Dataclysm: Who We Are (When We Think No One's Looking)

by Christian Rudder

  Essentially anything that, in theory, would make a group of people “less racist,” that’s what OkCupid users are. I point this out to people, who, like me, lead nice lives in large, diverse cities; who think of their opinions and tastes as nothing if not enlightened; who unwind at night with a glass of wine and a Facebook dose or two of progressive righteousness: When I show here that black women and later, black men, get short shrift, and that adding whiteness to a user’s identity makes him or her more attractive, I’m not describing some Ozark fever dream. I’m describing our world, mine and yours. If you’re reading a popular science book about Big Data and all its portents, rest assured the data in it is you.

  But look one more time at the match question above, which was written by one of OkCupid’s users and has been answered close to two million times: “vocalized” is an odd word. Get rid of it, and it still more or less reads “Would you date a racist?,” which I once assumed was the question’s real intent. The writer, however, understood the subtleties of the data set before I did. On a dating site you can act on impulses that you might otherwise keep quiet. On some level, the users come to judge and be judged by others, and each person joins the site free of the context of their everyday life. The site doesn’t connect you to your family. Nothing gets posted to your friends’ timelines. The game is: it shows you people, and you like them or you don’t; you talk to them or you don’t. There’s nothing else to it. In a digital world that’s otherwise compulsively networked, there’s an old-school solitude to online dating. Your experience is just you and the people you choose to be with; and what you do is secret. Often the very fact that you have an account—let alone what you do with it—is unknown to your friends. So people can act on attitudes and desires relatively free from social pressure.

  In the layperson’s mind, Facebook, “the social network,” is the sine qua non of online data sources. And it’s easy to see why: Facebook is huge and pervasive, and a sample of their users is pretty much a sample of people worldwide who have Internet access—in other words, you can easily get a representative corpus for whatever you want. And they have such robust and diverse data: they know who you went to high school with, what song you just listened to on Spotify, where your parents live, and so on.

  But as often as it is an asset, that richness can be a liability. You rarely meet a stranger on Facebook. The site is, by design, people you already know and whom you’ve already made up your mind about—they’re your friends, after all. Facebook’s data on race is the embodiment of the “But I have black friends” solipsism you often hear. How you treat your friends is, by definition, the exception to how you treat the rest of humanity. And you and your friends’ relationships were formed outside of the network first.

  Moreover, people become inhibited when their friends are watching. This fishbowl aspect is why the first step of most dating apps on Facebook is to get you off Facebook—your existence there is fully chaperoned. Long ago, we tried “social” features on OkCupid, and they bombed, as did similar features when Match.com gave them a go. For whatever reason, people don’t want their network along for online dating. The desire for solitude comes from the same place, I imagine, as the claustrophobia that would grip most of us if, on a promising first date at some restaurant, two old friends posted up at a nearby table. This is to take nothing away from the business or the community Facebook has created, but the “real life” relationships that both undergird and overarch the site give a different power to their data. When you want to look at something like race, where, at least among decent people, there’s pressure to behave a certain way in public, dating sites provide a uniquely powerful data set: everyone’s a stranger, alone, and there to tell you who they like and who they don’t.3

  So then let’s put OkCupid’s data up against data from other dating sites and see what shakes out. Looking at numbers made by other users, acting through other interfaces, gives us a much better sense of the real pattern. And that’s what we see below—this is data from OkCupid, DateHookup, and Match.com, sites that together signed up about 20 million Americans last year alone, presented side-by-side. In the particulars, the matrices vary—remember, these values reflect actions produced by different people using different software—but cutting through that difference is the same broad pattern. In terms of the “direction” of feeling, like or dislike, these matrices are very nearly identical:

  Match.com, you probably know. It’s been the most popular dating site in the United States for almost two decades. They buy tons of advertising on national television and, as a result, have exactly the broad “all-American” demographics you’d expect. DateHookup is a free site of several million members that is very popular among casual daters; its user base is just under 20 percent black and 13 percent Latino. It’s the most diverse of the three sites considered here. I think of it as the Atlanta or the Houston to OkCupid’s Portland and Match’s Dallas. But as you see, across all three sites, for men rating women, you get the same pattern wherever you go.

  The votes in the other direction, of women rating men, aren’t quite as uniform from site to site, though they’re still very similar:

  These matrices show two negative trends, and two positive. Blacks are again unappreciated by non-black users, but Asian men have joined them in the red. On the positive side, women clearly prefer men of their own race—they’re more “race-loyal” than men—but they also express a clear, secondary, preference for white men.

  Another way to dig into racial hierarchies is open to us on OkCupid, and it reinforces this “white preference.” Because the users are able to select more than one ethnic identity, we can study racial blends in an almost laboratory-like way. For example, we have men who check “Asian” as their ethnicity. We also have men who check both “Asian” and “white.” Comparing the two groups gives us some sense of what adding “whiteness” gets a person. It turns out: quite a bit. When you add white, ratings go up, across the board. I’ve just spilled out the complete data here. It’s a big, messy table, but it’s worth exploring.

