the computer spits out a result, we react—immediate relief or dismay, yes, but the real reaction is our swift replacement of the computer’s vision with our own. Yeah, I’m an 8.2/no way I’m an 8.2. Let’s try it again—which, if the reviews (and my own experience) are any testimony, we do, attempting to prove or disprove or simply test the limits of computer vision by making a weird face or giving it our best smile. We turn to computer vision because it seems objective, but we treat it as though it’s inherently flawed, and then use those flaws to deepen trust in our own vision. Computers don’t tell us anything about how we define beauty; instead, they illuminate how we want to define beauty, by our own metric.
In turning to our computers’ eyes to see ourselves, we may be trying to escape from the more severe scrutiny that we subject ourselves to through our phones: that of other humans. Of course, that ignores how computer vision came to be. Algorithms are still reliant upon what we humans have collectively and historically decided is attractive, and it’s still human programmers telling computers what to look for when scanning a good-looking face.
And in fact, the most successful beauty apps facilitate human interaction instead of replace it. When you search for “how pretty am I app” on Google Play, the most popular and most highly rated app that shows up is Hot or Not—which has people, not machines, rating your photograph. Apps that connect you with makeup artists and hairstylists are far more popular than any evaluation apps: Between GlamSquad, StyleSeat, and Beauty Booked, millions have given themselves immediate access to on-demand professionals, some of whom will come to your home (“It’s like Uber, but for hair!”). When researching this piece, the only app that prompted me to drop everything and immediately download it was Spruce, which hooks you up with a live, human dermatologist who recommends a treatment plan based on your selfies and sends a prescription to your pharmacy. I had fun with the goofy computer-vision apps, sure, but they’re gone from my phone now. Spruce remains.
At best, computer vision serves as a ruse to allow you to pretend you are seeing yourself through critical human eyes, facing an “objective” judgment that you can then admit or dismiss at will. It offers a respite from the pervasive human judgment that we still depend on—the judgment that goes beyond whether someone thinks we’re pretty and instead indicates whether we belong, whether we’re liked or valued, whether we’re weird or awkward or just not quite right or any of the things that we might erroneously code as appearance. Human attention is the goal. Machine attention is a break from having to constantly strive toward that goal.
I fantasized about The Strip as a teenager. But what I looked forward to then was a trip to Shopko, the Kmart-style store on the other side of town. There, nestled in the makeup aisle, was 1989’s version of diagnostic beauty apps. It was a stubby little computer screen with a warped protective film on its face—though it was a touchscreen, giving it the air of The Future. It would ask you questions in its bitmap font: Do you have “warm” or “cool” undertones? What color are your eyes? Your hair? Are your lips full, medium, or thin? Oily skin, dry, combination? It would then tell you what makeup—all conveniently found right there at Shopko—would make you look your best. The store even left a small bin with scrap paper and pencils, library-style, next to the screen so you could jot down the computer’s picks.
I couldn’t get enough of it. I’d jump at the chance to accompany my mother on any shopping trips there, asking if I could bring my best friend so we could do it together. We’d put in our own parameters first, but after that round was over—that’s what it felt like, a round, like in a video game—we’d play with different ones, where I had blonde hair and blue eyes instead of brown and more brown, or where my best friend had olive skin and curly hair, not straight hair and skin so pale it was nearly translucent.
On no visits I can remember did I actually buy something the computer recommended. Part of this was because my babysitting money allowed me to get Wet ‘n’ Wild, not Maybelline, but it was never about the products. It wasn’t even about educating ourselves on what would work with our coloring in those neophyte years of makeup-wearing. Nothing that computer ever told me nudged me to the right on The Strip. In fact, The Strip wasn’t on my mind as I’d stand there punching in my diagnostics. Nor were the actual people who might actually be judging me: the cute skater boys I liked, the popular girls I wanted to run with. The Shopko computer was a relief because for a change, I could think about beauty in a way that wasn’t about people at all. It was about inserting myself into this new world of makeup, seeing how my qualities intersected with this technology that was promising to grant me the know-how I hadn’t yet acquired, and getting 20 minutes of fun to boot. We weren’t in it to look better. We were in it for the game.