Simulating Democracy

“Which category have they put you in?”

This sinister question—at least, it was meant to sound sinister—headlined the advertising copy for The 480, a 1964 novel by Eugene Burdick. His previous best sellers, The Ugly American and Fail-Safe, had caused sensations in political circles, and the new one promised to do the same. Its jacket featured the image of a punched card. The title referred to 480 categories of voter, defined by region, religion, age, and other demographic characteristics, such as “Midwestern, rural, Protestant, lower income, female.” Many readers recoiled from the notion of being sorted into one of these boxes. The New York Times’s reviewer called The 480 a “shock novel” and found it implausible.

What was so shocking? What was implausible? The idea that a company might use computer technology and behavioral science to gather and crunch data on American citizens, with the nefarious goal of influencing a presidential election.

In the 1950s and 1960s this seemed like science fiction. Actually, The 480 was a thinly disguised roman à clef, based on a real-life company called Simulmatics, which had secretly worked for the 1960 campaign of John F. Kennedy. Burdick had been a political operative himself and knew the Simulmatics founders well. The company’s confidential reports and memoranda went straight into his prose. And the 480 categories—listed in an appendix to the novel—were the real Simulmatics voter types, the creation of what one of its founders called “a kind of Manhattan Project gamble in politics.”

Simulmatics was founded in 1959 and lasted eleven years. Jill Lepore mentioned its involvement in the Kennedy campaign in These Truths (2018), her monumental history of the United States; she was already on the trail of the story she tells in her new book, If Then. Lepore is a brilliant and prolific historian with an eye for unusual and revealing stories, and this one is a remarkable saga, sometimes comical, sometimes ominous: a “shadow history of the 1960s,” as she writes, because Simulmatics stumbled through the decade as a bit player, onstage for the Vietnam War, the civil rights movement, the Great Society, the riots and protests. It began with grand ambitions to invent a new kind of predictive behavioral science, in a research environment increasingly tied to a rising defense establishment amid the anxiety of the cold war. It ended ignominiously, in embarrassment and bankruptcy.

Irving Kristol, the future architect of neoconservativism, dismissed Simulmatics in 1964 as “a struggling little company which, despite the fact that it worked on a few problems for the Kennedy organization in 1960, has since had a difficult time making ends meet,” and he wasn’t wrong. Today it is almost completely forgotten. Yet Lepore finds in it a plausible untold origin story for our current panopticon: a world of constant surveillance, if not by the state then by megacorporations that make vast fortunes by predicting and manipulating our behavior—including, most insidiously, our behavior as voters.

The name “Simulmatics”—a mashup of “simulation” and “automatic”—was the kind of faux-sophisticated compound that tech startups favored (Teledyne, Biotronik, Microsoft), before a 1990s countertrend led to names like Yahoo, Google, and Twitter. Simulmatics’s creator and president was a former Madison Avenue advertising and public relations executive, Edward L. Greenfield—a glad-handing, scotch-drinking, Pall Mall–smoking huckster, as Lepore describes him. Greenfield saw national politics as virgin territory for a scientifically minded advertising business.

This was a fairly new idea. In 1952 the eloquent and erudite Democratic nominee, Adlai Stevenson, objected to paying professionals for tools of mass persuasion, while Dwight D. Eisenhower became the first presidential candidate ever to appear in television commercials. A word on people’s lips that year was “brainwashing,” coined by Edward Hunter, the rabidly anti-Communist author of Brain-Washing in Red China: The Calculated Destruction of Men’s Minds (1951). Stevenson wanted no part of that. Republicans spent $1.5 million on television advertising, which seems like a pittance until you learn that the Democrats spent $77,000.

Computers at the time were house-sized machines built from thousands of vacuum tubes, with tanks of liquid mercury serving as “memory.” The largest computer in the world, produced by the Remington Rand corporation for the Census Bureau, was UNIVAC. CBS News hired it for election night—Walter Cronkite called it “that miracle of the modern age, the electronic brain.” It was not just going to tally the numbers as they came in; it was meant to predict the outcome before the final results were known. “This is not a joke or a trick,” said his colleague Charles Collingwood. “It’s an experiment, and we think it’s going to work. We don’t know, we hope it will work.” Collingwood referred to the computer as “he”: “He’s sitting there in his corner, humming away. A few minutes ago I asked him what his prediction was, and he sent me back a very caustic answer.” In fact, UNIVAC’s operators saw early in the evening that Eisenhower seemed to be winning in a landslide, but they weren’t brave enough to trust their computer.

