You May Also Like

by Tom Vanderbilt

  This was the crux of what Liu was working on at Hunch, a site that had been formed, as its founder had described, as a “way to give you recommendations of any kind.” Hunch*1 invited the user to answer a series of simple, sometimes playful, seemingly unconnected questions: Have you ever purchased anything from an infomercial? Which of these greens would you usually prefer in a salad (images of iceberg, romaine, red leaf, and arugula)? Do you like it when the cabin crew cracks jokes on an airplane?

  Initially, Hunch was meant to be a personalized “decision engine,” a way to answer all kinds of questions (for example, where should I go to college?). But the line from how you sliced your sandwich (diagonally or across the middle) to “what Blu-ray should I buy?” could be tenuous. And how could you ever truly validate whether Hunch’s ultimate recommendation was correct? It also turned out that “people loved talking about themselves” via the quirky personal questions. So Hunch “took this taste component,” Liu told me, and made it the entire site. The idea was a kind of meta recommendation engine.

  Answering the Hunch questions fell, to my mind, somewhere between taking a magazine psychology quiz and playing the old artificial intelligence program ELIZA. You vaguely sensed you were being manipulated, but with a compulsive fascination you pressed on. Most questions were not indicative of anything in and of themselves; Hunch had no psychological theory about what kinds of people liked jocular cabin crews. Rather, Liu said, the questions were meant to be, above all, engaging. The average person, he noted, gave more than 110 responses. The questions were also designed to be occasionally jarring. “People have come up with baked answers for many things,” he said. Maybe they could avoid the typical biases. “If I ask, ‘Are you a good person?’ it’s just like asking someone if they’re middle class. Everyone in America is going to say they’re middle class!”

  But what if you are asked, “Will you go out of your way to step on a crunchy leaf?” It is probably not something you have thought much about. Would its answer betray any greater understanding beyond itself? Rather than asking, “Are you a good person?” Liu suggests, why not ask, “Would you drink from a public drinking fountain?” Does your inclination to answer yes to this question happen to correlate with your answer to the question of whether you would ever risk your life to save someone?

  The idea with Hunch was that if you asked enough of these questions—the slight ones and the seemingly meaningful ones and everything in between—and then correlated all the answers into a massive “taste graph,” a mathematical depiction showing where people and their collective preferences were in relation to each other, you could get a robust two-dimensional understanding of human behavior. You could get the “who” and the “what” and leave the “why”—why the leaf crunchers preferred Toyotas—for the psychologists.

  The correlations were striking. The magazine Wired described a few: “People who swat flies have a thing for USA Today. People who believe in alien abductions are more likely than nonbelievers to drink Pepsi. People who eat fresh fruit every day are more likely to desire Canon’s pricey EOS 7D camera. And respondents who cut their sandwiches diagonally rather than vertically are more likely to prefer men’s Ray-Ban sunglasses.” Whether any of these made sense, or were actionable on their own, was almost beside the point; by simply fathoming the vast web of associations, Hunch could know you. “A quietly radical promise” is what Wired called it, “implying that our tastes are defined not only by what we buy or what we’ve liked in the past but who we are as people.”

  —

  Except it was not so radical. It was not even particularly new. The sociologist Georg Simmel, writing in 1904, noted that fashion “signifies union with those in the same class,” while demarcating the “exclusion of all other groups.” It is little surprise that Simmel was writing in the Victorian period, obsessed as it was with social distinctions. Philosophers had begun wrestling with aesthetic taste in earnest in the salons of the eighteenth century, spurred by a rising bourgeois class in which, as the historian Jennifer Tsien suggested, “everyone felt that they had the right to make judgments about paintings and books.”

  In the nineteenth century, taste went from philosopher’s rumination to social obsession. As more people had more money, signifying who you were, socially, became a kind of game. Social and cultural identity was increasingly defined less by long-established institutions (the church, the aristocracy) and more by money—how much you had and, more important, how you spent it.*2 What you wore helped define who you were. And the more open to interpretation who you were was, the more important what you wore became.

