Two suspects are arrested. Police lack evidence to convict either, so they offer each a deal separately: testify against your partner and go free, while your partner serves ten years. If both testify, both get five years. If both stay silent, both get only one year on a lesser charge.

If the prisoners could talk, they'd obviously both stay silent. But they can't. And here's where it gets strange: the rationally selfish move, regardless of what the other person does, is to betray. If your partner stays silent, betrayal means you walk free instead of doing a year. If your partner betrays, betrayal means five years instead of ten. Betray always wins.

So two rational, self-interested prisoners both betray — and both serve five years. They would have each gotten one year by cooperating. This is the Prisoner's Dilemma, and mathematician John Nash proved that in one-shot games between strangers, mutual betrayal is the only stable outcome.

The dilemma shows up everywhere. Nuclear powers spending on arms neither side wants. Airlines undercutting each other's prices until everyone's broke. Climate treaties where every country benefits if all cut emissions, but each benefits more by cheating.

The escape hatch is repetition. When people play the game hundreds of times, strategies like 'tit for tat' — cooperate first, then copy what the other player did last — consistently win. Trust builds through repeated interaction, not through appeals to reason. Which is why long relationships work and one-time transactions often don't.