The last two lessons taught you to find the still point of a game — the cell nobody can profitably leave alone. But finding the resting point and liking it are two very different things, and this lesson is where that gap opens into a chasm. We’re going to meet a game where both players reason flawlessly, each plays their provably best move, and because of that they both land somewhere worse than a pair of cooperating fools would have. Nobody is stupid. Nobody is evil. The rational move and the good outcome are simply different moves. Once you feel this, you will never again confuse “stable” with “good” — and you’ll spot the same tragedy hiding under price wars, arms races, overfishing, doping, and climate talks. This is the prisoner’s dilemma, the most important game in the social sciences.
The story
Two partners in crime — call them You and your Partner — get arrested. The police have enough evidence to convict you both of a minor charge, but they want the big one, and for that they need a confession. So they lock you in separate rooms, cut off all communication, and offer each of you the exact same deal. Your move is either to stay silent (this is cooperating — with your partner, not the police) or to rat (this is defecting — selling your partner out for a lighter sentence).
Here’s how the deal shakes out:
- You both stay silent (both cooperate). They can only pin the minor charge on you, so you each do light time. Good outcome for the pair.
- You rat while your partner stays silent (you defect, they cooperate). You get the sweetheart deal — you walk with a reward for testifying — and your partner takes the full fall. Best possible for you, worst possible for them.
- Your partner rats while you stay silent (you cooperate, they defect). The mirror image: they walk, you eat the whole sentence. This is the outcome that keeps you up at night — being the sucker who cooperated with a rat.
- You both rat (both defect). The prosecutor doesn’t need either of you anymore, so neither gets the sweetheart deal, and you both do serious time. Worse for both than if you’d both kept quiet.
To reason about it, we strip the years off and turn everything into points where higher is better. Cooperating together earns each of you 3. Defecting on a cooperator earns the rat 5 (the temptation) and leaves the sucker with 0. Mutual ratting earns each of you a measly 1 (the punishment). Confirm the ranking in your head — 5 beats 3 beats 1 beats 0 — then play with the live matrix below.
Payoff matrix
The prisoner's dilemma (higher points = better)
Each cell is (Your points, Partner's points). Ringed = that player's best response to the rival's choice; a cell ringed for BOTH gets the NE badge. Read it, confirm the badge sits on (Defect, Defect) — then start nudging.
| Partner | ||
|---|---|---|
| Cooperate | Defect | |
| YouCooperate | 33 | 05 |
| Defect | 50 | NE11 |
A ringed payoff is that player’s best response to the rival’s choice. A cell where both are ringed is a Nash equilibrium.
What the matrix says
You — dominant strategy: Defect
Partner — dominant strategy: Defect
Nash equilibrium (pure): (Defect, Defect)
Why each of you defects
Put yourself in your own locked room and reason it through — the exact column-by-column dominance check from lesson 2. You don’t know what your partner will do, so consider both of their possible moves in turn.
- Suppose your partner cooperates (stays silent). If you cooperate too, you score 3. If you defect, you score 5 (you walk, they take the fall). 5 beats 3 — defecting is better.
- Suppose your partner defects (rats). If you cooperate, you’re the sucker and score 0. If you defect, you score 1 (you both do time, but at least you’re not the lone patsy). 1 beats 0 — defecting is better.
Look at what just happened: defecting won in both cases. It didn’t matter what your partner chose — defecting handed you a higher score either way. A move that beats its alternative no matter what the rival does is a dominant strategy, and here defecting is one. So you defect. But the game is perfectly symmetric, so your partner sits in their room running the identical arithmetic and reaches the identical conclusion: defect. Two flawless chains of reasoning, both ending at the same move, and you march together into (Defect, Defect) = (1, 1).
That cell is the unique Nash equilibrium — the only cell where neither of you can improve by switching alone. Try it: from (1, 1), if you unilaterally switch to Cooperate while your partner keeps defecting, you drop from 1 to 0. Strictly worse. Same for your partner. Nobody can profitably leave, so nobody leaves. The trap has closed.
The gut-punch
Now look at the cell you both walked away from. (Cooperate, Cooperate) = (3, 3). Three points each — three times the one point you each ended up with. Both of you strictly prefer (3, 3) to (1, 1). You both knew it going in. And you still couldn’t get there.
