The Numerati

by Stephen Baker

  Chapter 7

  * * *

  Lover

  MY WIFE SHOUTS from down in the computer room. She says the online dating site is asking her to describe the man she’s hoping to land. “What do I write?” she asks.

  “Just describe me,” I yell. I’m at the dining room table, filling out the same dating form on my laptop.

  I hear her muttering below. She’s not happy that I’m dragging her into an experiment that promises to reveal the algorithms of love. I want to see whether an online dating service I’m looking into, Chemistry.com, will sift through the answers and essays we provide, run them through their analysis, and offer each of us to the other as a good match. Later I’ll be talking to the Rutgers University anthropologist who devised Chemistry’s matchmaking formula, Helen Fisher. She was the one who suggested, in a phone call, that my wife and I try this test. I agreed and strong-armed my wife into joining me. We vowed, naturally, not to respond to any of our fellow daters on the service. We were testing the algorithms, not the people. Granted, there’s nothing scientific about studying one couple. For all I know, our marriage is a near-miraculous triumph against the odds.

  I embark on this research, I’ll admit, with plenty of doubts. An individual’s emotions are hard enough to decipher, let alone predict. We’ve seen in politics how tough it is to figure out what draws people to one party or another. How much more complicated will it be to match two people, each one as complex as a universe? The Numerati’s approach works best in areas where sets of consistent data reflect faithfully what they’re looking for. Our spending and earning patterns tell them about our risk as debtors. That’s easy. Supermarket managers given a list of papaya lovers can bet that a healthy share of them will respond to a discount on mangoes. Piece of cake. But what data best describes us as mates? How can you model someone as a lover?

  I remember having dinner a few years back with friends in Pittsburgh. The husband was saying that his wife’s brother really blew it when he broke up with his girlfriend. “She was un-be-liev-able,” he recalled, shaking his head. He piled on the adjectives. Incredible, amazing. Clearly, there was something about this woman, some piece of data that he viewed as highly desirable, perhaps essential. Would he dare say it in front of his wife, who was sitting right next to him? I finally asked him.

  “She had a great . . .” He paused to add emphasis, and I was thinking of all the possibilities. “Personality” would work. If he said “body,” or worse, “ass,” our friendly dinner was in trouble.

  “Job,” he said. She was a highly trained nurse who could make good money in any town in the whole country. His brother-in-law, if he had known what was good for him, could have enjoyed guaranteed income wherever he went. (I should note, just to place the comment in context, that this friend’s wife had lost her job about a month earlier.)

  In theory, finding an online partner for my friend would be a cinch. Just rank the dating prospects by income or credit score, almost as if they were applying for a loan. Then he could start from the top. But I’m betting this nurse he admired had certain other qualities he valued, perhaps ones he wouldn’t mention in front of his wife. Maybe he barely recognized them himself.

  So how do scientists break down love into pieces that can be fitted into a statistical hierarchy? Love has long resisted such measurement, which is why it has stuck stubbornly to the realm of poets—and frustrated scientists at every step. Shakespeare, no doubt, understood it far better than Newton, and was one of the greatest experts of his day. The challenge today is to find modern experts, whether from anthropology, like Helen Fisher, or psychology, and to team them up with today’s Newtons: the Numerati. What they have to do together—and this is the tricky part—is distill what they know about human love and relationships and fit it into a series of algorithms. Some think this is folly. But these online dating sites provide a bonanza of relationship data, more than Shakespeare or even Dr. Kinsey could have imagined. The Numerati are laboring in laboratories of love.

  To compare the old, intuitive style to the scientific approach on the website, I send an e-mail to my friend who many years ago, in El Paso, Texas, set me up with my future wife. Which bits of data, I ask him, led him to make this recommendation? I get back a response within minutes. The resulting data, if you could call it that, isn’t anything the most gifted of the Numerati could measure or model. He writes that he could communicate well with both of us, that we had a “similar sense of humor,” and he felt “an enormous amount of positive energy.” Pure fog, from a data-mining perspective. Small wonder, then, that Helen Fisher and her colleagues call their service Chemistry.com. They’re trying to uncover certain data at our essence—perhaps the qualities that Shakespeare compared to “a summer day” or “a worm ‘n the bud”—and package it in matchmaking algorithms.