  Down the right-hand column you see the improvement in scores created by whiteness in a person’s racial makeup. The biggest takeaway is that the racial discount applied to black men and women and Asian men in the tables above is significantly undone here. It’s the reverse of the old “one-drop” rule.

  Unfortunately, there aren’t enough people who select “black” and “Latino” or “Asian” and “black” to fully flesh out this alchemy, but it’s an intriguing glimpse at how we view the ethnic spectrum:

  Now, this is all taken from ratings on a dating website, but dating data is essentially data of the first impression, of the first blush—the users need to get to know each other, at least a little, before they’re going to want to kiss—and it’s in that same basic spirit that any pair of people come together: Well, what am I looking at? Who do I see? The data measures the frisson of meeting someone new: that burst of judgment and instinct and chemistry that determines whether you like a person or not, before you even really know much about them. Here are a few OkCupid users putting it in their own words:

  Then one day, I think I was looking through my daily matches and there he was. I instantly clicked on his profile … something about him, just made me smile.

  —Bella, on Patrick

  Well, it all began when one day I am looking through my matches and see this girl that I found attractive from first glance.

  —Dan, on Jenn

  But if there is love at first sight, there is dislike at first sight too, right? And is it not that same frisson of attraction, but in reverse, when someone flinches, however unconsciously, from a stranger? Here, again, someone in his own words:

  There are very few African American men who haven’t had the experience of walking across the street and hearing the locks click on the doors of cars. That happens to me.… There are very few African Americans who haven’t had the experience of getting on an elevator and a woman clutching her purse nervousl
y and holding her breath until she had the chance to get off. That happens often.

  —Barack Obama, July 19, 2013

  These flashes of intuition at the core of the data—extrapolations from just the smallest amount of information—pertain not just in romance, but in picking who you rent your apartment to, in deciding to approve a loan or not, and, surely, in police work, where there’s often no time for anything but a flash. Even in more deliberate situations, the first impression plays the heavy. One paper asked: “Are Emily and Greg More Employable Than Lakisha and Jamal?” and got a resounding “Yes” from our nation’s HR professionals. The scientists sent identical résumés, some with “black-sounding” names at the top and some with “white-sounding” ones, and found that the latter received 50 percent more responses, no matter the position or industry. And companies that say they’re “Equal Opportunity Employers” discriminate as much as anyone else.

  That kind of irony gets to why big studies are important, but small person-to-person measurements are essential: when you read findings like the one above, and see that Jamal doesn’t get the job, it’s easy to shake your head at the few racist hiring managers who’ve tilted the odds against him. But the data we see in this chapter shows racism isn’t a problem of outliers. It is pervasive. We’ve seen the same patterns repeated on three different sites, with different users and different experiences: men, women, free, subscription-only, casual, serious, “urban” demographics, and more “mainstream.” All told, the research set represents a large chunk of the young adults in this country, and the data uniformly shows non-blacks discount African American profiles. It’s not a problem caused by a small cluster of “ugly” black users or by a small group of unreformed racists throwing off an otherwise regular pattern.

  It is no longer socially acceptable to be openly racist. In response to that pressure, there is some portion of the public who have therefore slunk away: if I can’t shout hate at some schoolchildren anymore, well, fine, I’ll just shout it at the TV. This is not the typical American. Most of us—almost all, in fact—recognize that racism is wrong. But it is still implicit in many of the decisions we make.4 Psychologists have a name for the interior patterns of belief that help a person organize information as he encounters it: schema. And our schema is still out of step with how most of us know the world should be. By hundreds of small, everyday actions, none of them made with racist intent or feeling, we reflect a broader culture that is, in fact, racist. As we’ve seen, the pattern is so woven-in that relatively recent additions to our society, Asians and Latinos, have adopted it, too.

  When it comes to these patterns, the individuals are, in a way, blameless. That black people get three-quarters the affection on dating sites is practically an accident. I can’t fault someone for not wanting to go on a date with someone else. There’s rarely any malice in that decision. Judgments like votes are made in an instant, and are such small, seemingly meaningless, things. You browse around and maybe one face in twelve is black. And looking at that person your action at that time could go in any direction, just as it could if you were looking at a white user; you’re in the flow. And so what if you don’t like one particular person at one particular moment? It is everyone’s right to think what they want about any individual—in fact, seeing each person as an individual in the first place, and not as a category, is a huge step in the right direction. It’s just that the patterns in aggregate show that the dice, overall, are still loaded. Actually, a better metaphor from the same general category: they show that the house is still taking a rake—it’s not the dealer, it’s not the hand, it’s not even the play, it’s the rules of the game that make certain groups of people lose and others win.

  Sociology professor Osagie K. Obasogie recently produced some ingenious research—he interviewed people blind from birth and found the same attitudes about race as in the sighted world. His sample was relatively small—just 106 individuals, but he found my OkCupid data in the flesh. He cites numerous examples of a young blind person being happy on a date until some “tell”—usually the feel of the hair but occasionally a whisper from a stranger—revealed that the other person was black. The date was then over.

