On the twenty-fourth of November, 1960, ten men met in the basement of a Paris restaurant to discuss a literature that did not yet exist. They had no funding, no institution, and no readers. They gave themselves a name that sounded like a department of the post office: Ouvroir de littérature potentielle, the workshop of potential literature. Oulipo.
The membership was odd. Raymond Queneau was a novelist who had spent years on mathematics and once tried to axiomatise the plot. François Le Lionnais was a chemical engineer and chess theorist who had survived deportation to a concentration camp, Mittelbau-Dora, by reconstructing paintings from memory with another prisoner, the two of them describing brushwork to each other in the dark.
Claude Berge was a graph theorist; there were also poets, a publisher, and a man whose declared field was imaginary solutions. They met monthly, ate well, and set themselves one task: to invent constraints, and then to write under them.
The constraints were absurd. Queneau printed ten sonnets on strips so that any line could follow any other, producing a hundred thousand billion poems and, by his own calculation, nearly two hundred million years of reading at twenty-four hours a day. Georges Perec, who joined in 1967, wrote a three-hundred-page novel without using the letter e. Italo Calvino was elected in 1973 and built a novel out of a deck of tarot cards; he had already written Invisible Cities, one of the greatest books ever written.
The group's own account of itself, attributed to Queneau, is the best sentence any of them wrote: they were rats who build the labyrinth from which they propose to escape.
This runs against most of what we believe about creation. We treat constraints as the enemy of work and freedom as its precondition: give people room, clear the diary, and something will come. Oulipo held the opposite and demonstrated it for sixty years. A rule that is arbitrary, absolute and indifferent to your intentions closes off the routes you would have taken by habit and leaves you in front of the ones you would never have found. Perec did not lose the letter e. He gained a novel about absence that could not have been written any other way.
What interests me is what happens when the constraint is applied to a group instead of a page. Not what makes a person creative, which has a large and mostly disappointing literature (unless written by Robert Caro), but what makes an assembly of people produce work none of them could produce alone. The clearest model I know comes from ants.
E. O. Wilson spent the 1950s at Harvard working out how fire ants communicate. He suspected chemistry, dissected the ants gland by gland, and found the source in the Dufour's gland, a structure smaller than a comma. He crushed it, drew a trail across a sheet of glass with a sharpened stick, and watched a column of ants pour out of the nest and follow a road to nothing at all. It is a small experiment and it contains everything.
The ant has no map, but it has a rule about a smell. A colony solves problems that no single ant understands. It finds the shortest route to a food source, allocates workers between foraging and brood care and nest repair, and adjusts within hours to a flood or a raid, with no supervisor and no plan. The queen issues no orders; she lays eggs. What the colony has instead of management is a set of local rules that each ant applies to what it can smell and touch in the next few centimetres of its own life. Task allocation in harvester ants runs partly on how often a returning forager meets other ants near the nest entrance. The decision is made by contact frequency. No one individual decides.
Two features of this are important. The first is that the rules are tight about form and silent about action. An ant is not told where to go. It is told what to do when it meets a gradient. The constraint sits at the level of the mechanism, and above that level the ant is on its own, which is the only arrangement that can work, because the information is local and the situation changes faster than any instruction could travel (sound familiar).
The second is idleness. In colonies studied carefully, a large fraction of the workers are doing nothing at any given moment. They stand about. They groom. Under a consultant's gaze this is half a workforce being wasted, and the obvious efficiency would be a catastrophe. The inactive fraction is the reserve that lets the colony answer a perturbation it did not anticipate, and a colony run at full utilisation can respond to nothing. Slack is capacity held in a form that cannot be scheduled.
Local rules, no supervisor, unaccounted time. Where a group has done work that outran the sum of its people, some version of those three conditions is usually in place, occasionally by design and more often by luck.
The most deliberate case I know is at the Santa Fe Institute. Santa Fe was incorporated in 1984 by Los Alamos scientists, led by the chemist George Cowan and joined by Murray Gell-Mann, on the grounds that the interesting problems had fallen into the gaps between disciplines and there was nowhere to work on them. Their solution was to remove almost everything an institution provides. No departments. No permanent faculty, and so no tenure clock, no field to defend, no incentive to publish inside a lineage. Scientists come for months or years and leave. There are blackboards in the corridors and on the terrace, so that an argument begun over coffee can be fact checked, in writing, immediately.
Santa Fe is usually described as an experiment in removing structure, but I disagree. It removed one structure and installed a tighter one. The residency is a constraint: fixed period, people you did not choose, not enough time to build a fiefdom. So is the absence of departments, since you cannot address only your own field when your own field is not in the room.
Bell Labs had found this forty years earlier. Laying out the Murray Hill building in 1941, Mervin Kelly made the corridors long and put the theorists among the engineers, on the reasoning that a physicist walking to lunch should be unable to avoid someone with a different problem. Claude Shannon rode a unicycle down those corridors juggling, and in a decade of almost complete freedom from assigned work he invented information theory. The building arranged the collisions; the freedom allowed something to come of them.
Richard Hamming worked in the same building and paid closer attention than anyone to why a few of his colleagues mattered and most did not. His conclusion was that the difference was not ability. The people who kept their doors shut got more done on any given day and were, twenty years later, working on problems nobody cared about. The people who left their doors open were interrupted constantly, knew what everyone else was struggling with, and worked on things that mattered. He kept Friday afternoons clear for what he called great thoughts, and made himself unpopular at lunch by asking physicists what the important problems in their field were, and then whether they were working on them. The open door is a constraint voluntarily accepted.
Now consider the places most of us actually work. Many universities and most companies run the inverse arrangement: loose about form, tight about action. A doctoral student gets a five-year cycle, a supervisor to be updated, a literature to be positioned within, and a question chosen for the probability that it yields a publishable answer inside sixty months. What gets optimised is legibility, the ability to say in a report what was done and why it was worth doing, in advance of knowing either.
The open-plan office is the sharpest case, built on the right theory with the wrong mechanism. It was meant to be the Murray Hill corridor: take down the walls and the collisions will follow. When it was studied properly, at two firms whose staff wore sensors before and after the walls came down, face-to-face interaction fell sharply and email rose. Removing the walls removed the conditions for a private conversation, and people responded by having none, or by putting on headphones, which is a wall a person can afford. A corridor works because it is a constrained space you must pass through on the way to somewhere else.
Then the calendar, where the third condition dies. Unaccounted time cannot be defended in a planning meeting, because by definition you cannot say what it is for. So it goes first, and its removal is recorded as an improvement: the decision to run at full utilisation, taken by people who will not be there when the flood arrives.
Deborah Gordon has followed harvester ant colonies in the Arizona desert for four decades. A worker ant lives about a year; a colony lives twenty or thirty, and over that span its behaviour changes. Young colonies are erratic and overreact to disturbance. Older ones are steadier, less easily provoked. The colony learns, though no ant learns anything, no ant is present for more than a fraction of the learning, and there is nothing inside a single insect that could store it. The knowledge sits in the relations between them, and in rules that keep being applied after every one of the original ants is dead.
Oulipo still meets. Perec died in 1982 and Calvino in 1985, and both remain members, because the constitution the ten men wrote in that basement holds that nobody leaves, and the dead are recorded in the minutes as excused. The rules outlasted the rats. The labyrinth is still standing, and people are still working inside it, which was always the point of building one.