The pioneers of computing had a dream application for a machine capable of performing rapid calculations on large amounts of data: weather forecasting. Atmospheric behavior is just physics, and the relevant equations were well known. At the Institute for Advanced Study in Princeton, John von Neumann argued that computers would soon enable humans not only to predict the weather but to control it. He believed in a degree of perfection that we now know is unreachable, because the weather is a chaotic system, but the methods he invented are used in the supercomputers that run nonstop at all the world’s meteorological centers. Computers predict the weather by simulating it: by creating models and letting them run.

Ed Greenfield, the ad man, saw no reason not to try the same approach with the behavior of people. He created a Social Science Division in his company, with the idea of simulating the entire electorate and selling voting predictions to campaigns. He hired a computer guy and a social scientist, Bill McPhee and Ithiel de Sola Pool, and Eugene Burdick, then a political scientist. All of them grew obsessed in one way or another with the possibility of mathematizing human behavior. McPhee was ebullient about programming: “The computers (Univac, Manniac + ‘Johnniac’—I work on Johnniac) are just like big electric trains for grown-up boys.” Pool was working on a theory of “social networks.” He had a notion that “any two Americans, of any sort, can be linked by a chain of six acquaintanceships or less”—six degrees of separation, one might say.

Who were these people? “They were midcentury white liberals in an era when white liberals were not expected to understand people who weren’t white or liberal,” Lepore says. She calls them—they sometimes called themselves—the What-If Men. Deeply researched, written with elegance and passion, If Then gives a vivid picture of their lives, including their often miserable wives, suffering “the bad bargains of the middle-class marriages of the 1950s.” In The 480, Burdick describes the Simulmatics men as a “new underworld”:

Innocent and well-intentioned people who work with slide rules and calculating machines and computers…. Most of these people are highly educated, many of them are Ph.D.s, and none that I have met have malignant political designs on the American public. They may, however, radically reconstruct the American political system, build a new politics, and even modify revered and venerable American institutions—facts of which they are blissfully innocent.

Greenfield wanted to build a “Voting Behavior Machine.” He also called it a “People Machine.” It would treat voters as “discrete units at the ‘microscopic’ level,” and it would create a “macroscopic picture of how, when it is all put together, the aggregate system works.” For the first time, large quantities of data had been gathered through the booming business of political polling. When Greenfield and his team formed Simulmatics in advance of the 1960 election, they bought the results of Gallup polls from the past four national elections, 100,000 surveys on punched cards. They sorted the questions into fifty “issue attitudes,” and they sorted the people into demographic boxes.

First they tried selling their services to Stevenson, their favorite among the potential Democratic candidates. They proposed to test policy positions in the machine before revealing them to actual voters. Should the candidate speak out for civil rights in the South? “We will, from our model, be able to predict what such a speech would mean to each of 1,000 sub-groups of the population…and consequently to pinpoint the state where it could affect the electoral vote,” Greenfield wrote. This was liable to strike some as cynical or sneaky. It should strike us as a description of what every technically astute political consultant offers today.

When the nomination went instead to Kennedy, they sent their pitch to his campaign. It went like this:

With speed, accuracy, precision, and efficiency, simulation provides a unique test tube for the political strategist. Before he acts, it tells him what would be the effect of his handling of certain issues on the voting behavior of the electorate.

And like this:

The relationship between such current intelligence and a simulation model developed out of historical data is analogous to the relationship between current weather information and a climatological model.

The Kennedy team commissioned three reports—urgently, because, for the first time, the rival candidates had agreed to televised debates, and the dates were approaching.

The Simulmatics men worked fast. No one can say now how much of their advice came from the computer, an IBM 704 at Columbia University, and how much was old-fashioned political savvy. “There’s a lot of bluster and nonsense in the archival trail left behind by flimflam men,” as Lepore says. But the advice seems to have been good. Black voters were grossly underrepresented in the 480 categories because George Gallup and other pollsters had avoided surveying them—meekly accepting the reality of Jim Crow voter suppression and kowtowing to southern newspaper owners who expressly did not want Blacks polled. But Simulmatics counted them and made the case that Democrats could not win without supporting the civil rights movement. They also advised the candidate to speak out frankly about the anti-Catholicism that dogged his campaign. And they urged him to treat the debates as an opportunity: “The danger to Nixon is that Kennedy can make use of his more personable traits—including a range of emotions such as fervor, humor, friendship, and spirituality.” Kennedy did everything Simulmatics suggested, and when he carried off his narrow victory in November, the What-If Men were sure they deserved the credit.

