Prisoner's Dilemma

Prisoner's Dilemma is a foundational non-cooperative model in game theory demonstrating why two rational decision-makers may fail to cooperate even when mutual cooperation yields higher joint and individual payoffs. Formulated in 1950 by Merrill Flood and Melvin Dresher at the RAND Corporation and framed in terms of prison sentences by Albert W. Tucker, the dilemma illustrates the tension between individual rationality and collective welfare. Because each player possesses a dominant strategy to defect regardless of the opponent's action, the game converges on a unique Nash equilibrium that is suboptimal under Pareto efficiency.

Payoff Matrix and Preference Ordering

In the symmetric two-player strategic form, players A and B simultaneously choose from the action set S_i = {C, D}, where C denotes cooperation and D denotes defection. The payoff function u_i(s_A, s_B) maps strategy profiles to real-valued utilities governed by four structural parameters:

• Temptation (T): The payoff to a unilateral defector when the partner cooperates, u_A(D, C) = T.
• Reward (R): The payoff to each player under mutual cooperation, u_A(C, C) = R.
• Punishment (P): The payoff to each player under mutual defection, u_A(D, D) = P.
• Sucker's Payoff (S): The payoff to a unilateral cooperator exploited by a defector, u_A(C, D) = S.

For a strategic interaction to constitute a canonical Prisoner's Dilemma, the parameters must satisfy the ordinal inequality:

In repeated settings, an additional constraint prevents players from achieving higher payoffs by alternating between (C, D) and (D, C) across consecutive periods than by maintaining steady cooperation:

Dominance and Equilibrium Inefficiency

A strategy s_i^* is dominant for player i if it yields a larger payoff than any alternative strategy for every possible action chosen by the opponent j. Evaluating player A's best response:

1. If player B chooses C, player A compares u_A(D, C) = T against u_A(C, C) = R. Since T > R, defecting is optimal.
2. If player B chooses D, player A compares u_A(D, D) = P against u_A(C, D) = S. Since P > S, defecting remains optimal.

By symmetry, D is also dominant for player B. Elimination of dominated strategies isolates (D, D) as the sole Nash equilibrium, generating payoffs (P, P). However, because R > P for both participants, the profile (C, C) Pareto-dominates (D, D). In the continuous donation game formulation, each player chooses whether to incur a personal cost c > 0 to confer a benefit b > c upon the counterpart. Here, T = b, R = b − c, P = 0, and S = −c, satisfying b > b − c > 0 > −c for all b > c > 0.

Backward Induction in Finitely Repeated Horizons

When the Prisoner's Dilemma is repeated over a known, finite number of rounds N < ∞ with common knowledge of rationality, backward induction unravels cooperation. In the terminal period t = N, no future retaliation is possible, reducing the stage game to a one-shot interaction in which both players play D. Given that period N play is invariant to history, period N − 1 effectively becomes the final strategic decision point, inducing mutual defection at t = N − 1. Iterating this logic back to t = 1 establishes that defecting in every round is the unique subgame-perfect Nash equilibrium (SPNE).

Folk Theorems and Discount Factors in Infinite Horizons

When the game is repeated infinitely—or with an indefinite horizon where the continuation probability after each stage is γ ∈ (0, 1) and the intertemporal discount factor is δ ∈ (0, 1)—cooperative outcomes can be sustained as subgame-perfect equilibria. Under a Grim Trigger strategy, a player begins by playing C and continues cooperating as long as both players have cooperated in all prior periods, switching permanently to D following any single defection. The discounted present value V_C of perpetual cooperation is:

If a player deviates to D in period t = 0, they capture the immediate temptation payoff T in the current round but trigger perpetual punishment P from t = 1 onward:

Cooperation is incentive-compatible (V_C ≥ V_D) if and only if the discount factor exceeds the threshold δ^*:

In Robert Axelrod's computer tournaments of the iterated Prisoner's Dilemma, Anatol Rapoport's Tit-for-Tat (TFT) strategy—cooperating on the first move and subsequently replicating the partner's previous action—achieved the highest cumulative score among deterministic rules. Tit-for-Tat sustains cooperation when δ ≥ max[(T − R) ÷ (T − P), (T − R) ÷ (R − S)]. In environments with stochastic execution or perception noise ε > 0, Tit-for-Tat can enter prolonged cycles of alternating retaliation; Generous Tit-for-Tat and Win-Stay, Lose-Shift (Pavlov) restore cooperative stability by forgiving isolated defections or resetting after mutual punishment.

