Two players, each perfectly rational, each acting in self-interest — and together they land on the outcome both would trade away. The dilemma is not a mistake. Defection is the dominant strategy; mutual cooperation is Pareto-better yet structurally unreachable. This is the sharpest wedge ever cut between the individual and the group.
source Flood & Dresher, RAND experiment (1950); formalized in M. M. Flood, Some Experimental Games, RAND memorandum RM-789 (1952); the "prisoner's dilemma" framing due to A. W. Tucker (1950). No single stable canonical link for the 1950 origin — marked AMBER. rand.org/pubs/research_memoranda/RM789. Rendered, not quoted.
A symmetric 2×2 game. Each player picks Cooperate or Defect once, simultaneously. Four payoffs, one strict order:
T Temptation — defect while the other cooperatesNeighbour: a-the-nash-equilibrium.
A Nash equilibrium is a profile no one can improve on unilaterally. Usually that is where a game "solves." The prisoner's dilemma is the scandal at the heart of the concept: its unique equilibrium (Defect, Defect) is an equilibrium everyone regrets. Stability and desirability come apart. Every alignment problem, every tragedy of the commons, every arms race is this one 2×2 grid wearing a costume.Re-verifies the LIT invariant against the current payoffs on every change — dominance of Defect and the (D,D) equilibrium. Flips red the instant Red Team's tamper lands.
The payoff vector, in strict order:
"Rational players will just see that mutual cooperation is better and coordinate on it."
No. Rationality here is defined as best-response, and Defect is a best response to both opponent moves — the reasoning that reaches (C,C) requires binding agreement, repetition, or a payoff change, none of which exist in the one-shot game. Wanting the better outcome does not make it an equilibrium. The wall is that self-interest, correctly followed, walks you into the worse box."The players are irrational / made an error."
→ Both play their strictly dominant strategy. The outcome is fully rational and still bad.
"Communication would fix it."
→ Talk is cheap in the one-shot game; without enforcement each still defects. Only changed payoffs or repetition help.
"(D,D) is efficient because it's the equilibrium."
→ It is Pareto-dominated: (C,C) pays both strictly more. Equilibrium ≠ optimum.
Remove the temptation: reset payoffs so R > T. Now cooperating against a cooperator pays best, Cooperate becomes dominant, and the dilemma dissolves — so the "Defect dominates" invariant must fail. The Witness (7) catches it live.