The Prisoner's Dilemma

Picture two partners in crime, caught by the police and questioned in separate rooms. The police offer each of them a tempting deal. If you talk first, you walk free, and your silent partner gets 10 years. If you both stay loyal and keep quiet, there isn't enough evidence, so you each serve just 1 year on a minor charge.

But if you both betray each other and confess, you each get 5 years. Now think about what's going on in each prisoner's head. If the partner stays quiet and I confess, I go free (0 years). If the partner confesses and I confess too, I get 5 years instead of 10. Either way, confessing always looks like the better move.

So both prisoners pick the choice they think is smartest: confess. And they miss the best shared outcome, where both stay quiet and serve only 1 year. Instead, they each get a heavy 5-year sentence.

This dilemma shows up outside of prison every day. A classic example is two big tutoring academies fighting over tuition discounts or ad spending. If both cut back on ads, they save a huge amount of money and both profit. But each is tempted: if the rival cuts ads and we ramp ours up, we could scoop up all the students.

If the rival goes all out on ads and we sit still, we'd lose students and might have to close. So we reluctantly fire back with our own ads. In the end, both academies pour enormous sums into advertising and end up with the same number of students as before. Endless arms races between countries and cutthroat price wars between companies happen for the same reason.

Pollution works the same way. If every factory installs a wastewater treatment system, the river stays clean and everyone benefits. But if the others install one and you secretly dump waste, you save money. When everyone tries to save a buck, the river rots and everyone suffers.

In economics, a stable state that people reach by looking only at their own interests is called a Nash equilibrium. In the prisoner's dilemma, both players confessing is exactly this kind of equilibrium. As long as the other person doesn't change their choice, changing yours alone only makes things worse, so it's very hard to break out of it.

The paradox is that individual rationality doesn't add up to group rationality. A better result for both (both stay quiet, 1 year each) clearly exists. But even if the two promise each other to cooperate, with no way to enforce the promise they stay stuck in an equilibrium where both lose compared with cooperating.

So is there no way out of this trap? Things change if the game isn't played just once but is a relationship that repeats without end. If you punish a betrayer in the next round and reward cooperation with cooperation (a strategy called tit for tat), cooperation can last. In political scientist Robert Axelrod's computer tournaments, this simple strategy scored the highest. But it only works if the players expect to keep meeting in the future.