Each village's people share one temperament: a balance between their own gain and the right thing — what would be best if everyone did as they do. You set the balance. Then newcomers with other balances arrive. A newcomer that earns more than your villagers takes over; one that earns less dies out.
Five societies that differ in how often a newcomer meets its own kind. In the first three you find the unbeatable balance; in the last two you predict it in one try.
Built from Alger & Weibull (2013), Econometrica. About this game
Drag to set your villagers' balance. Then test it: eleven kinds of newcomer arrive in turn, with balances 0, 0.1, … up to 1.
The bar is the village: your villagers on one side, the newcomers on the other. Newcomers who earn more per meeting grow until they take over; those who earn less shrink to nothing.
One row per test, ruler from 0 (only own gain) to 1 (only the right thing). Brass wedge: your village. ✕: that newcomer took over. Small ring: it died out or blended in.
Each figure is one society. Across: how often a newcomer meets its own kind. Up: the balance you ended on (filled if it held). The dashed line is where the two are equal.
Play again: find each balance in fewer tests, or start from a different society.
Your aim: find the balance no newcomer can beat.
People in a village are paired up to deal with each other — sharing a chore, splitting a resource, doing business. Each villager carries a temperament, a balance between two goals: their own gain, and the right thing, meaning what would be best for them if the other person did exactly the same. Balance 0 cares only about own gain (economists call this creature homo oeconomicus); balance 1 only about the right thing (homo kantiensis, after Kant); anything between is homo moralis. Nobody can see anyone else's balance.
You set the balance for the whole village and test it. Eleven kinds of newcomer arrive in turn, with balances 0, 0.1, … up to 1. Everyone acts on their own balance and earns what the deal pays. A newcomer kind that earns more per meeting than your villagers spreads and takes over; one that earns less dies out; one that behaves exactly like your villagers blends in.
In the first three societies, test as often as you like until nothing beats you; the drawer shows which newcomers won and where they sit. In the last two, you get one test: predict the unbeatable balance from what you found.
They differ in how often a newcomer meets its own kind: strangers in a city (practically never), siblings (half the time), half-siblings (a quarter), classmates in business (about one in five). Three deals are played: a shared chore (how much effort to put in), a quarrel over a resource (grab or share), and a dilemma (cooperate or cheat).
After each test: how many of the eleven newcomers beat your village. Zero means you found it. Each society's verdict says how many tests it took, or whether your prediction held.
Which temperaments survive when people cannot read each other's minds. Play before reading the answer; the end screen and the About page explain it.
The game runs the paper's own model. It demonstrates how that model behaves: which temperaments resist invasion in five matching environments and three deals. It assumes the paper's setup (pairwise matching, private temperaments, rare newcomers), its homo moralis family of temperaments, and payoff numbers we chose within the paper's conditions. The paper establishes a theorem about this model — that the temperament whose moral weight equals the chance of meeting your own kind cannot be invaded, and that temperaments behaving differently can be — not that real people are like this. The paper suggests laboratory tests; none are shown here.
A model. People are paired to play a symmetric game. What each person wants is private. Pairing may be assortative: one number, which the paper calls the index of assortativity, is the chance that a rare newcomer is paired with one of its own kind. Homo moralis weighs her own payoff against the payoff she would get if the other person did what she does, Kant's "act as if your rule were a universal law". The paper's main result: the temperament whose weight on the second goal equals the index of assortativity — homo hamiltonensis — cannot be invaded, and any temperament that behaves differently can be (Theorem 1).
A counterintuitive result. Earlier work found that when preferences are private, evolution selects pure self-interest; and a natural guess is that more goodness is always safer. In this model, the survivor is a precise blend: exactly as moral as the chance of meeting your own kind. Too little loses; so does too much. And the blend does not depend on the deal being played: what is at stake changes the behaviour, not the balance. Both follow from the model, not from the game's scoring: the game counts which newcomers out-earn your villagers, which is the paper's own test of invasion.
You set the village's balance and test it against eleven newcomer kinds, balances 0, 0.1, … up to 1. For each, the game computes how your villagers behave, how the newcomers behave given that they meet their own kind with the society's chance, and what each earns per meeting — in the limit where newcomers are very rare, as in the paper. A newcomer that earns more takes over; one that earns less dies out; one that behaves exactly like your villagers blends in.
1. Strangers in a city, chance zero (Section 2.1), shared chore (the public-goods game of Example 1). 2. Siblings, chance one half (Section 5.1), same chore. 3. Siblings, a quarrel over a resource (the hawk–dove game, Section 4). 4. Half-siblings, chance one quarter (footnote 24), same quarrel. 5. Classmates in business partnerships, chance (40 − 1)/(100 − 1) × ½ ≈ 0.20 (Section 5.2), a prisoners' dilemma with the payoffs of the paper's Fig. 1. The first three are for finding the unbeatable balance; the last two ask you to predict it in one test.
The game's behaviour matches the paper's formulas (Fig. 1, the hawk–dove solution, Example 1's condition) to within 10⁻¹². In all five societies, the predicted balance is beaten by none of 201 newcomer kinds; every other balance that behaves differently is beaten by at least one of the eleven, and the winners always sit on the side of the unbeatable balance. The result survives sixty random alternative deals and chances, so it does not depend on our payoff numbers. In the classmates' dilemma, every balance up to one quarter cheats and survives, as the paper notes after Fig. 1.
Asymmetric games and role assignment; helping others and ultimatum bargaining; repeated play; morality versus altruism; morality conditional on group identity; newcomer kinds that are not homo moralis (the paper's proof uses committed types); multiple equilibria.
Newcomers come from an eleven-point grid of balances. The bar shows newcomers spreading or shrinking at a stylised speed; only the direction is the model's. The chore's benefit and cost curves and the quarrel's stakes are our choices within the paper's conditions. The balance moves in steps of 0.05. The count of tests and the predictions are the game's, not the paper's.
See the author note in the repository. This game has had the builder's review only; it has not been tested with real players.