The Numerati

by Stephen Baker

  This means that marketers must scope us out as individuals. One approach would be to deploy battalions of psychology and lit majors armed with clipboards to knock on our doors. That’s impractical. The sensible way to study us is to track and analyze the data we never stop spewing. And Morgan is stretching beyond that. He tells me of experiments his team is developing to monitor the spark of recognition in the brain as people look at online ads. The tests focus on a brain wave called p300. (The U.S. Navy has run similar tests to see how pilots distinguish friends from foes in the air.) If a p300 wave heats up within a fraction of a second of a subject’s seeing an ad, the Tacoda team will make the case that the viewer has not only looked at the spot but has processed it mentally. The next step? Figuring out which type of people process certain types of ads. Like other Numerati in a wide range of industries, Dave Morgan is scrutinizing humans and searching for hidden correlations. What do we do, he asks, that might predict what we’ll do next?

  WHEN I TELL people about this book, they often say, “We’re just going to be numbers!”

  Yes, I say, but we’ve long been numbers. Think of the endless rows of workers threading together electronic cables in a Mexican assembly plant or the thousands of soldiers rushing into machine-gun fire at Verdun—even the blissed-out crowd pushing through the turnstiles at a Grateful Dead concert. From management’s point of view, all of us in these scenarios might as well be nameless and faceless. We’re utterly interchangeable. Turning us into simple numbers was what happened in the industrial age. That was yesterday’s story.

  The Numerati have much more ambitious plans for us. Forget single digits. They want to calculate for each of us a huge and complex maze of numbers and equations. These are mathematical models. Scientists have been using them for decades to simulate everything from fleets of trucks to nuclear bombs. They build them from vast collections of data, with every piece representing a fact or a probability. Each model must reflect, in numbers, the physical truth: its size and weight, the characteristics of its metal and plastics, how it responds to changes in air pressure or heat. Complex models can have thousands, or even millions, of variables. And they must interact with one another mathematically just the way they do in the real world. Building them is painstaking work. And sometimes they flop. The dramatic market convulsions of 2008, for example, stemmed from faulty models that glossed over the complexity—and the risk—associated with real estate loans.

  Despite such stumbles, today’s Numerati are plowing forward, with an eye on us. They’re already stitching bits of our data into predictive models, and they’re just getting warmed up. In the coming decade, each of us will spawn, often unwittingly, models of ourselves in nearly every walk of life. We’ll be modeled as workers, patients, soldiers, lovers, shoppers, and voters. In these early days, many of the models are still primitive, making us look like stick figures. The ultimate goal, though, is to build versions of humans that are just as complex as we are—each one unique. Add all of these efforts together, and we’re witnessing (as well as experiencing) the mathematical modeling of humanity. It promises to be one of the great undertakings of the twenty-first century. It will grow in scope to include much of the physical world as mathematicians get their hands on new flows of data, from constellations of atmospheric sensors to the feeds from millions of security cameras. It’s a parallel world that’s taking shape, a laboratory for innovation and discovery composed of numbers, vectors, and algorithms. And you and I are in the middle of it.

  What will the Numerati learn about us as they turn us into dizzying combinations of numbers? First they need to find us. Say you’re a potential SUV shopper in the northern suburbs of New York, or a churchgoing, antiabortion Democrat in Albuquerque. Maybe you’re a Java programmer ready to relocate to Hyderabad, or a jazz-loving, Chianti-sipping Sagittarius looking for snuggles by the fireplace in Stockholm. Heaven help us: maybe you’re eager to strap bombs to your waist and climb onto a bus. Whatever you are—and each of us is a lot of things—companies and governments want to identify and locate you. Consider this: Google grew into a multibilliondollar sensation by helping us find the right Web page. How much more valuable will it be, in every conceivable industry, to find the right person? That information is worth fortunes, and the personal data we throw off draws countless paths straight to our door. Even if you hold back your name, it’s a cinch to find you. A Carnegie Mellon University study recently showed that simply by disclosing gender, birth date, and postal zip code, 87 percent of people in the United States could be pinpointed by name.

