Behavioral Game Theory

Game theory is the mathematics of interdependent decisions: what you should do depends on what others do, and what they do depends on what they expect you to do.

9 frameworks, from 1950 to today · Behavioral Economics · Economics

still a working view ended · each leads to its entry below

Overview

Game theory is the mathematics of interdependent decisions: what you should do depends on what others do, and what they do depends on what they expect you to do. Its classical form, built by John von Neumann and Oskar Morgenstern in Theory of Games and Economic Behavior (1944) and extended by John Nash's equilibrium concept in 1950, assumes players are perfectly rational, purely self-interested, and capable of unlimited strategic reasoning. Behavioral game theory is the empirical and theoretical project of finding out what happens when real people—with limited attention, social feelings, and habits learned from experience—actually play. Its practitioners run controlled, usually paid, experiments; build models that predict the observed deviations; and test those models against new data. What is at stake is whether the elegant equilibrium machinery of economics describes human interaction at all, and if not, what should replace it.

Testing equilibrium in the laboratory

The founding move was to put game theory into the laboratory. In the 1950s and 1960s, researchers such as Merrill Flood and Melvin Dresher at the RAND Corporation ran the first prisoner's dilemma experiments, and Anatol Rapoport and colleagues at the University of Michigan studied repeated games and coalition formation extensively. These early studies found that people cooperated far more than equilibrium predicted and that behavior shifted with the framing of the game.

The decisive methodological step came from Vernon Smith, whose work from the late 1950s onward established that laboratory markets could be made rigorous. His key device was induced value theory: give subjects a payoff schedule that makes their monetary reward depend on the outcome, so that the experimenter controls their preferences and can test theory cleanly. Smith's double oral auction experiments showed that even unsophisticated traders converged toward competitive equilibrium prices—evidence that markets can work without rational traders. For this body of work he shared the 2002 Nobel Memorial Prize with Daniel Kahneman. The induced-value approach became the standard toolkit of experimental economics and, by extension, of behavioral game theory: if you want to know how people play a game, pay them according to its payoffs and watch.

Social preferences: people care about others

The most robust departure from standard theory is that players are not purely selfish. In the ultimatum game—one player proposes a split of a sum, the other accepts or rejects, and rejection leaves both with nothing—self-interested responders should accept any positive offer, and proposers should offer the minimum. Experiments from the early 1980s onward, notably by Werner Güth, Rolf Schmittberger, and Bernd Schwarze (1982), found that proposers typically offered 40–50 percent and that low offers were frequently rejected at real cost to the responder. The dictator game, in which the responder has no veto, isolated pure generosity: many allocators gave away part of their endowment even when they could keep everything1.

Explaining this required models in which utility depends on others' payoffs. Gary Becker had proposed in 1974 that people may care about others' welfare, but the modern formalization came in the 1990s. Matthew Rabin's 1993 model of fairness made kindness and retaliation depend on intentions, not just outcomes. Ernst Fehr and Klaus Schmidt (1999) and, independently, Gary Bolton and Axel Ockenfels (2000) built inequity aversion models: players dislike unequal outcomes, both when they are behind and when they are ahead. A concrete prediction of Fehr–Schmidt is that a player will pay to reduce an advantage she holds over someone else—behavior seen in experiments where subjects give up money to lower a partner's payoff. These social preference models made sense of a wide range of results, including costly punishment of free-riders in public-goods games, where contributors pay to penalize non-contributors even though punishment yields no material gain.

Learning: how behavior changes with repetition

In one-shot games, players have no chance to adapt. In repeated games, they do—and the pattern of adaptation became a research program in its own right. Alvin Roth and Ido Erev's 1995 paper "Learning in Extensive-Form Games" showed that simple reinforcement models, in which strategies that paid off in the past are more likely to be chosen again, tracked experimental data better than equilibrium predictions in many settings. The idea is close to psychological learning theory: choices are strengthened by reward.

A rival family of models assumed players best-respond to beliefs about others, updating those beliefs from observed play. Colin Camerer and Teck-Hua Ho's experience-weighted attraction (EWA) model (1999) combined reinforcement and belief-based learning in a single framework, letting the data decide how much weight to give each. The broader lesson was that "equilibrium" is often better understood as a limit that learning processes may or may not reach, and that the path matters.

Limited reasoning: level-k and cognitive hierarchy

A different puzzle is that people seem to reason only a few steps deep. In beauty-contest games—guess a fraction of the average guess, so that the winner is the one who anticipates others' reasoning—equilibrium requires infinite iterated reasoning, yet experimental guesses cluster at low levels. Rosemarie Nagel's 1995 study of such games found spikes at guesses corresponding to one or two steps of reasoning, not at the equilibrium.