Oligopolistic Collusion and Price Competition

In industrial organization, firms competing inside an oligopoly confront a multi-player Prisoner's Dilemma when setting output quantities or prices. If two duopolists restrict output to the joint monopoly level, both earn supra-competitive profits R. However, each firm faces a unilateral incentive to undercut the cartel price or expand output toward the Cournot competition reaction curve, capturing market share T at the rival's expense S. When both firms expand production simultaneously, market price falls toward marginal cost, leaving both in the lower-profit equilibrium P. Antitrust leniency programs deliberately alter the payoff matrix (T, R, P, S) by offering complete amnesty to the first cartel member that confesses to regulators, destabilizing collusive equilibria.

Public Goods and Common-Pool Resource Depletion

When generalized to n > 2 agents, the Prisoner's Dilemma models the voluntary provision of public goods and the tragedy of the commons. Suppose n pastoralists or fishing fleets share an open-access resource. Restraint preserves the regenerative stock (R), whereas over-extraction yields an immediate private gain (T) while diffusing ecological depletion costs across all n users. Institutional remedies—studied in mechanism design and transaction cost economics—modify stage-game payoffs through enforceable contracts, Pigouvian tax schedules, or Elinor Ostrom's polycentric community monitoring governance.

Why do rational players not simply agree to cooperate before the game?

In non-cooperative game theory, pre-play communication ("cheap talk") without a binding, externally enforced contract does not alter the underlying payoff matrix. Even if both players verbally promise to play C, once they make their simultaneous independent choices, each player knows that playing D yields a higher payoff whether the counterpart honors the promise (T > R) or breaks it (P > S).

How does the Prisoner's Dilemma differ from the Stag Hunt game?

In the Stag Hunt (assurance game), the payoff ordering is R > T ≥ P > S. Mutual cooperation (C, C) yields a higher payoff than unilateral defection (R > T), so a player prefers to cooperate provided they trust the partner will also cooperate. Consequently, the Stag Hunt possesses two pure-strategy Nash equilibria—(C, C) (payoff-dominant) and (D, D) (risk-dominant)—whereas the Prisoner's Dilemma has only one pure-strategy Nash equilibrium at (D, D).

Do human subjects always defect in laboratory Prisoner's Dilemma experiments?

No. Empirical behavioral economics experiments consistently record cooperation rates of 35% to 50% in one-shot anonymous Prisoner's Dilemma trials. Behavioral models attribute this departure from narrow material self-interest to social preferences, conditional reciprocity, bounded rationality, and altruism.

1. Tucker, A. W. (1983). The Mathematics of Tucker: A Sampler. The Two-Year College Mathematics Journal, 14(3), 228–232. https://doi.org/10.2307/3027092
2. Axelrod, R., & Hamilton, W. D. (1981). The Evolution of Cooperation. Science, 211(4489), 1390–1396. https://doi.org/10.1126/science.7466396
3. Fudenberg, D., & Maskin, E. (1986). The Folk Theorem in Repeated Games with Discounting or with Incomplete Information. Econometrica, 54(3), 533–554. https://doi.org/10.2307/1911307
4. Nowak, M. A., & Sigmund, K. (1993). A strategy of win-stay, lose-shift that outperforms tit-for-tat in the Prisoner's Dilemma game. Nature, 364(6432), 56–58. https://doi.org/10.1038/364056a0
5. Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press. https://doi.org/10.1017/CBO9780511807763