What Is the Prisoner’s Dilemma and How Does It Work?

Search Fintrev

Search our guides, reviews and glossary.

The prisoner’s dilemma is a game theory model in which two players each do better for themselves by betraying the other, yet both end up worse off than if they had cooperated. Each player’s self-interested choice is individually rational, and the outcome it produces is collectively irrational. That gap between what is best for each person and what is best for the pair is what makes it a dilemma rather than a simple choice.

How the Prisoner’s Dilemma Works

The classic version runs like this. Two suspects are arrested for a robbery and questioned in separate rooms, with no way to communicate. The prosecutor has enough evidence to convict both on a minor charge but needs a confession to convict either one on the main charge. Each suspect is offered the same deal: testify against your partner and you go free.

Each suspect has two options: cooperate with the partner by staying silent, or defect by testifying. The sentences depend on both choices:

• Both stay silent: each is convicted on the lesser charge and serves 1 year (2 years combined).
• One testifies, the other stays silent: the one who testifies goes free; the silent partner serves 5 years (5 years combined).
• Both testify: each serves 3 years (6 years combined).

The same outcomes in a payoff matrix, with years in prison shown as (Suspect A, Suspect B). Lower numbers are better for the player.

Why both players defect

Look at the game from Suspect A’s side. If B stays silent, A can serve 1 year by also staying silent, or walk free by testifying. Testifying is better. If B testifies, A can serve 5 years by staying silent, or 3 years by testifying. Testifying is better again.

Whatever B does, A shortens her own sentence by testifying. In game theory terms, defecting is a dominant strategy: it beats the alternative in every possible case. B faces the identical calculation, so B testifies too. Both end up with 3 years.

The paradox is in the totals. The combined sentence is lowest (2 years) when both stay silent and highest (6 years) when both testify. Reasoning carefully about their own interests leads the two players to the worst joint outcome in the table.

The Nash equilibrium and why it is inefficient

The mutual-defection cell (3, 3) is the game’s Nash equilibrium: a set of choices from which no player can improve their own result by changing only their own move. If A switches to silence while B keeps testifying, A’s sentence rises from 3 to 5 years, so A has no reason to switch alone. The same holds for B.

Mutual cooperation (1, 1) is not an equilibrium, because each player could improve from 1 year to 0 by switching to testify. The cooperative outcome is better for both, but it is unstable without some way to make promises binding.

Economists describe the equilibrium as Pareto inefficient: another outcome (mutual silence) would make both players better off. Individual rationality does not guarantee an efficient result, so self-interest does not always add up to the best outcome for a group.

The Formal Structure of the Game

The prison story is only an illustration. Game theorists define a prisoner’s dilemma by the ordering of four payoffs, usually written as points where higher is better:

• T (temptation): the payoff for defecting while the other player cooperates
• R (reward): the payoff each player gets when both cooperate
• P (punishment): the payoff each player gets when both defect
• S (sucker’s payoff): the payoff for cooperating while the other player defects

A game is a prisoner’s dilemma when T > R > P > S. In the jail example, measured in years avoided rather than years served, the ranking holds: going free (T) beats 1 year (R), which beats 3 years (P), which beats 5 years (S).

For the repeated version, a second condition is usually added: 2R > T + S. It ensures that cooperating every round pays more in total than taking turns exploiting each other, which would otherwise change the character of the game.

A points-based example

A widely used set of teaching values is T = 5, R = 3, P = 1, S = 0.

Both conditions hold: 5 > 3 > 1 > 0, and 2 × 3 = 6 is greater than 5 + 0 = 5. Over 10 rounds, steady mutual cooperation earns each player 30 points and mutual defection earns each 10. A player who cooperates while the other defects every round collects 0 while the defector collects 50, and that fear of being exploited pushes both toward defection.