Two models formalized this. Level-k models, developed by Dale Stahl and Paul Wilson (1995) and extended by Costa-Gomes, Crawford, and Broseta (2001), posit a hierarchy: a level-0 player follows a simple rule, a level-1 player best-responds to level-0, a level-2 player best-responds to level-1, and so on, with the population spread across levels. Cognitive hierarchy models, introduced by Colin Camerer, Teck-Hua Ho, and Juin-Kuan Chong (2004), make the same move but let higher-level players believe they face a distribution of lower-level types rather than a single type. These models predict behavior in many games—entry games, auctions, p-beauty contests—better than equilibrium, and they connect game theory to the psychology of bounded reasoning.

Beyond the lab: culture, development, and the brain

From the 1990s, researchers asked whether the findings were universal. Joseph Henrich and colleagues' cross-cultural ultimatum and dictator game studies, published in Behavioral and Brain Sciences in 2001 and in the book Foundations of Human Sociality (2004), found enormous variation across small-scale societies: some groups made near-zero offers, others made generous ones, and market integration and cooperation in everyday life predicted game behavior. This undercut any simple claim that fairness preferences are fixed human constants. A parallel developmental literature, using simplified games with children, traced how strategic reasoning, fairness, and punishment emerge with age—work associated with researchers such as Ernst Fehr, Bettina Rockenbach, and their collaborators.

Meanwhile, neuroeconomic game theory began in the early 2000s to ask what happens in the brain during strategic play. Studies using functional magnetic resonance imaging, such as Alan Sanfey and colleagues' 2003 paper on the ultimatum game, found that unfair offers activated the insula and anterior cingulate—regions linked to negative emotion—and that activity predicted rejection. This gave a physiological correlate to the social-preference findings and helped motivate dual-process accounts in which emotional reactions and deliberative control compete.

The present

Behavioral game theory today is a mature experimental science with several live strands. Social-preference models remain central but are increasingly tested against alternatives such as guilt aversion (players dislike letting others down) and conditional cooperation (people cooperate when they believe others will). Learning models and level-k/cognitive-hierarchy models are standard tools for predicting behavior in auctions, markets, and strategic settings, and are routinely compared against equilibrium in new experiments. Cross-cultural and developmental work has expanded into field experiments and large-scale replication efforts, and neuroeconomic methods continue to probe the mechanisms behind strategic choice. The open questions are substantial: how stable preferences are across contexts, how learning and reasoning interact, whether laboratory findings generalize to high-stakes field settings, and how to build models that are psychologically realistic without losing predictive power. The field's founding insight—that the rational-actor benchmark is a starting point, not a description—remains its organizing principle.

References

  1. Dictator game, Wikipedia

The frameworks of Behavioral Game Theory

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1950-Present

    Experimental game theory

    How do people actually play strategic games?

    Fly to its tower in Nousfarer

  2. 1976-Present

    Experimental Economics Methodology

    How can laboratory experiments test economic theory reliably?

    Fly to its tower in Nousfarer

  3. 1993-Present

    Social preferences

    Do people care about others' payoffs?

    Fly to its tower in Nousfarer

  4. 1995-Present

    Learning models

    How does behavior change with repeated play?

    Fly to its tower in Nousfarer

  5. 1995-Present

    Level-k reasoning models

    How many steps deep do people reason?

    Fly to its tower in Nousfarer

  6. 2000-Present

    Developmental game theory

    When do fairness and strategy emerge in children?

    Fly to its tower in Nousfarer

  7. 2001-Present

    Cross-cultural experimental economics

    Are fairness preferences the same across cultures?

    Fly to its tower in Nousfarer

  8. 2003-Present

    Neuroeconomic game theory

    What happens in the brain during strategic play?

    Fly to its tower in Nousfarer

  9. 2004-Present

    Cognitive hierarchy theory

    How do people form beliefs about others' reasoning?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Experimental game theoryExperimental Economics MethodologySocial preferencesLearning modelsLevel-k reasoning modelsDevelopmental game theoryCross-cultural experimental economicsNeuroeconomic game theoryCognitive hierarchy theory

Beyond the field

Where Behavioral Game Theory's ideas came from in other fields, and what they led to there: the strongest 12 of 43 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Behavioral Game Theory is a planet in Nousfarer

Its city has a tower for each of these 9 frameworks. Land, walk in, and climb one concept at a time.

Land on Behavioral Game Theory

Related fields

Fields its ideas are linked to

More in Behavioral Economics

See where Behavioral Game Theory sits on the Map of Human Knowledge