And that was it. The sum total of Simulmatics’s intervention in this (or any) political campaign amounted to a few bits of advice, ostensibly computer-generated but really, at most, computer-assisted. No voters were manipulated or deceived. No one’s privacy was invaded.

The Kennedy team thought they had bought confidential research reports, but the Simulmatics people were desperate for publicity. One of them, Thomas B. Morgan—a friend of Ed Greenfield who had helped him edit the three reports—approached Harper’s pretending to be a disinterested journalist, and the magazine featured his sensational account in January 1961, the month of Kennedy’s inauguration. “The People-Machine,” screamed the headline. “The First Report on a Computing Device Secretly Designed for the Democratic Presidential Campaign—and on Its Consequences for Political Strategy.” It was a puff piece posing as an exposé. It quoted another Greenfield friend, the Yale political scientist Harold Lasswell, as saying, “This is the A-bomb of the social sciences.”

All hell broke loose. United Press International reported:

A secretly designed robot campaign strategist nicknamed a “people-machine” was said today to have been put to work by President-Elect John F. Kennedy’s top advisers to suggest alternative methods of influencing voters.

The New York Herald Tribune said Kennedy’s “secret weapon” had been “a big, bulky monster called a ‘Simulmatics.’” An Oregon paper editorialized, “The voters—you, me, Mrs. Jones next door, and Professor Smith at the university” were reduced “to little holes in punch cards, or whatever device our new Ruler uses.” The Kennedy people panicked and tried to deny the whole thing.

These still unfamiliar machines—thinking machines—scared people, for good reasons and bad. Machines couldn’t really think, by any definition. They were hardly monsters or robots. They might or might not be threatening to take jobs from people: Kennedy himself had warned on the campaign trail of “the growing crisis of automation—the replacement of men by machines.” Even so, reliance on computers seemed an attack on human agency: you and me, reduced “to little holes in punch cards.” Simulmatics hyped its own power and prowess but also tried to dampen these fears with boilerplate reassurances: “Machines can do nothing but speed up communication.” By doing so, “they restore the possibility of ready discourse about important matters in large societies.” Nowadays, that sounds naive. The same kind of hopeful innocence was heard again three decades later, with the rise of the Internet.

Computers were, of course, poised to grow in speed and power more rapidly than the most optimistic What-If Man could have dreamed. For Simulmatics, the promise of computing and the hope of predicting human behavior were tied together. “The scientists of the Simulmatics Corporation acted on the proposition that if they could collect enough data about enough people and feed it into a machine,” Lepore writes, “everything, one day, might be predictable, and everyone, every human mind, simulated, each act anticipated, automatically, and even driven and directed, by targeted messages as unerring as missiles.” For Simulmatics, Bill McPhee wrote down what he called “a set of rules of thinking” to “reproduce or better simulate the thinking behavior of voters.” (That he wrote the rules from inside the psychiatric ward of Bellevue Hospital, where his wife had temporarily committed him, is the kind of detail a novelist might be ashamed to invent.)

Around the same time, the science-fiction writer Isaac Asimov began his Foundation series of novels, premised on the idea of a science of psychohistory. The sages who mastered this science could predict the rise and fall of empires: “Psychohistory was the quintessence of sociology; it was the science of human behavior reduced to mathematical equations.”¹ Asimov’s fictional future didn’t need computers to predict its future—just geniuses, making magic with their formulas. The difference between Asimov and Simulmatics was that Asimov knew he was spinning a fantasy. People do not behave as predictably as billiard balls or water molecules. Atmospheric simulation relies on variables like temperature, pressure, and humidity; the science of psychology has discovered no mental or social quantities that can be defined as simply or measured as precisely. No matter how powerful our computers are, we have no scientific laws of behavior to feed them.