  The Numerati also want to alter our behavior. If we’re shopping, they want us to buy more. At the workplace, they’re out to boost our productivity. As patients, they want us healthier and cheaper. As companies such as IBM and Amazon roll out early models of us, they can predict our behavior and experiment with us. They can simulate changes in a store or an office and see how we would likely react. And they can attempt to calculate mathematically how to boost our performance. How would shoppers like you respond to a $100 rebate on top-of-the-line Nikon cameras? How much more productive would you be at the office if you had a $600 course on spreadsheets? How would your colleagues cope if the company eliminated their positions or folded them into operations in Bangalore? The Numerati will be placing our models in all kinds of scenarios. They’ll try out different medicines or advertisements on us. They’ll see how we might respond to a new exercise regimen or a job transfer to a distant division. We don’t have to participate or even know that our mathematical ghosts are laboring night and day as lab rats. We’ll receive the results of these studies—the optimum course—as helpful suggestions, prescriptions, or marching orders.

  The exploding world of data, as we’ll see, is a giant laboratory of human behavior. It’s a test bed for the social sciences, for economic behavior and psychology. Researchers at companies such as Microsoft and Yahoo are busy hiring scientists from fields as diverse as medicine and linguistics to help them grapple with the bits of our lives that are pouring in. These streams of digital data don’t recognize ancient boundaries. They’re defined by algorithms, not disciplines. They can easily cross-fertilize. This means that psychologists, economists, biologists, and computer scientists can collaborate as never before, all of them sifting for answers through countless details of our lives. Jack Einhorn, the chief scientist at a New York media start-up called Inform Technologies, predicts that the great discoveries of the twenty-first century will come from finding patterns in vast archives of data. “The next Jonas Salk will be a mathematician,” he says, “not a doctor.”

  IT’S MIDSUMMER gridlock in Manhattan. By the time I reach the French bistro in Chelsea, Dave Morgan’s already sitting at a table by an open window, reading e-mails on his Treo. He seems distracted as we eat, glancing from time to time at the handset. Just as the waitress drops the dessert menus on our table, his machine beeps. Morgan looks at it, apologizes, and hurries off into the summer heat. From my seat at the window, I watch him angling across the street and trotting up the far sidewalk.

  The next time I see Morgan, it’s October. He’s moved from Tacoda’s Seventh Avenue offices and is newly installed at the headquarters of AOL, high above the skating rink at Rockefeller Center. I meet him at the door of what he calls 75 Rock, and we walk to a café. He tells me that on the day we had lunch, he and his investors agreed to sell Tacoda to AOL. (The reported price was $275 million. The Numerati, it should be noted, tend to make a lot of money.) Morgan is working, at least for the time being, as a senior advertising exec at AOL. He certainly doesn’t need the salary. But he says he’s tempted to stick around. By tapping AOL’s resources and its millions of users, he says, he can learn even more about Web surfers and target us with ever greater precision. It’s a long process, he says. “We’re just at the beginning.”

  I ask him about the correlation he told me about earlier, the one between romantic-movie fans and Alamo Rent A Car. It takes a moment for him to recall it. “Oh yeah. They w
ere off the charts.” Did his researchers, I ask, ever come up with an explanation for it? He nods. “It had to do with weekends.” It was Alamo ads promoting “escapes” that attracted the attention of these Web surfers, he says. The romantic-movie fans booked leisure rentals, largely for weekend getaways. Perhaps they wanted to act out the kinds of scenes that drew them to the cinema. Banners for weekday rentals apparently left them cold.

