In 1959, a company was founded in New York that sought to achieve what then sounded like science fiction: predicting human behaviour with a computer. It was called Simulmatics. It gathered over one hundred thousand old survey interviews, divided voters into four hundred and eighty categories, and simulated how each group would react to campaign issues. It worked for the Democratic Party in the 1960 election, which Kennedy won. Subsequently, the company itself propagated the notion that its machine had been the secret weapon of that victory. The legend exceeds the facts: there was a contract and there were reports, but the claim that it won the election is a narrative Simulmatics sold very effectively. The history has been reconstructed by the historian Jill Lepore. She is primarily interested in what the machine signified for politics.
For Simulmatics did not intend to remain confined to the ballot box. One of its technicians, Alex Bernstein, attempted to construct models for book and magazine publishers as well as film distributors. The premise remained identical to that applied to voters: to forecast sales volumes, subscription rates, and cinema attendance. To translate the calculus of the elector to that of the reader and the spectator. Yet, within the realm of culture, the project faltered. This was not due to moral scruples or theoretical limitations, but for a purely material reason: insufficient data existed to populate the models. The ambition was present; the infrastructure required to fulfil it was not.
That failure marks precisely what has changed. What Simulmatics envisioned and could not accomplish in 1960—anticipating which culture each type of audience would consume and adjusting what is offered in advance—is, almost word for word, what recommendation platforms now do effortlessly. What was lacking then is exactly what is now in abundance: data. Every playback, every pause, every abandonment after ten seconds, every search, now feeds the models that the company could not build.
I am not suggesting that Simulmatics invented Netflix, nor that there is a direct line from that company to current algorithms. There is not: today's systems were born independently, using different techniques and in a different world. What exists is not an inheritance, but a repetition. The same intellectual operation—converting audiences into calculable types, predicting their response, and modulating what is made visible to them—appears in 1960 without the means to be realised, and reappears today with every means available. It is not a lineage. It is an idea that waited sixty years for technical reality to catch up with it.
And the implications of that idea, now realised, are significant. When a system predicts your preferences and organises the content presented to you accordingly, it does more than facilitate your daily experience. It intervenes in a more profound matter: which works are permitted to appear before you and which are not. That which does not conform to any calculable category, that which bears no resemblance to your previous consumption, the unusual, the deliberate, the signal-less, tends to become invisible. Not censored, but something more efficient: simply not recommended. A work that remains undiscovered is, for all practical purposes, a work that barely exists within the public sphere, regardless of its presence on a server.
One need not be nostalgic or technophobic to recognise the issue. Prediction is not inherently detrimental; organising abundance is beneficial, and no one wishes to return to searching blindly through millions of files. What must be addressed directly is who determines the criteria for such prediction, to what end, and what is systematically excluded. For the taste these systems claim to serve is, in part, a taste they themselves manufacture: they offer more of what has already been chosen, and in doing so, they gradually narrow the scope of what one may eventually choose.
There were those who foresaw it at the time. When it was proposed during the 1960 campaign to condition the candidate's message upon what the machine dictated, the historian Arthur Schlesinger wrote that the idea of a man not speaking until he had consulted the computer made him tremble. He was speaking of politics, but the warning applies equally to culture: the problem is not that the machine calculates, but that we allow its calculation to decide in advance what deserves to be said, shown, or seen. Simulmatics could not close that circle with books and films because it lacked the data. We have provided it with all of it. The question they could not answer in 1960 was not technical: it was not whether one can predict what we will watch, but what is lost when someone predicts it for us.
On the open conversation
This text originates from an episode in the history of computing—the Simulmatics Corporation as reconstructed by Jill Lepore—to examine a question central to culture: what occurs when the prediction of our behaviour determines which works are made visible to us? I carefully distinguish the historical fact from the interpretation, and I propose a genealogy of a logic, rather than a causal continuity between that enterprise and contemporary systems. Should anyone wish to contribute from the perspectives of the history of technology, platform studies, the sociology of culture, or the practice of those who publish and distribute works under these systems, this notebook remains open.
Sources
Lepore, Jill. If Then: How the Simulmatics Corporation Invented the Future. New York: Liveright, 2020. (Historical basis regarding Simulmatics; the theses concerning its legacy are the author's interpretation.)
Lepore, Jill. 'Democracy v the machine', The Guardian, The Long Read, 18 August 2026, in relation to her book The Rise and Fall of the Artificial State (2026).
Pool, Ithiel de Sola, and Robert Abelson, 'The Simulmatics Project' (report on the work for the Democratic Party in the 1960 campaign).
Regarding early warnings: Arthur M. Schlesinger Jr. (1959-1960); Eugene Burdick, The 480 (1964); Lewis Mumford, The Myth of the Machine (1967).