  I slog through scores of questions on the Chemistry site, while working to quell a mounting insurrection from downstairs. Clearly, many are designed to gauge whether I’m outgoing, adventuresome, cautious, a stickler for details. Some, though, are harder to read. I’m shown a photo of a man and a woman having a drink on a terrace. Brother and sister? Lovers? Husband and wife? I guess lovers, but it’s hard to know what that answer says about me. Then I’m asked to compare the length of my second and fourth fingers. That too is a mystery. The form asks about relationships that have thrived. I describe my wife. Finally, I get to the essay where I’m asked to define myself. This scientific system, I’m assuming, will analyze the word selection and the length and syntax of my sentences, and then use them to pigeonhole me as one type of lover or another. My choice of words will even reveal my inner secrets, I’ve been told. But for this to work, I figure the system has to see the free-flowing, unedited me. So I relax and, for their computer’s benefit, type copiously about myself. I write about a year I spent long ago in South America as a near hermit. I write about this book, and about going to coffee shops and putting on noise-canceling headphones and writing. I go on and on with this stream of consciousness. Then I hit the send button. A few minutes later, I’m appalled to see this blather appear verbatim on the profile, right next to the blank space where my photo would appear. This essay was no quiet tête-à-tête with a computer! It was ad copy for myself. What does that have to do with data-crunching Numerati?

  They’re only collaborators in this effort. Much of it, at least at this stage, involves self-promotion. It’s the same process we go through when we apply for the class of 2014 at Harvard or try to land a six-figure consulting job at McKinsey and Co. We put forth the most flattering image—much as we have throughout history. And we look for the Numerati to find us the right match. But what if we’re both lying?

  That’s the risk we take. Now that we’re advertising ourselves, many of us may choose to litter our profiles with little white lies and deal with the consequences later. What counts, as any advertiser will attest, are results. Look at my possibilities. I have a 1999 picture of myself that might give a good bounce to my profile on Chemistry.com. If I Photoshopped away that small fin-de-siècle zit on my nose, lightened the bags under my eyes, and added $50,000 to my reported income, who knows what the youthful, wealthier (and not entirely honest) me could accomplish on this site?

  Is this science? That depends. Consider the data I provide to Chemistry.com. I start with demographic details, and I hand them over knowingly. I have a pretty good idea how they’ll be interpreted. Some may flatter me. Others I part with in the resigned spirit of full disclosure, or a believable facsimile. So my prospective dates learn my age, address, profession, education, religion, even my income if I choose to provide it (which I don’t). The questionnaire asks if I have children living at home. (It politely refrains from asking if my children and I are living there with my wife.) It then asks about the women I’m willing to consider. Do I prefer to exclude certain religions or body types? (Exclude no one, I say.) Am I interested in teetotalers, high school dropouts, or women taller than me? (Yes, I want them all.) These ty
pes of questions date from the infancy of computer dating, in the 1960s. They don’t demand sophisticated analysis or modeling. They only ask the computer to carry out simple matching chores, to put us into piles. These are the details marketers and politicians have feasted on for decades. Many still swear by them. And these preferences are accompanied by what many consider to be the most important piece of data in the entire process: the photograph. (I withhold mine. This may raise suspicions among my dating prospects, but I’m interested in how the Numerati interpret my data. They’re not analyzing the photos, though perhaps they will in years to come.)