Having begun the 1960s with a storm of publicity, Simulmatics set out to expand its reach from campaign politics to consumer marketing. The company pitched projects to film studios and music labels, along with advertising agencies—anyone they thought might buy the notion of simulating the needs and wants of masses of consumers. It offered, Lepore writes, “a simulated population, a miniature United States, consisting of three thousand perfectly representative but entirely imaginary people.” Punch, the British satirical magazine, mocked this new approach to reading society’s mind: “Punch proposed that Simulmatics’ Media-Mix add a few more categories,” Lepore writes,

including “dog-lover, flat-earther, doughnut dunker, milk-in-firster,” so as to be able to determine, for instance, whether a new, seventy-five-cent pink ballpoint pen ink refill would be desired by consumers who shop on Wednesdays, have Republican sympathies, “fruit-juice breakfasts, bouts of depression, slight astigmatism, fitted carpets, thick eyebrows, and one or more cousins in the armed forces.”

Which is all very amusing, and exactly what Facebook, Amazon, and Google do today.

The company also pursued government contracts. Its social scientist, Pool, now chairing MIT’s new political science department, got the Defense Department’s Advanced Research Project Agency to finance a foray into psychological warfare to be called Project ComCom, for Communist Communications. As the United States committed itself more and more deeply to the war in Vietnam, Simulmatics went there, too. After all, psychological warfare was just political persuasion militarized, and here was fertile new territory. “Vietnam is the greatest social-science laboratory we have ever had!” Pool told a colleague. Robert McNamara, the numerically oriented secretary of defense, wanted to win hearts and minds, and along with the soldiers he brought in behavioral scientists as civilian contractors.

Simulmatics and its much larger competitor, the RAND Corporation, set up shop in nearby Saigon villas. One of RAND’s analysts was a former marine named Daniel Ellsberg, who later helped write and then leaked the devastating secret report known as the Pentagon Papers. These researchers were an odd presence in a bloody war. “Young men from RAND and Simulmatics bounded about the countryside in Land Rovers studying ‘upward mobility among village elites’ or ‘the interrelationship of land reform with peasant political motivation,’” wrote Frances FitzGerald in her prize-winning history Fire in the Lake. Twenty-six years old when she arrived in Vietnam, she thought they were remarkably naive.

By Lepore’s account, the Simulmatics operation was something of a clown show. Its staffers, who spoke little or no Vietnamese, got their fingers into various pieces of the US adventure, including the notorious pacification program as well as a psychological warfare project known as Chieu Hoi. These relied more on old-fashioned propaganda methods—and brutal coercion—than anything involving special data analysis. Ed Greenfield arrived in 1966 with a plan to set up a “paramilitary action research center.” But the Pentagon was realizing how much the company had overpromised and how little it had done, and at the end of 1967 the team was sent home.

On their way to obscurity, the Simulmatics people played minor parts in major events, appearing Zelig-like at crucial moments of 1960s history. After race riots in Detroit, Los Angeles, and other cities, Simulmatics conducted a study for the Kerner Commission, established by Lyndon Johnson to investigate the riots and their causes. One sociologist told the commission, “Until we can predict riots with at least as much reliability as the weather forecast predicts rain or snow, it will be very difficult rationally to devise policies of control.” That was a last gasp. In 1968 the company shut down for good, bankrupt. It stiffed its last consultants, including future senator Daniel Patrick Moynihan. It had accomplished nothing.

What seemed frightening and even immoral fifty years ago we now mostly take for granted. We shouldn’t. Facebook places its billions of customers into many more than 480 categories, based not on voluntary surveys but on pervasive surveillance. Machine-learning techniques intuit cultural “affinities” and political preferences. The algorithms sort users by location, education level, languages, financial standing, property ownership, occupation, age, gender, sexual preference, and relationship status. They track almost everything you buy, read, or watch. Facebook knows who is connected to, related to, and interested in whom.

Everyone knows about Facebook. Hardly anyone had heard of Cambridge Analytica four years ago, when it helped elect Donald Trump. Established in London by two wealthy conservatives, Robert Mercer and Steve Bannon, it pitched itself as a behavioral research and political consultancy, “operating through different vehicles, in the shadows.” It offered “psychographics.” It secretly harvested personal Facebook data from more than 50 million Americans, ostensibly in violation of Facebook’s policies.

Through Cambridge Analytica and Trump’s separate digital operation, run by Brad Parscale, users of Facebook, Google, Twitter, Snapchat, and others were targeted with thousands of messages, many of them false, continually updated and tweaked. “We bombarded them through blogs, websites, articles, videos, ads, every platform you can imagine,” said one Cambridge Analytica executive:

Until they saw the world the way we wanted them to. Until they voted for our candidate. It’s like a boomerang: you send your data out, it gets analyzed, and it comes back at you as targeted messaging to change your behavior.