This is the moment the whole course pivots on, so let’s say it plainly. (Defect, Defect) is the only stable outcome, and it is not the good one. Stability and goodness are different properties of a cell, and this game rips them apart to show you they were never the same thing. A Nash equilibrium answers the question “can anyone improve by moving alone?” It does not answer “is this good for the group?” — that’s a separate, often heartbreaking question with a different answer.
Reconnect it to the alone clause you learned in lesson 1. Why isn’t the lovely (3, 3) an equilibrium? Because from (3, 3), either one of you alone is tempted to defect: you’d jump from 3 to 5 by ratting while your partner stays loyal. That single unilateral temptation is enough to disqualify (3, 3) as a resting point — it fails the “no unilateral regret” test even though it’s better for everyone. The cooperative cell is good but unstable; the defection cell is stable but bad. The dilemma is precisely the clash between those two.
The exact shape of the tragedy
The mutual-defection cell (1, 1) is the Nash equilibrium: neither player can improve by switching alone (cooperating while your partner defects drops you from 1 to 0). The mutual-cooperation cell (3, 3) is better for both — but it is not an equilibrium, because from it each player can do even better by defecting (jumping 3 to 5). The only stable outcome is the one both players wish they could escape together. Individually rational, collectively self-defeating.
Mutual cooperation (3, 3) gives both players more than mutual defection (1, 1). Nash equilibrium is supposed to be where rational play 'lands' — so why doesn't the Nash concept just move both players to the better (3, 3) cell?
Pareto efficiency, defined
We need a precise word for “better for everyone,” because “good” is too vague to build on. That word is Pareto efficiency.
An outcome is Pareto-efficient if there’s no way to make one player better off without making another player worse off — you’ve squeezed out all the free wins. Its opposite is a Pareto-inefficient outcome, where a change exists that helps someone and hurts nobody: pure waste left on the table. And we say outcome A Pareto-dominates outcome B when A makes at least one player better off and nobody worse off — a strict, uncontroversial improvement that no rational player could object to.
Now apply it to the dilemma. Does (Cooperate, Cooperate) = (3, 3) Pareto-dominate (Defect, Defect) = (1, 1)? Move from (1, 1) to (3, 3): You go from 1 to 3 (better), and your Partner goes from 1 to 3 (also better). Both improve, neither is harmed. So yes — (3, 3) Pareto-dominates (1, 1), which means the Nash equilibrium (1, 1) is Pareto-inefficient. There was a strictly better outcome sitting right there, available to both, and rational play walked straight past it.
Here is the sentence to carve into your brain: a Nash equilibrium need not be Pareto-efficient. The equilibrium is where the game lands; Pareto efficiency is whether the landing wasted anything. The prisoner’s dilemma is the cleanest proof imaginable that these two properties can come apart completely — its unique equilibrium is Pareto-dominated by an outcome both players preferred.
Efficient doesn't mean fair — and stable doesn't mean either
Pareto efficiency is a low bar: it only asks whether any free improvement remains, not whether the outcome is fair. The temptation cell (5, 0) is actually Pareto-efficient — you can’t make the sucker better off without taking from the rat — yet it’s grotesquely unfair and not an equilibrium either. So a cell can be efficient-but-unfair, stable-but-inefficient, or good-but-unstable. “Stable,” “efficient,” and “fair” are three separate questions. The prisoner’s dilemma is famous because it drives a wedge between the first two.
The dilemma is everywhere
The prison is just a vivid costume. The structure — two parties, each with a dominant incentive to defect, whose mutual defection leaves them both worse off than mutual cooperation would — is one of the most common shapes in the entire social world. Once you know the skeleton, you see it wearing new outfits constantly. In every row below, defecting is individually tempting, and universal defection is a shared disaster everyone would have paid to avoid.
