Nobody wants to
go first.

The hall of mirrors

Every action waits on a guess about a guess.

Set two people down who have never met and give them a shared problem — how to greet, how to divide, how to begin. Each is, in the jargon, an alter ego to the other: a centre of choice that could do otherwise. So Ego reasons: I should do what fits what Alter will do. But Alter is not a rock with a fixed behaviour to be predicted; Alter is reasoning the same way — I should do what fits what Ego will do. To know my best move I must know yours; to know yours you must know mine. The regress has no bottom. This is what Parsons called double contingency: not one uncertainty but two, each feeding the other, my expectation propped on my expectation of your expectation of my expectation, all the way down to no floor at all.

On paper it is a perfect deadlock. If each strictly waits for certainty about the other before moving, neither ever moves; the encounter should freeze. And yet encounters do not freeze. Strangers do greet, negotiators do settle, the floor does fill. Something gets the system off the ground that pure logic says should be stuck. The instrument above is the smallest honest machine that reproduces the deadlock and escapes it — so you can watch exactly what does the lifting. Each agent holds a guess about the other (the two distributions), acts on it, sees what actually happened, and quietly revises. Nothing else. No dictionary, no designer, no first mover who simply decides.

What to try

Four things the instrument will show you.

Cold-start two strangers and watch an order condense. Leave the defaults and press Run. The two distributions start nearly flat — a tie in every direction, the deadlock made visible. But no tie is perfect: a little sampling noise tips Ego a hair toward one greeting, Ego acts on it, Alter sees that and tips the same way, which makes Ego surer, which makes Alter surer. Positive feedback does the rest. Within a few thousand rounds both bars spike on the same move — the same colour on both sides — and the coordination line climbs off its chance baseline toward one. A little social order, condensed out of nothing but mutual guessing.

Reseed, and get a different order. Press ⟳ Reseed and run again. Order forms just as reliably — but often on a different greeting. Which convention wins is a frozen accident of who trembled which way first; that some convention wins is all but guaranteed by the feedback. The order is contingent; its existence is not. Reseed a few times and you will see several different little societies, each stable, none inevitable, none designed.

Crank the noise and watch it never settle. Push Noise up past a third. Now hands tremble too often for expectations to earn their keep: every time the beliefs begin to sharpen, random moves muddy the record they are learning from. The bars never fully spike, the coordination line hovers, restless, above chance but far below one. Some encounters are simply too turbulent for an order to crystallize — and the model will not pretend otherwise.

Deadlock two stubborn strangers. Press Two stubborn strangers: it sets five possible greetings and a learning rate near zero, so neither can revise fast enough to get any traction on the other. Coordination sits pinned at the chance floor for tens of thousands of rounds — the pure Parsons deadlock made flesh, two people who could mesh in an instant if either would bend, held apart precisely because neither will move until the other has. Watch long enough and it does, finally, tip: a coordinating pair, given endless patience, always escapes eventually. What stubbornness buys is not a permanent freeze but an agonizingly long one — a truly permanent deadlock needs conflicting interests, a case this coordinating model cannot reach.

Order from instability

Luhmann's inversion: the uncertainty is the raw material.

Parsons saw double contingency as a problem that had to be solved before society could exist — and reached for a solution outside the encounter: a pre-given shared culture, a common code of values already in both heads, that tells each what to expect. Luhmann's move was to refuse the rescue. There need be no code beforehand, he argued, because the instability is self-resolving. Precisely because each is uncertain what the other will do, each watches the other with total attention; every actual move is enormous information, seized on and built into expectation. The very openness that looks like a bottomless pit is what makes the two hypersensitive to each other — and that sensitivity is the coupling that lets an order nucleate. Uncertainty is not the wall. It is the reactivity that lets the first accident propagate.

What stabilizes in the instrument is not the moves — it is the expectations. Ego does not just learn to greet a certain way; Ego learns to expect Alter to greet that way, and to expect that Alter expects the same of Ego. That interlocking — expectation of expectation — is, for Luhmann, the first genuinely social structure, the elementary particle of a system. It is more than the sum of two guesses, because each guess now includes the other. Once it locks, the pair can coordinate without deliberation, without checking, almost without attention: the structure carries the load that logic could not. Order here is not something added to the encounter from outside. It is the encounter's own instability, folded back on itself until it holds a shape.

From expectation to norm

How an "is" quietly becomes an "ought."

A settled expectation starts as a mere prediction — this is how we greet — but it does not stay merely predictive for long. Turn on Sanction and mismatches are now penalized: to break from the expected move is not just to fail to mesh but to cost both parties. It widens the gap between conforming and deviating, so each agent's best reply to its expectation sharpens; the order locks faster and holds harder. The very same mutual expectation that was once a fragile guess becomes a thing you are punished for violating. The prediction has grown teeth.

This is, in miniature, the birth of the ought from the is. Once "we all expect each other to do X" is common ground, doing X stops feeling like a choice and starts feeling like a requirement; not-X reads as a breach, a rudeness, a betrayal — and gets met with a frown, a fine, an exclusion. Luhmann called stabilized expectations that survive their own violation normative: you keep expecting the greeting even when someone flouts it, and you sanction the flouter rather than update your expectation. That is the moment a coordinated accident hardens into a norm — an order that no longer merely describes what the two do, but demands it. The hall of mirrors has grown a floor, and the floor has grown a rule.

The mapping

Mechanism ↔ social life.

Where it tears

Three honest limits.