  This brings Morgan to a different insight, one that involves not just who we are but how we feel. No doubt plenty of romantic-movie fans, he says, rent cars for business trips. But after reading the review of the latest candlelight-and-kisses movie, they’re thinking about getaways to Napa Valley or Nantucket. Work, at least for the moment, is far away. The challenge ahead is to map not just our tastes and preferences but our shifting moods. “If you think about it,” he says, “the movies and music that people click on tell us a lot about their state of mind at that moment. Are they happy? Are they reflective?” He considers the trove of mood messages that pour through our cell phones. That’s a new frontier and a potential gold mine of behavioral data. He goes on about the advertising possibilities of music sites, including AOL’s, where they can see us clicking on cheerful, sad, or inspirational songs.

  I’m not so sure about that. If I click on a happy song, I say, maybe I’m just looking for a pick-me-up. Morgan shrugs. He won’t know until he does more research. This means more of our data to collect and more numbers to run. Just thinking about it makes him smile. Outside the sky grows dark, and rain scatters the crowd at Rockefeller Center. As Dave Morgan heads back to his behavioral laboratory at 75 Rock, he covers his head with his hands and sprints.

  Chapter 1

  * * *

  Worker

  IT’S RUSH HOUR in New York. I stop by Hank’s stand on 47th Street, spend a buck and a quarter for a sweetened coffee, carry it to the elevator, and ride high up in a Midtown skyscraper. A big pile of Wall Street Journals used to wait at reception, one for each of us. No more. Now we’ve been instructed to read the paper online. With that, even more of our work moves onto the computer.

  I pry the lid off the coffee. I call up Yahoo, read my personal mail, and type a quick reply to an e-mail from my sister. Then I check the Philadelphia papers for baseball news. The Phillies got crushed . . . It’s 10 A.M., the coffee’s a brown stain on the bottom of the cup, and I’m just getting to the Wall Street Journal online. Or maybe I’m not.

  Office workers have had pleasant little stalling routines forever, and it hasn’t mattered much. Other laborers haven’t been so lucky. A century ago, men carrying notebooks and stopwatches made their way into factories and started to measure the movements of workers. They turned industrial production into a science, which reached its zenith in Japanese auto plants. They perfected Statistical Quality Control, and today they can analyze each spray gun, each furnace, and, by extension, each worker, minute by minute. If any one of these elements is missing a beat, they can adjust it on the spot. Many office dweebs, by comparison, luxuriate in privacy. Unless we happen to be snoring louder than usual in our cubicle when the boss strolls by, our work habits remain our own little secret. We’re scored on results, not process. Sell a house, win a trial, wow the boss with elegant lines of software code, and we’re golden.

  Things are changing, though. In the past decade, much of the work we do has moved away from the piles on our desks, the notebooks and newspapers and Post-its stuck to the door. It has migrated right onto the computer, which is now linked to a network. We’re tied to a workmate equipped with a phenomenal memory, an uncanny sense of time, and no loyalty to us. He works for the boss, who can measure our efforts with no need for a notebook or a stopwatch. The computer will rat on us, exposing each one of our online secrets without a nanosecond of hesitation or regret. At work, perhaps more than anywhere else, we are in danger of becoming data serfs—slaves to the information we produce. Every keystroke at the office can now be recorded and mathematically analyzed. We don’t own them. If our bosses wanted to, they could order up an e-mail chart for each of us. It would display the words we write most often, in proportionally sized fonts. You could only pray that movies or beer wouldn’t show up bigger on your chart than the medicines you sell or the stocks you recommend. That online version of the Wall Street Journal? Our employers can follow which articles we read. They can also buy software to create maps of the people we communicate with—our social networks. From these, they can draw powerful conclusions about our productivity, our happiness at work, and our relations with colleagues. Just what kind of team player are you, anyway? Microsoft even filed in 2006 to patent a technology to monitor the heart rate, blood pressure, galvanic skin response, and facial expressions of office workers. The idea, according to the application, is that managers would receive alerts if workers were experiencing heightened frustration or stress. Such systems are in the early stages of research. But even with today’s technology, if your company is not scouring the patterns of your behavior at the keyboard, it’s only because it doesn’t choose to—or hasn’t gotten around to it yet.

