Ultimatum Game
Proposer is given $100. They offer a split to Responder: x for me, 100−x for you, take it or leave it. If Responder accepts, they each get their share. If Responder rejects, both get nothing. Game theory says the rational Responder accepts any non-zero offer (1¢ > 0). Humans reject offers below ~30% — even at cost to themselves. The most studied game in behavioral economics.
when fairness contradicts game theory
The subgame-perfect equilibrium: Responder is rational, so they accept any offer ≥ $0.01 (something beats nothing). Proposer knows this, so they offer $0.01 and keep $99.99. This is what classical game theory predicts.
What actually happens: when humans play this game (in experiments across dozens of cultures since Güth-Schmittberger-Schwarze 1982), the mean offer is around 40-50% of the pot. Offers below 20% are rejected ~80% of the time. People burn money to punish unfair proposers.
The most extensive cross-cultural study (Henrich et al., 2005, in Nature) ran the game in 15 small-scale societies — from Machiguenga foragers in Peru to Hadza hunter-gatherers in Tanzania to Lamalera whale-hunters in Indonesia. Modal offers ranged from 25% to 58% depending on the culture's degree of cooperation in everyday life. The Machiguenga accepted nearly any offer; the Lamalera (who share whale meat communally) frequently rejected offers above 50% as "too generous, what's wrong with you."
The deep claim: humans evolved negative reciprocity as a stable social strategy. Punishing free-riders (even at cost to yourself) sustains cooperation in repeated interactions. The "irrational" rejection of low offers may be a single-shot expression of a multi-shot strategy that's optimal in the ancestral environment. The Nash equilibrium is wrong about humans not because we're confused but because we're optimized for a different game than the one being played.