  “The more nervous the age,” Simmel wrote, “the more its fashions change.” Consider the Victorian “extreme makeover.” When one upper-middle-class client called on the famed London decorative arts firm Morris & Co., asking Dante Gabriel Rossetti what should be done with his home, Rossetti’s answer was as swift and emphatic as any latter-day reality television host’s: Begin by “burning everything you have got.” The client later praised the firm for saving innumerable people from “sitting on shepherdesses, or birds and butterflies, from vulgar ornaments and other atrocities in taste.”

  In a novel like Elizabeth Gaskell’s North and South (1855), the geographic divide is really a taste divide, polite London society versus the emergent merchant classes of the North. Nothing is too minor to broach the fault lines of taste—from the pattern of the wallpaper to the appropriate “number of delicacies” on the dinner table. The very phrases “good taste” and “bad taste” actually did not surface in earnest until the twentieth century (according to Google’s book database Ngram). They seemed to plateau in the 1950s, when the “middlebrow” reached its ascendance (as one wag described the concept, “People who are hoping that some day they will get used to the stuff they ought to like”).

  But no one quite so thoroughly plumbed the taxonomy of taste—what it was, what it was for—as the French sociologist Pierre Bourdieu. Distinction, his landmark 1979 book, was dubbed a “Copernican revolution in the study of taste.” Bourdieu created “taste profiles”—as today’s Internet sites would call them—of some 1,217 French subjects. He combined ethnographic observation with an exhaustive and innovative survey, which asked scores of questions: “Which are your three favorites among the painters listed below?” “Where did you get your furniture?” He even wanted to know how people did their hair.

  He tallied all that against people’s demographics, rigorously and rigidly sorted into groups like “Executives, engineers” and “Clerical, junior executives” (he warned it was a very “French book”). He found, through statistical correlation, that “social subjects, classified by their classifications, distinguish themselves by the distinctions they make.” This itself was not novel, but Bourdieu emphasized just how minute these taste distinctions could run, how firmly tied to one’s place in society they seemed to be, and how often they were determined less by one’s wealth than by one’s education.

  The correlations were strong: In music, the “dominant classes” preferred works like Ravel’s Concerto for the Left Hand; the “middlebrows” liked Hungarian Rhapsody (so often used in mid-century cartoons); while the “popular” classes went for “lighter” fare like The Blue Danube. One’s “cultural” capital was a stronger predictor of taste than one’s actual capital. More than money, cultural capital bracketed people: Parisian architects liked Kandinsky; dentists preferred Renoir.

  You expressed your taste not simply through the films you saw. The way you talked about them also served as a none-too-covert display—a “rank to be upheld or a distance to be kept”—of your cultural capital. Did you go see the latest “George Clooney movie,” or did you go see the latest “Alexander Payne film”? Talking about directors becomes a signal that you belong to a certain place in the social hierarchy. It is a subtle badge allowing admission into a kind of club, like knowing that the American designer Ray Eames was a woman and that Ortega y Gasset were not two different people but one Sp
anish philosopher (two of my early mistakes).

  Bourdieu insisted these “oppositions” were found not only in “cultural practices” but in more mundane things, like “eating habits.” He wanted to tear down the old Kantian divide between “aesthetic consumption”—the art we liked—and “the world of ordinary consumption”: the baser pleasures of what we ate and bought. He saw taste at work everywhere. “Taste is the basis of all that one has—people and things—and all that one is for others,” wrote Bourdieu. “The science of taste has to abolish the sacred frontier which makes legitimate culture a separate universe, in order to discover the intelligible relations which unite apparently incommensurable ‘choices,’ such as preferences in music and food, painting and sport, literature and hairstyle.”

  —

  Hunch, in its own way, was trying to discover those “intelligible relations.” Hunch was like Bourdieu on steroids, with none of the sociological weight but with a vastly wider data set (some fifty-five million responses), spanning a ridiculous number of behaviors. It was no longer just Bourdieu’s paintings and food but what kind of Christmas tree you preferred (real or fake), what sort of french fries you liked, and how you came down on the question, “Is it wrong to keep dolphins in captivity and teach them to do tricks?” Every time you answered a Hunch question, as Liu told me, “you add[ed] some kind of clarification to your coordinate on the taste graph.” In the way that GPS uses triangulated latitude and longitude coordinates to map your location on earth, Hunch had a fifty-coordinate system to place you in society.