| Situation | ”Cooperate” (the 3-3 move) | “Defect” (the tempting 5-0 move) | Why mutual defection lands at (1, 1) — worse for both |
|---|---|---|---|
| Two firms setting prices | Hold prices high | Undercut the rival to steal share | Both undercut, a price war guts everyone’s margins to the bone |
| Arms race between nations | Disarm / don’t build | Build more weapons | Both arm, vast money burned, relative power unchanged, everyone poorer and no safer |
| Shared fishery | Fish modestly | Catch all you can today | Both overfish, the stock collapses, and there’s nothing left for anyone |
| Doping in sport | Stay clean | Take the banned drug | Both dope, the rankings end up identical, but both wreck their health |
| Advertising spend | Modest ad budget | Blitz the airwaves to out-shout them | Both blitz, market shares barely move, but both bleed cash on ads |
| Climate / emissions | Cut emissions | Keep emitting freely | All keep emitting, everyone shares the climate damage nobody wanted |
Every row is the same trap. Defecting grabs a private, immediate gain — steal share, feel safe, catch more fish, win the race, out-shout the rival, skip the costly cutback — and mutual defection is a collective catastrophe both sides would have signed a treaty to avoid. That shared skeleton is why the prisoner’s dilemma is sometimes called the engine of the tragedy of the commons: a shared pasture overgrazed because each herder gains from one more cow while everyone splits the cost of the ruined field is just a many-player prisoner’s dilemma. “Graze modestly” is cooperate, “add another cow” is defect, and universal defection destroys the common resource. (There’s a whole course on that model nearby — same engine, bigger stage.)
Each situation has a 'cooperate' and a 'defect' move. Which ones are genuine prisoner's dilemmas (defecting is individually dominant AND mutual defection is worse for both than mutual cooperation) versus merely bad outcomes that lack that exact structure?
Place each item in the right group.
- Two drivers deciding which side of the road to drive on, both just wanting to match
- A single hiker deciding whether to pack an umbrella for a solo trip
- Rival airlines deciding whether to launch a fare war
- Fishers on a shared lake each choosing how hard to fish
- Two nations each deciding whether to keep building missiles
- Teammates each deciding whether to slack off on a shared-grade project
- Two startups hoping to pick the SAME connector standard so their gadgets interoperate
The one crack in the trap
If the numbers stay exactly as written, there is no clever argument that escapes the dilemma. Swearing an oath beforehand doesn’t help — cheap talk changes no payoffs, so from the cooperative cell each of you still gains by breaking the promise. Being smarter doesn’t help either; smarter players just defect faster. So what actually cracks it?
The single most powerful crack that needs no change to the payoffs at all is repetition — and the reputation it creates. If You and your Partner play this game once and never meet again, betrayal is free: there’s no future in which it can be punished. But if you meet again and again, today’s defection invites tomorrow’s revenge, and the shadow of the future can make cooperation the rational move after all. Reputation is the same lever aimed outward: when others watch and remember, a defector pays a long-run cost in every deal they never get offered again.
This is only a teaser — the full theory of repeated games, tit-for-tat, and the famous computer tournament lives in the game-theory-basics course, and the broader art of deliberately changing the game (contracts, enforcers, quotas, merging the deciders) gets its own treatment in lesson 06 of this course. For now, just hold the headline: you don’t beat a prisoner’s dilemma by being cleverer inside it — you beat it by changing the game around it.
Match each idea to its precise role in the dilemma.
Pick a term, then click its definition.
Capture the whole lesson in one line.
Pick the right option for each blank, then check.
In a prisoner's dilemma, is a dominant strategy for each player, so the unique Nash equilibrium is mutual defection at (1, 1) — an outcome that is but , because mutual cooperation at (3, 3) would leave both players strictly better off.
The classic misreading
The most common mistake in all of game theory is hearing “Nash equilibrium” and quietly translating it to “the good outcome,” “the fair outcome,” or “the outcome everyone agreed to.” The prisoner’s dilemma exists to kill that instinct. Its equilibrium is none of those things — it’s the outcome both players wish, with all their hearts, they could escape together, and can’t, precisely because it’s the only stable one. Equilibrium means nobody can move alone, not nobody wishes they could.
When to use it
Reach for the prisoner’s dilemma the instant you notice individual incentives pointing one way and collective interest pointing the other. The diagnostic is two quick questions: Is defecting dominant for each party? and Is mutual cooperation better for everyone but unstable? If both are yes, you’ve found the trap — and you now know the fix is never to moralize or to be cleverer inside the game. The fix is structural: change the payoffs. Introduce repetition so betrayal has a future cost, install an enforcer who punishes defection, build reputation so watchers remember, or merge the deciders so there’s no one left to defect against. Every real dilemma that humans actually solve gets solved by re-engineering the game, not by out-thinking it from inside.
That’s the thread we pull hard in lesson 06, “Changing the Game.” But first, a twist: so far every game we’ve met has had a single clean equilibrium. What happens when a game has several stable resting points, all equally valid, and the players have to somehow guess the same one? That’s the puzzle of coordination — and the surprising human trick that solves it — in the next lesson, “Many Equilibria: Coordination and Focal Points.” See you there.