Iterated Prisoner's Dilemma¶

The prisoner's-dilemma stage game (payoffs T > R > P > S) played over many observed rounds, where a high enough continuation probability lets the threat of future punishment deter present defection — turning the one-shot game's unavoidable mutual defection into sustainable cooperation.

Version: v2 · 2026-09-19 · History
Domain-specific #: 580
Domain group: Social Sciences
Origin domain: Economics & Finance
Subdomain: Game Theory and Strategic Interaction → Economics & Finance

Core Idea¶

The iterated prisoner's dilemma is the canonical repeated game: two players face the prisoner's-dilemma stage game (payoffs T > R > P > S) across many rounds, each observing the other's prior choices. The one-shot game's unique equilibrium is Pareto-inferior mutual defection; the iterated structure dissolves this when the interaction repeats with a high enough continuation probability δ, because future punishment deters present defection. The folk theorem shows any individually rational payoff is then sustainable, and Axelrod's tournaments found Tit-for-Tat robust.

How would you explain it like I'm…

Share or Grab, Again

Two friends play a game many times, and each time they can share or be selfish. Being selfish wins a bit more once, but if both are always selfish, both do badly. Because they meet again and again, each friend can remember what the other did. That makes sharing worth it, since being selfish today can get you treated badly tomorrow.

The Many-Rounds Trust Game

In the prisoner's dilemma, two players each choose to cooperate or to cheat. Cheating against a cooperator pays the most, but if both cheat, both end up worse off than if both had cooperated. Played only once, cheating is the 'logical' choice, so both lose out. In the iterated version they play many rounds and can see what the other did before. Now cooperation can last, because cheating can be punished in later rounds. In famous computer contests, a simple strategy called Tit-for-Tat did very well: cooperate first, then copy whatever the other player did last time.

Repeated Prisoner's Dilemma

The iterated prisoner's dilemma repeats a stage game in which each player cooperates (C) or defects (D). The payoffs satisfy T > R > P > S and 2R > T + S, where T is the temptation to defect, R the reward for mutual cooperation, P the punishment for mutual defection, and S the sucker's payoff. In a single round, mutual defection is the only Nash equilibrium, even though both would prefer mutual cooperation. When the game repeats with a high enough chance of continuing, the threat of future punishment can sustain cooperation. In Axelrod's computer tournaments, Rapoport's Tit-for-Tat won repeatedly. Successful strategies were nice (never defect first), retaliatory, forgiving, and clear. Later work found other strong strategies, like Generous Tit-for-Tat and Win-Stay-Lose-Shift. The model is used to study cooperation in biology, international relations, business collusion, and computer trust systems.

The IPD is the canonical repeated game. The stage game has payoffs T > R > P > S with 2R > T + S, and players condition their actions on the observed history. The one-shot game's unique Nash equilibrium is mutual defection, which is Pareto-inferior to mutual cooperation. With indefinite repetition and a sufficiently high continuation probability δ, credible threats of future punishment make cooperative equilibria sustainable. The folk theorem generalizes this: with patient players in an infinitely repeated game, any individually rational payoff vector, including mutual R, can be supported as a subgame-perfect equilibrium. Axelrod's tournaments, won repeatedly by Tit-for-Tat, suggested that successful strategies are nice, retaliatory, forgiving, and clear. Evolutionary analyses by Nowak, Sigmund, and May showed strategy populations cycling under selection, with Generous TFT and Win-Stay-Lose-Shift performing well. Press and Dyson's zero-determinant strategies can unilaterally set the ratio of payoffs regardless of the partner's strategy. The IPD models cooperation without enforceable contracts, in contexts such as reciprocal altruism, arms control, tacit collusion, and reputation systems.

Scope of Application¶

Because it is a paradigmatic model, it applies wherever its precondition holds: agents interacting repeatedly, observing past moves, under high δ and the stage-game payoff ordering.

Game theory of repeated interaction — the formal home and teaching vehicle.
Evolutionary biology — reciprocal altruism, cleaner-fish interactions, microbial public-goods games.
International relations — arms control and deterrence between repeatedly-interacting states.
Industrial organization — tacit oligopoly collusion held by the threat of price wars.
Multi-agent computer science — reputation, trust, and peer-to-peer protocols.

Clarity¶

The IPD makes visible that the same agents with the same payoffs can lock into mutual defection or sustain cooperation purely by changing the repetition structure. It separates agent preferences, stage-game payoffs, and temporal-relational structure, promoting interaction design to a first-class lever. It sharpens the indefinite-versus-finite contrast (a foreseeable endgame is a specific killer of cooperation) and, via Axelrod, makes legible which strategies are robust: nice, retaliatory, forgiving, clear.

Manages Complexity¶

The strategy space of a repeated game is intractable — a strategy is a function from the whole history to the next move. The IPD compresses this: the folk theorem characterizes the sustainable-payoff set as a region rather than a list, and the Axelrod tournaments compress the exponential space to a handful of short-memory rules and four legible properties. Deeper still, it isolates the few relational parameters (δ, horizon, observability, reputation) off which every qualitative result reads.

Abstract Reasoning¶

The IPD holds preferences and payoffs fixed and treats relational structure as the deciding variable. It licenses diagnostic prediction from the repetition dials, endgame-unraveling reasoning by backward induction, characterizing the equilibrium set (via the folk theorem) rather than enumerating it, strategy-class evaluation by the four legible properties, interventionist design keyed to the same dials, and transfer-by-shared-structure treating the model as a template instantiated by any setting's structural facts.

Knowledge Transfer¶

As a paradigmatic model, the IPD transfers as mechanism wherever an interaction genuinely carries the stage-game inequality plus a repetition profile — biology, international relations, industrial organization, multi-agent CS all inherit the whole analysis as genuine co-instances, not metaphors. Beyond the model itself the honest read is shared-abstract-mechanism: each domain fits the same game-theoretic substrate, so the most-exported lesson ("repetition can sustain cooperation") is really a claim about the parent primes Social Dilemma, Cooperation, Reciprocity, Iteration, and Game-Theoretic Strategy. Those carry the cross-domain weight; the folk theorem, Axelrod results, and named strategies stay home.

Relationships to Other Abstractions¶

Current abstraction Iterated Prisoner's Dilemma Domain-specific

Parents (3) — outgoing typed relations

Iterated Prisoner's Dilemma includes Iteration Prime

The game contains repeated application of the same stage interaction with prior-round output carried into the next strategy decision.
Iterated Prisoner's Dilemma includes Social Dilemma Prime

The iterated prisoner's dilemma contains the one-shot social-dilemma payoff matrix as the stage game repeated in every round.
Iterated Prisoner's Dilemma is a decomposition of Shadow Of The Future Prime

Stripping the named matrix leaves the horizon-times-observability mechanism by which discounted future punishment makes present cooperation self-interested.

Children (1) — abstractions linked here by typed relations

Driver Safety Arms Race Domain-specific is a kind of Iterated Prisoner's Dilemma

Driver Safety Arms Race is a domain-specific kind of Iterated Prisoner's Dilemma: The driver-safety arms race is a prisoner's-dilemma dynamic in which individually protective vehicle enlargement makes the fleet progressively larger and collectively more dangerous.

Hierarchy paths (4) — routes to 4 parentless roots

Iterated Prisoner's Dilemma → includes Iteration

Show alternative paths (3)

Neighborhood in Abstraction Space¶

Iterated Prisoner's Dilemma sits in a crowded region of the domain-specific corpus (30^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Game-Theoretic Models & Paradoxes (38 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25