  If Hunch.​com had a pop Bourdieu feel, it is little surprise to learn that Liu, in his MIT days, was inspired by the Frenchman. But much of that research had come from 1960s France. Subsequent scholarship had cast doubt on Bourdieu’s rigid hierarchies of class and taste; broadly speaking, many contended that taste was no longer an upper-class strategy to vertically dominate the lower classes but a horizontally dispersed system of coexisting communities of interest, of “taste worlds.”

  Traditional taste signifiers had gotten a bit slippery and, in theory, more democratized. At a place like Hunch, as with other Internet startups, where everyone looked the same age, no one had an office, and it appeared that everyone wore jeans and a T-shirt, there was no immediate way to discern social hierarchy. It seemed to reflect a new rule in America: As income inequality rises, people dress more alike. When an executive such as Google’s Sergey Brin wears sandals, is this counter-signaling, dressing down to disguise his wealth? Or is it just more signaling? My seeming disregard for my look is actually a powerful connotation of my power. When everyone sets the same smartphone on the table, where does one discern socioeconomic difference? In the gigabytes, which are not visible? Is it the hand-tooled ostrich-skin case, with one’s initials burnished in a provocative typeface, that sets one apart? Or is it that your phone has no case at all, an indicator that you care not for its fate because you will have the next version on its release date anyway? The ultimate social demarcation may be to have no phone.

  Signaling, Liu argued, had blurred. The expensive-looking shirt was a bargain at H&M. Much of the world Bourdieu charted in Distinction had moved online. One’s habitus could be expressed in the casual Instagram post of the vintage modernist chair passed down from one’s grandparents or the richness of the crema (a word no one knew a few years ago) in one’s single-origin espresso.

  The anxious positioning Bourdieu had noted could be felt in a tweeted “humblebrag,” an attempt to claim cultural capital without looking as if one were doing so. Thus the up-and-coming band tweets, “Our song has just come on the radio in our taxi. Awkward!” People’s musical likes, among others, could be displayed in their Facebook profiles. And not idly: One university study of Facebook accounts found that only people who put “classical” and “jazz,” and not “indie” or “dance,” in their “likes” encouraged others to follow suit. Only the former categories had an aura of prestige.

  Teasing explanation from all the Hunch data could breed questionable correlations and fanciful theories. Liu suggested that someone’s propensity to walk out of a movie he did not like could be a psychological surrogate for being more predisposed toward divorce. “A bad marriage is like a bad movie,” he told me. “Do you stick around?” At moments like this, it seemed hard to take Hunch as little more than a data-driven gimmick. But then, back in the Decision Lounge (that is, the only enclosed space) at Hunch’s offices, Liu ran me through the site’s “Twitter Predictor.” Hunch took my Twitter followers, and the people I followed, mapped all their taste coordinates, and then generated one for me. “This is taste by association,” he said.

  The Twitter Predictor then asked me questions and guessed how I would answer them. “Given the name of a well-known foreign country, would you know whether their time zone is ahead or behind yours?” Yes. “Did you vote in your country’s last major election?” Yes. “Do you watch documentaries?” Yes. So far, the Twitter Predictor had me figured out quite well. I felt as if I were on OkCupid and had found myself.*3 But were they obvious questions? Or was I simply falling for the so-called Forer effect, that tendency, in places like psychological tests or fortuneteller readings, to see ourselves sharply revealed in what are actually very broad assertions?

  As the questions kept coming, they seemed to get more specific and less naturally aligned by factors like politics: “Do you think giving clean needles to addicts is a good idea?” “Do you play games on Facebook?” “Should doctors be able to assist a patient with suicide?” But it did not falter. Liu checked my score. “Hunch is up 19–0.” He told me they have achieved roughly 90 percent accuracy on predicting answers. As Hunch’s founder Chris Dixon put it, “People in our studies are actually only consistent with themselves about 90 percent of the time.”

