the second hundred · metaphor 143
Why do some customs hold for generations with nobody policing them — how you queue, which side you drive on, when you fight and when you back down — while a newcomer who breaks them simply loses, and drifts back into line without anyone saying a word?
Not every rule needs a rulebook. Some hold because breaking them doesn't pay. The village where everyone is courteous isn't kept polite by a magistrate; it's kept polite because a lone boor, dropped into a town of the mild-mannered, does worse than the mild-mannered do — gets frozen out, loses the deals, wins the argument and loses the friend. Their rudeness can't spread, because it can't gain a foothold. The norm defends itself.
This is a strange and powerful kind of stability: no leader, no punishment, no coordination — just a mix of behaviours so balanced that any rare deviation earns less than the crowd it tries to invade, and so fades. Biologists found the idea while asking why animals bluff and posture instead of fighting to the death. The answer turned out to describe conventions everywhere: an evolutionarily stable strategy — a way of behaving that, once common, no mutant can beat.
The idea
Take the classic contest. Two animals meet over a resource worth V. A Hawk escalates and fights; a Dove displays and yields. Two Hawks fight and risk injury worth C; a Hawk against a Dove takes the prize; two Doves split it. The trick is that no single behaviour is best against itself. In a world of all Doves, a Hawk cleans up — so Hawks spread. In a world of all Hawks, everyone's bloodied and a peaceable Dove does better — so Doves spread. Each strategy is beaten when it's common.
So the population doesn't go to all-one-thing. It settles at a mix — a fraction of Hawks p* = V/C — where a Hawk and a Dove earn exactly the same. That balance is the evolutionarily stable strategy. Its defining test is invasion: seed the stable mix with a few mutants of any kind, and because the mix already equalizes the payoffs, the rare mutant earns less than the residents the moment it tips the balance — so it shrinks back. The engine that shrinks it is replicator dynamics: whatever earns above the average grows, whatever earns below it fades, generation by generation, with no one steering.
Notice what's absent: no referee, no agreement, no memory. The stability is a property of the payoffs, not of any authority. That's why an ESS is the natural mathematics of a convention that holds itself up.
What to try
Let it run and the hawk share climbs or falls to the gold line — the ESS at p* = V/C — and sits there. Now press Invade with hawks: a burst of aggressive mutants floods in and the share jumps. But the population is now too hawkish, fights are everywhere, and the bottom panel shows it plainly — the dove payoff line is suddenly above the hawk line. So hawks earn below average and dwindle, and the mix slides back to p*. Try Invade with doves and the mirror happens: too many pushovers, hawks feast, hawks grow, the share climbs home. The deviation never wins.
Then move the cost of a fight. Raise C and fights get dangerous, so the stable share of hawks drops — a costlier world sustains more restraint, automatically. Lower C toward the value V and hawks pay so little that aggression always pays: the ESS marches to 100% hawk, a pure norm of escalation. The presets stage a balanced convention, a peaceable world where fights are ruinous, and a ruthless one where they're cheap — each a different self-enforcing normal, none of them chosen by anyone.
The mapping
Read Hawk and Dove as any pair of "push" versus "yield," and the ESS becomes the mathematics of unenforced custom. Manners and civility: in a courteous town, boorishness loses more than it gains, so it can't take over — but a little sharpness survives, because in a town of total pushovers, some assertiveness pays. Queuing, tipping, driving conventions, who apologizes first: each is a behaviour that's only worth doing because almost everyone else does it, and deviating quietly costs you. Bluff and restraint in conflict — the posturing that stops short of ruin — is the original case, and it's everywhere humans threaten without wanting to actually fight.
The deep point for human life is that a norm can be real, robust, and completely unpoliced. We reach for enforcers — laws, shame, authority — to explain why people behave, but many conventions need none: they persist because, given that others follow them, following them is simply the better move, and breaking them is its own quiet penalty. It also warns us: "stable" is not "good." The mix that resists invasion can be one nobody would have designed — a level of aggression, or waste, or inequality that holds firm precisely because no lone individual can profit by stepping out of it.
