22 November 2020

Book Review: If Then by Jill Lepore

If Then: How the Simulmatics Corporation Invented the Future
By Jill Lepore
Liveright Publishing, 432 pages

The guys who invented predictive analytics never saw failure coming.

That’s the upshot of Jill Lepore’s latest book, If Then: How the Simulmatics Corporation Invented the Future

Ostensibly, it’s the story of Simulmatics, founded in 1959 on the idea that with enough data collected in one place, everything and everyone would become predictable. The name is an attempted portmanteau combining the words “simulation” and “automatic.” You’ve probably never heard of Simulmatics because it folded in 1970, but during its short history it played a role in electing John F. Kennedy, mismanaging the Vietnam War, seeking answers to 1960s social upheaval, and speeding the presence of mainframe computers at advertising agencies.

If Then: Book Summary

The founder of Simulmatics was Ed Greenfield, a midcentury ad man, but not like Don Draper. Lepore delightfully introduces him: “He was like a ten-million-volt Looney Tunes electric magnet, a giant red-handled iron U that pulled everyone toward him.” His personality, his ability to influence others, was what propelled him. As evidence, the story includes a lot of bold-faced names, especially from Democratic Party politics, which is what Greenfield cared about most.

Indeed, he built an impressive team. Lepore introduces the other main players early, and efficiently. Harold Laswell, the influential communications theorist. Eugene Burdick, novelist and self-styled adventurer. Alex Bernstein, mathematician and computer programming pioneer. Ithiel de Sola Pool, a social scientist specializing in technology. Bill McPhee, a FORTRAN programmer – and this is such an emblematic aspect of the story – who wrote “the core intellectual property” of Simulmatics while he was committed to Bellevue. Yes, a mental hospital.

on parade

Like any startup, the group had big plans. They bragged they had invented “the A-bomb of the social sciences.” They called it a “People Machine” that could predict the outcomes of advertising campaigns and government policy initiatives. Sadly, they couldn’t get out of their own way. They overplayed their true role in JFK’s winning presidential campaign of 1960. They overpromised how they could help the New York Times analyze the 1962 midterm elections in real time. They overestimated, tragically, how Western-style social science techniques could understand Vietnamese culture. They oversold their value to blue chip brands but opened the door to a legion of market research providers still selling soap today.

One gap in the story: What projects did they actually finish? The only projects fully described were the political ones, and there was only fleeting mention of having sold studies to various corporations, like Bristol Laboratories, Philip Morris, P&G, and some others. Simulmatics was always starved for data, so most of the projects had little effect. Still, it would have been interesting to read more about those episodes.

Eventually Simulmatics folded, although some of its work survived in projects undertaken by individual team members, thus laying the groundwork for today’s data-driven marketing. They accomplished just enough to push things forward, but not enough to get pinned with credit or blame for what we have now. Oddly, Simulmatics’ most accurate predictions came not from data but from the very human insights of Ithiel de Sola Pool. He envisioned with eerie accuracy the role of technology in our lives today: the interconnectedness of the World Wide Web, the ubiquity of social media, and the rise of “mobile computers,” today’s smartphones.

Why Simulmatics matters now

Lepore’s book is thoroughly researched and well-written. It’s a solid history, which is why Simulmatics matters: because we learn from history. Here’s what I took away:

  • No data. It shouldn’t have been surprising, but was nevertheless shocking, how Simulmatics never seemed to have data that were complete or accurate. In an almost poignant moment, Lepore writes, “Pool raised the question that Simulmatics would never really answer: ‘What is the data we would need for this model?’” Ad agencies, which had data, filled the gap, bringing in their own IBM mainframes and offering the services to clients directly. Today we have plenty of data, but we still have to answer the question: Which data do we need to solve this problem?
  • No humility. The Vietnam phase of the book is a troubling read. Defense Secretary Robert McNamara in 1962: “Every quantitative measurement we have shows we are winning the war.” That might have been all too true; Lepore points out that military progress was measured by “the number of insurgents killed,” with the implication that indiscriminate killing ran up the numbers. Humility is a function of introspection. Are we thinking things through? Are we seeing the big picture? Are tracking the right metrics? These questions are relevant to the work we do today.
  • No humanity. Lepore points out that computers can simulate a flight because physical laws like F=ma are constant. “But the computer simulation of human behavior … is much more difficult. Behavior is not a law.” If, as some Artificial Intelligence experts say, the brain is just a very sophisticated machine, then eventually we will create a machine that can think like a human brain. But there is a (so far) unquantified human element that no series of If-Then scenarios in FORTRAN, C++ or Python could ever predict.

Simulmatics failed where other succeeded. There’s still lots of room for modern failure, which is why these lessons from the past are important.

1 comment: