Loading field map
Game theory studies how decision-makers behave when the outcome each person receives depends not only on their own choices but also on the choices made by others. This interdependence—strategic interaction—is the field's defining subject. The central tension running through game theory's history is between the ideal of rational, self-interested calculation and the many ways real strategic situations resist that ideal: players may lack information, may interact repeatedly, may follow social or biological rules rather than deliberate reasoning, or may simply behave in ways that standard rationality cannot explain. Each major framework in the field emerged by relaxing, reframing, or rejecting the assumptions of its predecessors, and the frameworks that remain active today coexist because they address different aspects of this core tension.
The first comprehensive framework, Cooperative Game Theory, was laid out by John von Neumann and Oskar Morgenstern in The Theory of Games and Economic Behavior (1944). Their central idea was that when players can make binding agreements—contracts enforceable by an outside authority—the analysis can focus on what coalitions will form and how they will divide the gains from cooperation. The solution concept they proposed, the stable set (or von Neumann–Morgenstern solution), described which distributions of payoffs could survive coalitional objections. Cooperative theory treated the formation of agreements as given, not explained, and it assumed that players could commit to any deal they reached.
In 1950, John Nash published two papers that together split the field into complementary but competing traditions. The first paper, "The Bargaining Problem," founded Bargaining Theory as a framework within cooperative game theory. Nash proposed a set of axioms—symmetry, efficiency, independence of irrelevant alternatives, and scale invariance—that uniquely picked out a single bargaining outcome, the Nash bargaining solution. This solution did not explain how bargainers reach agreement; it specified what a rational agreement should look like. The second paper, "Equilibrium Points in N-Person Games," launched Non-Cooperative and Nash Equilibrium Theory. Here Nash abandoned the assumption of binding agreements entirely. Instead, each player chooses a strategy independently, and an equilibrium is a profile of strategies in which no player can gain by unilaterally changing their own choice. The Nash equilibrium became the central solution concept for situations in which agreements are either impossible or unenforceable.
These three frameworks—cooperative, bargaining, and non-cooperative—were not a simple sequence of replacements. They coexisted from the start, each answering a different question. Cooperative theory asked what coalitions could achieve; bargaining theory asked what a fair or rational division of surplus looked like; non-cooperative theory asked what stable outcomes emerged from uncoordinated individual choice. The non-cooperative approach gradually became the dominant backbone of the field because it required fewer institutional assumptions—no external enforcement of promises—and because it could be extended to dynamic and informational settings. But cooperative and bargaining frameworks never disappeared; they continued to inform analyses of negotiation, market design, and political economy.
Nash equilibrium assumed that players move once, simultaneously, and with complete knowledge of each other's payoffs. Two major extensions relaxed these assumptions in different directions.
Repeated Games (1956–present) emerged from work by Lloyd Shapley and others on infinitely repeated interactions. When the same strategic situation recurs, players can condition their current behavior on past actions. This creates the possibility of cooperation even in a non-cooperative setting: a player who defects today may be punished tomorrow. The key result, the folk theorem, showed that any feasible, individually rational payoff can be sustained as an equilibrium in an infinitely repeated game if players are sufficiently patient. Repeated games thus transformed the study of cooperation by explaining how it could arise without binding agreements—a question that cooperative theory had simply assumed away. The framework preserved Nash equilibrium as its solution concept but embedded it in a dynamic structure, showing that equilibrium could support a much wider range of outcomes than in one-shot play.
Bayesian Games and Incomplete Information (1967–present) addressed a different limitation: the assumption that everyone knows everyone else's payoffs. John Harsanyi, in a 1967–68 trilogy, introduced a method for converting games of incomplete information into games of imperfect information. He modeled each player's private information as a "type" drawn from a probability distribution, and defined equilibrium as a profile of strategies in which each player maximizes expected utility given their type and their beliefs about others' types. The resulting solution concept—Bayesian Nash equilibrium—extended non-cooperative theory to settings in which players are uncertain about each other's preferences, costs, or valuations. This framework became essential for analyzing auctions, bargaining with private information, and any strategic situation in which one side knows more than the other.
Repeated games and Bayesian games were parallel extensions, not competitors. Repeated games added a time dimension; Bayesian games added an information dimension. Both kept the core logic of Nash equilibrium—no unilateral profitable deviation—but applied it to richer environments. Together, they vastly expanded the range of economic problems that game theory could address, from oligopoly collusion (repeated games) to auction design (Bayesian games).
Mechanism Design Theory (1960–present) reversed the usual direction of analysis. Instead of taking the rules of a game as given and asking what outcomes would emerge, mechanism design asks: given a desired outcome, what rules would produce it? This is the "inverse problem" of game theory. The framework grew out of work by Leonid Hurwicz and was formalized through the revelation principle, which showed that any outcome achievable by some mechanism can also be achieved by a direct, truthful mechanism in which players report their private information and the mechanism implements the outcome. Mechanism design relies heavily on Bayesian Nash equilibrium as its solution concept: the designer must anticipate how players with private information will behave, and the equilibrium concept from Harsanyi's framework provides the standard tool for that analysis.