Read as life lessons
A custom can hold with no police at all, if breaking it simply pays less. The penalty isn't punishment — it's the worse outcome the deviant earns among people who don't deviate. Self-interest does the enforcing.
Any behaviour that thrives by being unusual can't take over — as it spreads it undermines its own advantage. That's why populations settle into mixes, not monocultures, and why the odd contrarian always survives.
An ESS resists change; it doesn't promise welfare. A wasteful or aggressive equilibrium can be rock-solid because no individual gains by leaving it. Robustness is not endorsement — some healthy-looking normals are traps.
In the wild
Ritualized fights, sex ratios near 50/50, alternative mating tactics, and territorial bluffing are read as ESSs — mixes no rare mutant can beat, which is why they're so widespread and so stable.
Evolutionary game theory models conventions, bargaining norms, and the spread of cooperation as strategies that persist by being un-invadable, not by being decreed — custom as equilibrium.
Replicator dynamics and ESS analysis tune multi-agent learning, routing, and evolutionary algorithms — engineering stable behaviour into populations of agents that no single defector can exploit.
The mapping, exactly
| Mathematics | Life |
|---|---|
| a strategy | A way of behaving in a recurring situation — escalate or yield, cooperate or defect, the custom you follow by default. |
| the payoff matrix | What each behaviour earns against each other — the real incentives of the situation, before anyone moralizes about them. |
| the frequency p | How common a behaviour is in your world — the prevailing normal you're born into and matched against. |
| replicator dynamics | Better-than-average behaviours spreading, worse-than-average ones fading — imitation, learning, or reproduction with no one in charge. |
| the ESS p* = V/C | The self-enforcing custom — the mix at which no rare deviation pays, so it holds without an enforcer. |
| failed invasion | The newcomer whose different behaviour earns less, gains no ground, and quietly conforms — the norm healing itself. |
The honest model
The Hawk–Dove payoffs are built live from your two sliders: Hawk-vs-Hawk earns (V−C)/2, Hawk-vs-Dove earns V, Dove-vs-Hawk earns 0, Dove-vs-Dove earns V/2. Given a hawk share p, the script computes the expected payoff to each strategy against the current mix, and evolves the share by the replicator equation ṗ = p(1−p)(W_hawk − W_dove) in small time-steps — nothing scripted, just this one line integrated.
That difference works out to W_hawk − W_dove = (V − pC)/2, so the share stops moving exactly at p* = V/C — the value the readout reports and the gold line marks — or at 100% hawk when fights are cheap enough that C ≤ V. The population grid recolours to the live share; the trajectory plots it over time; the bottom panel draws the two fitness curves crossing at p*. Invasions don't teleport the answer: they shove p off the ESS, and the same equation carries it back, so what you watch return is a computation, not an animation played to look right.
Where the metaphor tears
The model assumes everyone is equally likely to interact with everyone — a well-mixed population. Human norms live in structure: families, neighbourhoods, networks, reputations, repeat encounters. On a network a strategy that would fail against the average can survive in a cluster, and cooperation that "can't invade" globally invades locally. Structure changes which norms are stable, and the clean mix is only the well-mixed case.
Here V and C are dials we hold still. In life, institutions, technology, and the norms themselves move the payoffs: build courts and the cost of a "fight" changes; shift the culture and yielding starts to mean something new. When behaviour feeds back onto the incentives, the stable point wanders, and a snapshot ESS can quietly stop describing the world it was fitted to.
An ESS resists invasion; it does not maximize anyone's welfare, and a game can have several. Which self-enforcing custom a society lands in is often historical accident — drift, founder effects, who arrived first — not merit. Reading "it's stable" as "it's right," or as "it was inevitable," is exactly the error the mathematics should inoculate against.