Learn it → Apply it → Save as Skill
Two parties acting in their own self-interest can both end up worse off than if they had cooperated. The Prisoner's Dilemma is the classic game-theory situation where the individually "rational" choice, made by each side alone, produces a jointly worse result than mutual cooperation would have.
The Dr. Wallace's Favorite Mental Models is designed to turn useful ways of thinking into reusable tools. Each model gives you a fast visual overview, deeper interactive learning, practical application exercises and an AI-ready Markdown skill you can save for future use.
Instead of merely reading about mental models, you can gradually build your own AI skills library: a collection of portable reasoning skills that can help ChatGPT, Claude, Grok and other AI assistants apply the mental models you value.
When two parties each choose the option that is best for them individually, regardless of what the other does, they can land on an outcome that is worse for both than if they had trusted each other and cooperated.
Both cooperate: good outcome for both.
Both defect: worse outcome for both, yet each acted "rationally."
One defects, one cooperates: the defector does best of all; the cooperator does worst of all.
Important: the dilemma only exists because defecting looks individually safer no matter what the other side chooses, even though it is not the jointly best path.
The model applies to two-sided (or multi-sided) situations where cooperation would help everyone, but each side also has a private incentive to break that cooperation.
The dilemma is a structure, not a mood. Confirm the structure is actually present before assuming defection is inevitable.
Name the two (or more) parties whose outcomes depend on each other's choices, not just their own.
Estimate what each side gains or loses under mutual cooperation, mutual defection, and each one-sided outcome.
Ask whether each side does better by defecting no matter what the other side chooses. That is what makes it a true dilemma.
A one-shot interaction and an ongoing relationship behave differently. Repetition opens the door to reciprocity and trust.
Open a use case to see what the model changes.
Both firms would earn more if neither cut prices, but each fears the other will cut first and steal share. Map the actual payoffs before assuming a price war is unavoidable; sometimes public commitments or repeated interaction change the incentive.
Both sides gain from an honest, flexible agreement, but each also gains short term by quietly cutting corners. Building in verification, reputation, or repeated dealings changes what "rational" behavior looks like.
Everyone benefits if all contribute fully, but any individual can benefit personally by contributing less while others carry the work. Visible tracking and repeated collaboration reduce the payoff to free-riding.
Restraint benefits the group only if enough others also restrain. This overlaps with the Tragedy of the Commons but focuses on the strategic incentive each party faces rather than the resource itself.
Both nations would be safer or wealthier under mutual restraint, but each fears falling behind if it restrains alone. Verification agreements and repeated diplomacy are attempts to change the underlying payoffs.
A roommate or partner can benefit short term by contributing less to shared effort or expense. Clear agreements and an ongoing relationship usually do more to sustain cooperation than appeals to goodwill alone.
This is an illustrative example built to teach the payoff structure, using round numbers rather than real company data.
Expand for the origin of the model and the discovery that changed how people think about it.
Mathematicians Merrill Flood and Melvin Dresher formulated the underlying game in 1950 while working at the RAND Corporation, as part of research into strategic decision-making. Princeton mathematician Albert W. Tucker later gave the game its now-famous framing involving two prisoners being interrogated separately, which is where the name comes from.
In the one-shot version, defecting is a dominant strategy: each player does at least as well by defecting no matter what the other player chooses. When both players follow their dominant strategy, the result (mutual defection) is a Nash equilibrium, but it is not Pareto optimal, meaning both players could have been made better off by mutual cooperation.
In the late 1970s and 1980, political scientist Robert Axelrod ran computer tournaments pitting strategies against each other in the repeated (iterated) version of the game. The simplest submitted strategy, Tit for Tat, contributed by psychologist Anatol Rapoport, performed best overall. It cooperates on the first move, then simply mirrors whatever the other player did last time. Axelrod's key finding was that when interactions repeat, niceness, reciprocity, and forgiveness can outperform pure exploitation over the long run.
No. The one-shot Prisoner's Dilemma and the iterated Prisoner's Dilemma are different games with different stable outcomes. Repetition, reputation, and the possibility of retaliation change what counts as a rational move, which is why real-world relationships often sustain cooperation that a single isolated encounter would not.
Not every competitive situation has this exact payoff structure. Check whether each side truly has an individual incentive to defect regardless of the other's choice before applying the label.
Treating an ongoing relationship as if it were a single, anonymous encounter overstates the pull toward defection. Reputation and future interaction change the payoffs.
Defection is only the dominant move under specific assumptions: no communication, no enforcement, and no future consequences. Change any of those and the calculation changes too.
Talking can build trust, but without some form of commitment, verification, or repeated stakes, promises to cooperate may not be credible on their own.
The conclusion depends entirely on the size and shape of the payoffs. Rough or biased estimates can flip which strategy actually looks dominant.
Both involve a gap between individual and group interest, but the Prisoner's Dilemma is about strategic interaction between a small number of identifiable parties, while the Tragedy of the Commons is about overuse of a shared resource by many, often anonymous, users.
the two sides' interests are actually aligned, not partly opposed, so there is no real temptation to defect.
one side has overwhelming leverage over the other. A lopsided power relationship is not a symmetric strategic dilemma.
binding commitments, verification, or enforceable agreements are available. These can convert the game into a cooperative one entirely.
This browser tool structures your thinking. It does not send your information anywhere or perform AI analysis.
Save as Skill converts the model into portable Markdown instructions that teach an AI when and how to apply the Prisoner's Dilemma, what questions to ask, what evidence to seek, what mistakes to avoid and how to structure its analysis.
Save the skills you find useful and organize them into your own reusable library. The Markdown is intentionally model-neutral, so it can be used wherever an AI system accepts uploaded files, project knowledge, custom instructions or reusable text context. Platform capabilities differ, so this is a portable skill file rather than a claim of one-click installation.
No. They are close relatives. The Prisoner's Dilemma is usually framed between two identifiable parties choosing to cooperate or defect. The Tragedy of the Commons extends the same underlying tension to many parties sharing a limited resource, where each user's individual incentive to take more can deplete the resource for everyone.
Not by itself. Talking can help parties coordinate and build trust, but if promises are not enforceable, the underlying incentive to defect can remain. What tends to help more is verification, repeated interaction, reputational cost, or binding commitments that change the actual payoffs.
Tit for Tat is nice (it never defects first), retaliatory (it responds to defection immediately), forgiving (it returns to cooperation as soon as the other side does), and clear (its pattern is easy for an opponent to recognize and adapt to). Axelrod found that this combination outperformed more complex, exploitative strategies across repeated rounds.
No. Behavioral experiments consistently find higher rates of cooperation than pure self-interest would predict, especially when players can build reputation, communicate, or expect to interact again. The model describes an incentive structure, not a guaranteed prediction of human behavior in every case.