  Next comes a stream of involuntary data that was unavailable to the punch-card pioneers of computer dating. This is our behavior on the Chemistry.com site. Analysts at the company, like their colleagues throughout the e-commerce universe, record our every click. They can measure which types of potential dates appear to interest us the most. Then they can showcase more of the same genre to us (and people like us). They can break us down, category by category, and look for trends. This analysis is nearly identical to that done by Tacoda’s Dave Morgan and other online advertisers. They don’t pretend to know us in any depth. Our thumping hearts and quivering gills remain a mystery to them. They simply count our clicks, study our behavior, and then put us in buckets and market to us.

  Finally, we venture into the newest realm of data: the survey responses, which, when interpreted by scientists like Helen Fisher, create our love profile. This is where most of us lose control of the process. It’s hard to figure out what kind of self-portrait we’re creating because many of the questions seem mysterious. But who wants to fine-tune at this juncture, anyway? If we’re paying money to find the ideal date, most of us (at least in theory) want the service to understand us as well as it possibly can. So we scrutinize the length of our fingers, puzzle over that picture of the smiling couple on the terrace, and fudge only in cases where the question appears to probe something we’d rather not admit to. (If the questionnaire lays traps for child molesters or pornographers, I haven’t spotted them.) The idea in this psychological section is to advance far beyond demographics and behavior, to burrow deep within us, and untangle our heartstrings. This is to understand us at our most basic level—as beings engaged in mating rituals we share with other animals, from shad to kangaroos. I have no doubt that this test will help Helen Fisher and her team understand at least some of the cravings and neuroses operating inside me. But do these profiles really line us up with the right person? Or do they just give us something to talk about on our first date? That’s what I’m out to learn.

  So I ask Helen Fisher when I catch up with her by phone. Fisher is both an anthropologist and a neuroscientist. Matchmaking, she says, “is the most important game we play. From a Darwinian perspective, if you have four children and I have none, your genes win.” And she thinks that the standard data used in matchmaking sites, the shared hobbies and interests, are nearly worthless for finding a spouse. “You can have the same ethnic background, the same socioeconomic standing, the same general level of intelligence,” she says. “Your looks, religion, politics, and goals can all be aligned. You can walk into a room full of people like that, and you don’t fall in love with them. I can’t tell you,” she adds, “how many relationships I walked out of where the person was perfect on paper.” She’s confident that her method will decode the human lover just the way other efforts I’ve been telling her about model us as shoppers, voters, and workers. “We’re going to get to the bottom of this, just like IBM and Yahoo,” she says. “The human animal does have patterns.”

  Fisher says that in the late 1990s she began looking into the biology of personality, the genes, neurotransmitters, and specifically, the hormones. She did this in part by studying brain scans of “romantically obsessed” people. Her theory is that four different hormones—estrogen, testosterone, dopamine, and serotonin—mold our personalities and that we look for people who complement us, who provide what we’re missing. Her questionnaire is designed to divide us into four different types, each one with a dominant hormone. Some of the questions focus on the moods and personalities she associates with each hormone. Others, such as the question about the length of our fingers, zero in on the chemical itself. Research shows, she says, that those with an index finger shorter than the ring finger have often been exposed to more testosterone while in the womb, while those with longer index fingers will have more estrogen.

  Fisher outlines the different hormones and personalities for me. Those with lots of dopamine, she says, are likely to be “Explorers,” optimistic risk takers. Serotonin breeds “Builders,” who tend to be calm and organized and work well in groups. Those brimming with testosterone she calls “Directors.” Two thirds of them are men. They’re analytical, logical, and often musical. (They sound suspiciously like Numerati to me.) In the fourth group, their brains coursing with estrogen, are the Negotiators. They’re verbal and intuitive, and have good people skills. You’d think they’d be built for relationships. But sometimes, Fisher says, “they’re so pliable that they turn into placaters. You don’t know who they are.”

  People leave personality footprints everywhere, Fisher tells me, even in the sentences they write. She gives me common words used by each group. Explorers use words like excite, spirit, dream, fire, and search, while more community-minded Negotiators talk about links, bonds, love, team, and participate. Builders are more liable to discuss law, honor, limits, and honesty. And that Numerati-infested bucket of Directors? Their words focus largely on the physical world, where aim, measure, strong, hard, and slash have currency. Not surprisingly, they also talk a lot about “thinking.”