  Why would companies intrude like this? Very simply, to boost our productivity. For centuries they’ve concentrated on results because, like the newspaper advertisers now rushing to Dave Morgan’s offices at Tacoda, they haven’t had the means to monitor and dissect what we actually do. Now the tools are at hand. Don’t they have a responsibility to shareholders to put them to use and pump up productivity and profits? That’s the way they see it.

  Now as I look at the workplace through their purposeful eyes, I’m already feeling a trace of nostalgia for the idle moments and wasteful routines that brighten my days. Sitting in my 43rd-floor office, I call up YouTube and click on a silly Morphing Pug video. An animated dog dances and sings a ridiculous song. I wonder what that investment of 45 seconds of utter nonsense could possibly say to my bosses about me. Is there a correlation between Morphing Pug watchers and prizewinning journalism? It’s doubtful. And it’s a matter of time before management starts recording such behavior. The very thought fills me with such regret that I click on the video once more, not so much to laugh at the dog as to soak up the on-the-job freedom it represents.

  On a late spring morning I drive over the Tappan Zee Bridge, across the wide expanse of the Hudson. Then I hook left, away from New York City and up into the forests of Westchester County, to the headquarters of IBM’s Thomas J. Watson Research Laboratory. It sits like a fortress atop a hill, a long, curved wall of glass reflecting the cotton-ball clouds floating above. I have a date there with Samer Takriti, the Syrian-born mathematician who launched me on this entire project. He was the one who described to me early on how his team was building mathematical models of thousands of IBM’s tech consultants. The idea, he said, was to piece together inventories of all of their skills and then to calculate, mathematically, how best to deploy them. I came away from that meeting convinced that if Takriti could model people as workers, then eventually we’d also be modeled as shoppers and patients—in short, in a whole range of our activities as humans. Now I’m going back to find out how Takriti and his team plan to turn IBM’s workers into numbers—and what they’ll do with them (and with us) if they succeed.

  Takriti, a slim 40-year-old with wide, languid eyes, opens the door of his small office. He wears a rugby shirt tucked tightly into blue jeans. He’s on a conference call but waves me in. On one wall of his windowless office is a whiteboard covered with math calculations that mean nothing to me. Takriti is quiet on the call, just saying, “A hum, a hum.” I look to the other wall, which is decorated with an electricity grid of New York and Pennsylvania. This is an artifact from Takriti’s previous life, when he used math to model chunks of the old economy, things like steel mills and power plants. Story has it, Takriti says after he hangs up, that the original Takritis were warriors who marched from Saddam’s native city, Tikrit, in Iraq. His branch of the family, he tells me, eventually settled in Syria. A top engineering student in Da
mascus, Takriti won a fellowship in the mid-1980s to study at the University of Michigan. He fell head over heels for math. In 1996, by then a Ph.D., he landed a research job at IBM’s fabled Watson Research Center, a half-hour drive north of New York City. This son of Tikrit warriors now walked among the gods of math.

  Takriti’s specialty was stochastic analysis. This is the math that attempts to tie predictions to random events. Say it rains in Tucson from zero to six times per month, and you listen to the weather report, which has been right 19 of the past 20 days, only three times a week. One of your three jackets is suede. What are the chances it’ll get drenched tomorrow? Imagine that same question with one thousand variables, and you’ve stepped into the stochastic world.

  A generation ago, a crew of math whizzes led by Myron Scholes and Fischer Black focused their mastery of probability on finance, where they calculated risk and put prices on it. This led to a panoply of new financial products, from options to hedging strategies. It was a math revolution on Wall Street. The mathematicians were replacing hunches, wholesale, with science. Takriti says that by the time he reached IBM, many of the same math tools were being refitted for other industries.