Mechanism design absorbed elements of both cooperative and non-cooperative traditions. It borrowed cooperative theory's interest in designing institutions that achieve efficient allocations, but it adopted non-cooperative theory's insistence that players must be motivated to participate and reveal information truthfully. The framework has been applied to auction design, public-good provision, matching markets, and regulation. Its relationship to earlier frameworks is one of reframing: rather than predicting behavior under fixed rules, mechanism design treats the rules themselves as choice variables, making game theory a tool for institutional engineering.
Two frameworks questioned the cognitive assumptions underlying Nash equilibrium and its extensions, but they did so from different directions.
Evolutionary Game Theory (1973–present) originated in biology with John Maynard Smith and George Price's 1973 paper "The Logic of Animal Conflict." Instead of assuming that players consciously calculate optimal strategies, evolutionary game theory models behavior as the result of selection pressures operating on a population. The central solution concept, the evolutionarily stable strategy (ESS), describes a strategy that, if adopted by most of the population, cannot be invaded by any alternative. This framework does not require rationality or even awareness; strategies are simply traits that replicate through success. Evolutionary game theory coexists with non-cooperative theory by providing a different justification for equilibrium: rather than being the outcome of deliberation, equilibrium emerges from a dynamic process of adaptation. Within economics, it has been used to study social norms, conventions, and market competition, and it has partly merged with behavioral and experimental approaches that test whether populations converge to predicted equilibria.
Behavioral Game Theory (1982–present) took a different path. Inspired by experimental anomalies—such as the ultimatum game results reported by Werner Güth and colleagues in 1982—behavioral game theory relaxes the assumptions of perfect rationality and purely self-interested preferences. It incorporates psychological findings about fairness, reciprocity, bounded rationality, and social preferences into formal game-theoretic models. Colin Camerer's 2003 book Behavioral Game Theory synthesized these developments, showing how models of "cognitive hierarchy" (players with different levels of strategic sophistication) and social preferences could explain experimental data that standard theory could not. Behavioral game theory does not reject Nash equilibrium as a benchmark; it treats equilibrium as a special case that holds only under strong assumptions about cognition and motivation. The framework's relationship to evolutionary game theory is complementary: evolutionary models explain how behavioral rules might survive in a population, while behavioral models test those rules in controlled laboratory settings.
Today, no single framework dominates game theory. Instead, the field operates as a division of labor among approaches that are best suited to different questions.
Non-cooperative and Nash equilibrium theory remains the default backbone for most applied work in microeconomics. Its extensions—repeated games and Bayesian games—are standard tools for analyzing dynamic competition, auctions, bargaining, and contracting. Mechanism design has become a central framework for market design and public policy, especially in settings where information is private and incentives must be aligned. These four frameworks (non-cooperative theory, repeated games, Bayesian games, mechanism design) share a commitment to equilibrium analysis and rational choice, and they agree that the Nash equilibrium concept, in its various forms, is the appropriate way to model strategic stability.
Evolutionary and behavioral game theory occupy a different position. They are not replacements for the non-cooperative tradition but critical complements that identify its limits. Behavioral game theory has been particularly influential in experimental economics and in policy design, where understanding actual behavior—not just idealized rational behavior—matters. Evolutionary game theory has found a home in theoretical biology and in parts of economics that study long-run institutional change.
Cooperative game theory and bargaining theory remain active but are more specialized. Cooperative theory is used in political science to study coalition formation and in cooperative game theory's own subfields (such as the Shapley value and the core). Bargaining theory straddles the cooperative and non-cooperative traditions: the Nash bargaining solution is still widely used in applied work, but it now coexists with non-cooperative bargaining models (such as Rubinstein's alternating-offers model) that derive the bargaining outcome from strategic interaction rather than axioms.
The leading frameworks today agree that strategic interaction is best analyzed through equilibrium concepts that specify mutual consistency of strategies. They disagree about the appropriate behavioral foundations of those equilibria. The non-cooperative tradition assumes that players are rational, self-interested, and capable of sophisticated strategic reasoning. Behavioral game theory argues that these assumptions are often violated and that models must incorporate psychological realism. Evolutionary game theory sidesteps the question of individual rationality entirely, focusing instead on population-level dynamics. The most active frontier areas—behavioral mechanism design and algorithmic game theory—represent syntheses across these traditions. Behavioral mechanism design combines the institutional engineering of mechanism design with the psychological realism of behavioral game theory, asking how mechanisms perform when participants are not fully rational. Algorithmic game theory, which emerged from computer science, studies the computational complexity of finding equilibria and designing mechanisms, adding a new constraint—computational tractability—to the traditional concerns of game theory. These syntheses suggest that the field's future lies not in a single framework but in the productive tension among them.