  My wife and I, we learn later in the afternoon, are both Explorer-Negotiators. (Each person gets a dominant and a secondary label.) This sounds promising enough. “You tend to be focused and resourceful, and you are able to juggle a lot of projects at the same time,” I read. As a result, we’re both “sometimes a whirlwind of activity.” But a pairing of Explorers, Fisher warns me, can be risky. “Explorers fly off in different directions the minute they get bored,” she says. “They get into relationships fast, wonder how they got there, and then try to weasel their way out.”

  Okay. Maybe each of us really needs a no-nonsense Builder to keep our finances in order, map our vacations, and make sure the cats have their latest battery of rabies shots. Perhaps that would make sense. But is that what our hearts secretly ache for? As Fisher starts out, her evidence is mostly anecdotal. She describes a classic match. Picture a hard-driving man, a fabulously successful business executive. He bangs heads, slashes the payroll, drives would-be challengers into oblivion. This guy gets things done. He’s a Director. And chances are, Fisher says, he has a smooth-talking, problemsolving wife, who quietly patches together all the friendships he shatters. She’s a Negotiator. Those two types, Fisher says, “are very symbiotic. They will gravitate toward each other.”

  Clearly, the service doesn’t sense the same gravitational pull between my wife and me. When I log on to the website, I find a list of five women who have the right levels of serotonin and estrogen for people like me. My wife isn’t one of them. There’s an insurance manager from West Orange—a Negotiator-Explorer—who says “we all have to laugh every day, especially at ourselves.” A Negotiator-Builder, from Rochelle Park, works in information security and likes ballroom dancing. These and three others are the machine’s choices. Many other subscribers, however, have access to my profile. And regardless of the chemistry, they’re free to express interest. Whether they’re Builder-Directors from Tarrytown or fellow Explorer-Negotiators from Toms River, I learn that each one is a “great match.” It gets to the point, as I click through my prospective matches, where the word great starts to sound pretty ordinary.

  What gives? The automatic system, in all modesty, recognizes its limits and bows to the human brain. As the science stands now, it can make introductory suggestions. But far be it for a machine, at least at this point, to overrule
the vastly more sophisticated human and nix a potential Romeo. “Great match,” it says.

  Even if it doesn’t dare second-guess our judgments, the Chemistry computer should be able to suggest smarter pairings as it sees which combinations work. Fisher tells me she has data from 1.6 million people who took the test. She can see which types of people are most likely to pursue which others. Statistics indicate, as she predicted, that Negotiators gravitate toward Directors, and vice versa. Explorers are attracted to Negotiators. No-nonsense Builders are often drawn to Explorers, who help them “lighten up,” Fisher says. But just as often, Builders opt for a less combustible combination and seek out their own kind. With these insights, she can refine the recommendations—and perhaps lead the system to help me find my wife.

  Of course, the personality groupings are only one smidgen of data in our dating profiles. I talk to the analysts at Match.com, the parent company of Chemistry.com. Each of us, they say, instinctively seeks out a match from our own cultural and educational level. We can find this compatibility simply by reading the other person’s essay. As our level of education rises, we use larger words and longer sentences. Daters naturally tend to select people at their own level. Dating services can accelerate this process by presenting us first to people with similar vocabularies. We also focus on other similarities, no matter how silly. The Match.com team has found that if they can show people that a potential match has three things in common, interest soars. “We can send them e-mail saying that there’s someone else in their town who also likes dogs and whose favorite color is red,” says Jim Talbott, manager of web analytics. He speculates that people look at these as “a little piece of coincidence, a little piece of fate.” Regardless of the cause, it’s easy to hunt down these similarities, use them as marketing tools, and then measure which combinations appeal most to each type of customer. The dynamic, once again, is nearly identical to targeting advertising. This proves to be a much comfier niche for the Numerati than the tangles of human attraction.