  —

  It was a curious and powerful moment. In an age of individualism, many of us have convinced ourselves that we are complex creatures marching to our own drummers, unable to be pinned down into safe assumptions. “My own taste reflects my specialness,” summarizes the music critic Carl Wilson, where “it’s always other people following crowds.” But here I was, in the Decision Lounge, seemingly pinned like a butterfly to a fifty-coordinate wall, my preferences clearly outlined in a connect-the-dots pattern. “What’s so fascinating is that we aren’t capturing your answers to these questions directly,” Liu said. “We’re capturing you as a location in taste space.”

  Actually, it is even one step removed: Because I had not previously answered any of Hunch’s questions, I was being captured simply by the aggregation of all the answers to these questions given by all the people I am following on Twitter. “Taste is a space on a graph,” Liu said. “Someone can inhabit it without necessarily knowing the specifics of what they believe and their experiences.” This underscores the social homophily—that tendency to cluster—discussed in the last chapter: I was not motivated to answer any of these questions a certain way because I was influenced by someone’s individual tweet (though “a lot of users,” notes Liu, “suspected we were reading tweets” to make the Twitter Predictor work). Rather, I was associating with a lot of people on Twitter who were like me to begin with: Birds of a feather tweet together.

  Because people are often puzzled by other people’s tastes, it is easy to accept the maxim “There’s no accounting for taste.” “People just assume that tastes are inexplicable,” Liu told me. They will say, “I am unique, just like everyone else.” “Of course there’s accounting for taste,” he added. “You have to look for the right features.”

  For Bourdieu, one thing stood, above all else, as a shortcut to cracking someone’s taste. “Nothing more clearly affirms one’s ‘class,’ nothing more infallibly classifies,” he wrote, “than tastes in music.”

  MUSIC IN THE KEY OF YOU

  What sort of music do you like?

  Is there a question that at once seems so reductionist yet so open-ended, so banal yet so freighted with meaning?

 
But it comes up: In studies of “zero acquaintance,” where people were meant to try to get to know one another, music was the first topic broached (granted, they were college students). It is not just small talk: People’s music preferences are potent in drawing accurate inferences about their personality, or at least the personality they are trying to project.

  Likes seem easier to discuss than dislikes. Likes are public, Hugo Liu had told me. A person’s clothes reveal his likes, but not necessarily his dislikes. Dislikes—even though they are so crucial to taste—tend to be private. Sites such as Facebook do not even offer a “dislike” option.*4 Talking about likes might be a good way to find out if someone could be a possible friend. But discussing dislikes is generally reserved for those already in your social network; Liu compared dislikes to gossip you exchange with friends, a way to groom relationships. Simply expressing your musical preferences depends on any number of factors: who’s asking, what you’ve listened to lately, where you are, what you can remember.

  These kinds of questions animate the Echo Nest, a “music intelligence” company in Cambridge, Massachusetts, that is a kind of mash-up between the neighbors MIT and the Berklee College of Music, data geeks playing with music geeks. The essential job of the Echo Nest, owned by Spotify, is to help solve the dilemma of matching people to music in an age when the latter is in virtually inexhaustible supply.

  When I arrived at its offices one afternoon, it probably should not have come as a surprise that the very first interaction I had was about musical taste. As I sat down with Glenn McDonald, the company’s principal engineer, I asked what was playing on the stereo. In an office where everyone must be bristling with opinions, how could they decide what to play? “The rule is ‘anything but Coldplay,’ ” he said sardonically. There it was, that line in the sand, delivered half in jest but still able, in one cutting thrust, to divide the population into those who liked Coldplay, those who did not, and those who did not feel strongly either way—but could still perhaps get the joke. Coldplay may be a particularly good litmus test for taste. Type in “Coldplay is,” and Google autocompletes, in this order, “Coldplay is the best band ever” and “Coldplay is the worst band ever.” Much of the venom for Coldplay is no doubt driven by that very adoration. Whatever the reason, people are taking sides. Take enough of these sides, and you begin to locate “your music”—and yourself—on the taste graph.