  Obasogie asserts that blind people’s attitudes on race reflect a lifetime of cultural absorption, as opposed to any visual reality. From his data, it seems impossible to argue otherwise. Moreover, he observed that sex is the locus of the sharpest discord between what we’re looking at and what our culture tells us we see. As he puts it to the Boston Globe, he was struck by the vigilance with which, even among his blind subjects, “racial boundaries get patrolled, primarily in the realm of dating.” To take his metaphor one step further, a patrol protects the interior, and here dating is just the frontier of a vast cultural mass that will take decades to rearrange.

  Anyhow, I’m well aware of the long and embarrassing history of “science” by white researchers conducted to “prove” the scientist’s belief that white people are better. And I’m equally well aware of how data showing that, just for example, “women find white men attractive” can come across. It is not my claim that white men are unusually good-looking. Nor am I claiming that the data “proves” black people aren’t attractive. In fact, OkCupid’s patterns change in places outside the United States. In the UK, the site’s black members get 98.9 percent of the messages white members do. In Japan, 97.8 percent. In Canada, 90 percent. Many of the black users in the former two countries, especially Japan, are Americans abroad.

  Sex sometimes has nothing to do with bone structure and muscle and flesh—the flaws and boons of which all races share in equal amounts. There is culture there too, and expectation, and conditioning. That’s what this data shows, and because it’s person-to-person, and collected in fine detail, it can show it in a way that no other research can.

  I was an exchange student in Japan for a summer in high school, and the agency officials in my host town, Utsunomiya, would occasionally collect me and the other Americans to visit a school or a factory nearby. The goal was as much for us to see the country as for it to see us. This was the early ’90s, pre-Internet, and Japan, not China, was still our big economic rival. There was tension; they had bought Rockefeller Center a few years before; the yen was threatening the dollar. The name of my exchange program captured the timbre of the visit in three words: Youth for Understanding.

  The name notwithstanding, I found the culture baffling. I remember even the characters’ names in Street Fighter II were all wrong; Vega was called Balrog and Balrog was M. Bison.… I was like, This is madness. But they did have American television; Baywatch would soon be the number one show in the country. At one school they bundled us off to, we had to get up and say a few words in front of the student assembly. I rose from the floor to the podium, said something dumb, and stepped down. The next person due up was the only blonde in our little troupe, and as she stood, and I’ll never forget it, there was an audible gasp. The person standing there was just a regular girl—we were sixteen and all lumpy and horrid—but a shudder went through the crowd as if Pamela Anderson were there in the flesh.

  Many people have taken that shudder at face value. And for decades, phrenologists, racialists, and quacks have jumped through hoops to give that essentially cultural response a biological (and therefore immutable) basis. Nell Irvin Painter’s book The History of White People gives an excellent overview of “race science,” and in the course of it she offers up a quote from an Enlightenment-era text on the wonders of the “Caucasian” race, written, naturally, by a white man:

  The blood of Georgia is the best of the East, and perhaps in the world. I have not observed a single ugly face in that country, in either sex; but I have seen angelical ones. Nature has there lavished upon the women beauties which are not to be seen elsewhere … it would be impossible to point to more charming visages, or better figures, than those of the Georgians.

  Johann Blumenbach was the writer here; he developed his racial theories by collecting and comparing huma
n skulls. Scholarship, perhaps, has progressed. The subconscious is another story.

  1 Of course, not every person on OkCupid puts themselves in one of these neat categories. However, to simplify and focus the discussion, we’ll limit our analysis to users who have selected one of the four.

  2 Black women get roughly 75 percent of the number of first messages that other women do. Their messages are replied to about 75 percent as often.

  3 Now, of course, dating sites are far from a perfect general source. As we both know, almost every user is single, and that has consequences. Using our data, if I were to sit here and research, say, spending habits, and thus conclude that the average American man spends all his disposable income on restaurants and movie tickets, I’d be making a fool of myself. A claim like this, oblivious to the special nature of my source, would be absurd.

  4 To be clear, “we” isn’t rhetorical. It means me, too.

  7.

  The Beauty Myth in Apotheosis

  I work in a universe where people identify themselves along almost every conceivable axis—as smokers and non-; as Christians and atheists; as nerds or geeks, or maybe dorks; to say nothing of black or white or Asian or gay or straight, or neither, or both. Mankind is tribes within tribes. Or, putting it more beautifully, like the Korean proverb: “Over the mountains, mountains.” That’s the ruggedness of their peninsula and the endless difficulty of our fractured human terrain.

  Running a dating site you become aware of a subdivision that on the one hand seems frivolous but on the other is as inborn as a person’s race or sexuality, and like those latter traits it’s often resistant to direct analysis. On OkCupid—as on Match, as on Tinder—a prime divide, perhaps the deepest, is between the beautiful and the rest. These are our haves and have-nots, our rich, our poor, and when it comes to sexual attention, the haves reap the benefit of their inheritance just as surely as any heir, while the have-nots largely go without. Not unlike race, beauty is a card you’re dealt, and it has huge repercussions.