The prisoner’s dilemma is well-known:

  Bob cooperates Bob defects
Alice cooperates A: -1, B: -1 A: -3, B: 0
Alice defects A: 0, B: -3 A: -2, B: -2

Here’s the setup: Alice and Bob are suspects being interrogated separately. They can snitch on the other (defect) or not (cooperate). If both snitch, they get two years in prison (the lower right corner). If neither snitches, they get a year each, presumably on a lesser charge (upper left). If one snitches and the other doesn’t, the defector goes free while the sucker gets three years in prison (the other two corners).

Given these conditions, if the prisoners are rational and care only about their own freedom, you’d expect both to snitch. This is because no matter what the other does, one gains more by defecting. You can verify this: if Bob cooperates, Alice should defect (going free (0) is sweeter than a year in prison (-1)). If Bob defects, Alice should defect (snitching (-2) is better than getting suckered (-3)). Bob reasons in an exactly symmetric way. Outcome: both defect and get two years in prison each. Tragedy. They could’ve agreed to cooperate and gotten just a year instead.

In real life, of course, we don’t care much about prisoners determining their fates in contrived settings. The prisoner’s dilemma is a metaphor for scenarios we do care about. The problem is that it is often hard to tell whether a real-world scenario is a prisoner’s dilemma or this other game called a stag hunt:

  Bob cooperates Bob defects
Alice cooperates A: 2, B: 2 A: 0, B: 1
Alice defects A: 1, B: 0 A: 1, B: 1

Here’s the setup for the stag hunt: Alice and Bob can cooperate to hunt a stag (a significant undertaking), earning 2 lbs of meat each (upper left). Or they can each hunt a hare alone, earning 1 lb of meat each (lower right). But if either tries to hunt a stag alone while the other hunts a hare, they’ll fail and the other guy will earn 1 lb of meat from the hare (the other two corners).

In a stag hunt, there is no obvious strategy X like “No matter what the other guy does, I should do X.” If Bob cooperates, Alice should, too (2 > 1). If Bob defects, Alice should also defect (1 > 0).

So, prisoner’s dilemmas and stag hunts are markedly different. In the real world, how do you tell one from the other? Here’s a quick heuristic: games where individual contributions are critical are stag hunts. For example, founding a startup is a stag hunt. It’s a sufficiently hard problem that if one of the co-founders slacks off, the thing just won’t get done. But if a co-founder sees the others cooperating, they have an incentive to cooperate themselves — if everyone’s working hard, there’s a good chance they’ll succeed. Similarly, small teams of scientists pursuing a research question are playing a stag hunt.

On the other hand, lots of diffused social interactions are n-person prisoner’s dilemmas. They’re called, of course, tragedies of the commons. A classic example is the overfishing story. If your fellow fisherman have a pact to restrict their fishing (so that enough fish remain to mate and yield a harvest next year), you should overfish. Your overfishing is unlikely to shrink next year’s pie too much, while you get a larger slice this year. If there is no such pact, you again overfish (otherwise you’re getting suckered). Another classic example is the global warming story: if there’s an international treaty to limit carbon emissions, you over-emit. (Burning fossil fuels without the costs of climate change is great!) If there isn’t a treaty, you again over-emit — the world will heat up whether or not you build those factories, so build them.

In general, projects where the individual contributors are non-critical are prisoner’s dilemmas. But these projects need not be society-scale. Small teams doing projects which are not truly hard play prisoner’s dilemmas. Final projects in college are an example. Usually, one or two teammates can pull off an A-grade project just fine — so it should not be surprising how common free-riding is in college projects.