The Trump campaign designed negative ads to inhibit at least three categories of voter: African-Americans, former Sanders supporters, and young women who might resent Bill Clinton. Apart from all that, Trump’s sometime campaign manager Paul Manafort is known to have shared sensitive polling data with his associate Konstantin Kilimnik, who has been identified by the Senate Intelligence Committee as a Russian intelligence officer.

That was 2016. This fall, a new barrage of campaign messaging is well under way, as covert and hydra-headed as ever. Cambridge Analytica is gone, but Facebook has only grown more powerful. In September, its founder and chief executive, Mark Zuckerberg, announced that the company would refuse new political advertisements in the final week before the election—a token gesture that served mainly to highlight the extraordinary control one man exercises over the core communications platforms serving the United States and most of the world. Zuckerberg continues to allow, for example, videos and photographs falsified by the Trump campaign to depict Joe Biden as a dotard. Officially Facebook bars “misinformation,” but the rules don’t apply to political advertising.

Jaron Lanier, a virtual-reality pioneer who has become a forceful critic of social media, describes this as a “mass behavior modification machine.”² Its power comes from a feedback loop: algorithms choose what information to present each user at each moment, aiming to catch and hold the user’s attention; the user, by viewing or “liking” particular shiny objects, trains the algorithm to perform better and better. Fueling anger and outrage has proved especially effective. The machine may not be a monster, but it brings out the monsters in us. Truthfulness is of no interest to the algorithm. Facebook’s goal is to get everyone addicted to Facebook, but the company also sells its manipulation services to companies with different goals. “For instance,” Lanier says,

if you’re reading on a device, your reading behaviors will be correlated with those of multitudes of other people. If someone who has a reading pattern similar to yours bought something after it was pitched in a particular way, then the odds become higher that you will get the same pitch. You might be targeted before an election with weird posts that have proven to bring out the inner cynic in people who are similar to you, in order to reduce the chances that you’ll vote.

Lepore’s What-If Men didn’t cause any of this, of course. One of them saw it coming, though, with unusual prescience. In 1968 Ithiel de Sola Pool contributed an essay to Toward the Year 2018, a book of “amazing predictions of what life will be like 50 years from today.” Everything now filed on paper will be stored in computers, he said: “Tax returns, social security records, census forms, military records, perhaps a criminal record, hospital records, security clearance files, school transcripts…bank statements, credit ratings, job records.” And so?

By 2018 the researcher sitting at his console will be able to compile a cross-tabulation of consumer purchases (from store records) by people of low IQ (from school records) who have an unemployed member of the family (from social security records). That is, he will have the technological capability to do this. Will he have the legal right?

Fifty years later, we can answer Pool’s question. Yes, the researcher—and the world’s most powerful corporations—have the legal right to compile these dossiers. They also have the legal right to feed us a diet of misinformation, conspiracy theories, and clickbait for the benefit of marketers and unscrupulous politicians.

It didn’t have to be this way. That is Lepore’s final message: history is not inevitable. “Plainly, all of this might have gone differently,” she says. “Plenty of people believed at the time that a people machine was entirely and utterly amoral.” The Internet was born in an ethos of free-market utopia and contempt for centralized authority.

Previously, radio and then television broadcasters had submitted to federal regulation. The Federal Communications Commission had the difficult responsibility of balancing issues of free speech, fairness, and truth in advertising. But when people’s personal computers gained the ability to communicate with one another as if by magic, across a worldwide web, hardly anyone wanted the government to lay its heavy hand on this free-wheeling, decentralized, individualistic phenomenon. And hardly anyone wanted to pay for information, even if it came from newspapers or books—which meant that corporate entrepreneurs turned instead to advertising and personal-data collection as sources of revenue.

“The new Internet followed no rules but many mantras,” Lepore writes scornfully. “Content must be free. Media solves all problems. Data drives predictions.” To this day the government has failed to establish effective rules and standards to safeguard the ownership, collection, and marketing of personal information. Facebook, Google, and Amazon all lobby hard to prevent that.

When Eugene Burdick walked away from the Simulmatics Corporation, he did leave us a warning. “This may or may not result in evil,” he wrote. “Certainly it will result in the end of politics as Americans have known it.”