Origins of the Concept

Mathematicians Merrill Flood and Melvin Dresher developed the game in 1950 at the RAND Corporation, where game theory was studied for its military and strategic uses. Princeton mathematician Albert Tucker gave it the prison story and the name. John Nash had just published his equilibrium concept, and the prisoner’s dilemma became the best-known example of an equilibrium that is worse for everyone involved.

The Repeated Prisoner’s Dilemma

A single round always ends in defection for rational, self-interested players, but real economic relationships rarely last one round. Suppliers deal with the same buyers, competing firms set prices every quarter, and countries negotiate trade terms year after year. When the same players meet again and again, the game becomes an iterated or repeated prisoner’s dilemma, and the logic changes.

In a repeated game, today’s choice affects tomorrow’s treatment. A player who defects now gains T once but may face retaliation in every later round. If players care enough about future payoffs, the long-run value of continued cooperation outweighs the one-time gain from betrayal, and cooperation can become a stable outcome.

Tit for tat and other strategies

In the early 1980s, political scientist Robert Axelrod ran computer tournaments in which strategies submitted by researchers played the repeated game against each other. The winner was tit for tat, submitted by mathematician Anatol Rapoport. Its rules are simple:

1. Cooperate in the first round.
2. In every later round, copy whatever the other player did in the previous round.

Tit for tat is never the first to defect, punishes a defection immediately, and returns to cooperation as soon as the other player does. Axelrod’s analysis credited its success to being nice, retaliatory, forgiving and easy for opponents to read.

Tit for tat is not the best strategy in every setting; its results depend on which other strategies are in the population. It also struggles with mistakes: one accidental defection can lock two tit-for-tat players into an endless cycle of alternating retaliation. Variations address this:

• Generous tit for tat occasionally forgives a defection instead of retaliating, which breaks echo cycles.
• Grim trigger cooperates until the other player defects once, then defects forever. It deters betrayal strongly but cannot recover from an error.
• Win-stay, lose-shift (also called Pavlov) repeats its last move after a good payoff and switches after a bad one, which helps it recover from mistakes.

Why a known end is dangerous for cooperation

Cooperation in the repeated game depends on the shadow of the future. If both players know exactly when the game ends, that shadow disappears in the last round. With no future retaliation possible, the final round is effectively a one-shot game, so both defect.

The second-to-last round then has no future worth protecting either, so both defect there too. Reasoning backward this way, called backward induction, unravels cooperation all the way to the first move. Cooperation is easier to sustain when the end is uncertain or distant, which is why long-term business relationships behave differently from one-off transactions.

Real-World Examples

Cartels and output quotas

A cartel is a prisoner’s dilemma among producers. If every member restricts output, prices stay high and all earn more, but each can earn even more by quietly producing above its quota while the others hold back. If everyone reasons this way, output rises and prices fall, which is why cartel agreements are prone to cheating.

From society’s point of view, this dilemma is useful. Cartel members fail to cooperate, and consumers benefit from lower prices. Antitrust laws that ban binding collusion keep competitors locked in the dilemma deliberately.

Price wars and advertising

Two competing firms may both be better off holding prices steady, but each is tempted to undercut the other to win market share. If both cut, both earn lower margins with roughly the same market share. Advertising works the same way: when rivals mostly win customers from each other, heavy spending by both cancels out, leaving each with higher costs, and neither wants to cut back alone.

Arms races

The Cold War nuclear buildup is often described in these terms. Each country preferred to be armed whether the other disarmed or not, so both armed and bore enormous costs while being no safer relative to each other. Arms control treaties with inspection provisions try to make cooperation verifiable.

Trade policy

Two countries may both gain from open trade, but each may be tempted to impose tariffs to protect domestic industries while still exporting to the other. If both impose tariffs, both lose some of the gains from trade. Trade agreements with dispute mechanisms try to make the cooperative outcome enforceable.

How Cooperation Is Achieved

Every way out of the dilemma works by changing the payoffs or the information available to players.

Repetition. Turning a one-time interaction into a repeated one gives players a reason to protect their future dealings. Long-term contracts, customer relationships and memberships all create repetition.

Enforceable agreements. Contracts, laws and regulations add a penalty to defection. If the penalty is large enough that T minus the penalty falls below R, the game is no longer a prisoner’s dilemma.

Reputation. When others can observe how a player behaves, defecting against one partner damages trust with many. Credit histories, seller ratings and professional licensing all make reputation visible.

Communication and commitment. The classic game assumes the players cannot talk. Allowing communication does not solve the problem by itself, since promises can be broken, but combined with enforcement or repetition it makes cooperation easier.

Norms and preferences. Trust, reciprocity, guilt and concern for others change what players actually value, so people who feel bad about betraying a partner are not playing with the textbook payoffs.

Limitations of the Model

The prisoner’s dilemma is a simplification, and its predictions depend on assumptions that often fail in practice.

• Payoffs are assumed to be purely self-interested. Real people weigh fairness, loyalty and the welfare of others, which alters the ranking of outcomes.
• Behavior in experiments differs from the prediction. Laboratory studies of one-shot versions regularly record cooperation.
• Real situations are rarely two-by-two. Firms choose among many prices and quantities, not just “cooperate” or “defect,” and often face more than one rival.
• Information is rarely symmetric. Players may not know the other side’s payoffs, costs or intentions.
• Many situations are not prisoner’s dilemmas. A different payoff structure implies different solutions, so the label can mislead.

What the Prisoner’s Dilemma Is Not

Not the same as the Nash equilibrium

The Nash equilibrium is a general solution concept that applies to any game. The prisoner’s dilemma is one specific game, notable because its equilibrium is inefficient. Many games have equilibria that are perfectly efficient.

Not a zero-sum game

In a zero-sum game, one player’s gain is exactly the other’s loss. In the prisoner’s dilemma, the total payoff changes with the choices: both can gain together through cooperation or lose together through defection. That is precisely what gives cooperation its value.

Prisoner’s dilemma vs. tragedy of the commons

The tragedy of the commons describes a shared resource, such as a fishery, pasture or aquifer, that anyone can use and no one can be excluded from. Each user gains the full benefit of taking more but bears only a fraction of the cost of depletion, so the resource is overused. It is often treated as a many-player prisoner’s dilemma.

The difference is scale: the prisoner’s dilemma is usually framed as two players making a single choice, while the commons involves many users drawing on a resource over time. Economist Elinor Ostrom documented communities that manage shared resources successfully through local rules and monitoring, work that earned her the 2009 Nobel Prize in economics.

Prisoner’s dilemma vs. the free-rider problem

A free-rider problem arises when people can enjoy a public good without paying for it, so each has an incentive to let others contribute. It is closely related and often modeled as a multi-player prisoner’s dilemma, but the focus is on underfunding a shared benefit rather than mutual betrayal.

Prisoner’s dilemma vs. the stag hunt

In the stag hunt, two hunters can cooperate to catch a stag or each catch a hare alone. Mutual cooperation is the best outcome and, unlike in the prisoner’s dilemma, also a Nash equilibrium: neither hunter gains by switching away from it if the other cooperates. The problem is trust, not temptation. Bank depositors deciding whether to keep their money in resemble the stag hunt more than the prisoner’s dilemma, because everyone staying calm is stable as long as each believes the others will.

Prisoner’s dilemma vs. chicken

In the game of chicken, two drivers head toward each other and the one who swerves loses face. The worst outcome is mutual defection (a crash), which is worse than being the one who swerves. That changes the ranking to T > R > S > P, and players try to commit visibly to not swerving rather than simply defecting.

The Bottom Line

The prisoner's dilemma matters to anyone studying competition, pricing, cartels, arms control or trade policy, where self-interest can produce a result worse for everyone. Cooperation becomes more likely when players expect to meet again with no known end, face enforceable penalties for defecting, or care about visible reputations. The model simplifies real behavior, so check whether a situation actually has its payoff structure before applying the label.