Behavioral game theory
Game theory rebuilt around how people actually play.
What it means
Behavioral game theory is the study of strategic interaction that incorporates how real people actually behave, rather than assuming perfectly rational, self-interested, and unboundedly calculating players. It modifies the three core assumptions of classical game theory by building in bounded rationality and limited strategic reasoning (people perform only a few steps of 'I think that you think'), social preferences such as fairness, reciprocity, and inequity aversion, and learning dynamics through which behavior adjusts over repeated play. The mechanism is to retain the formal structure of games while replacing idealized players with empirically grounded ones, then to test the resulting predictions against behavior in economic games run in the lab and the field. This matters because standard equilibrium concepts often mispredict real choices: the ultimatum game's frequent rejection of unfair offers, positive contributions in public-goods games, and trust and reciprocity in the trust game all defy pure self-interest yet are systematic and replicable. A nuance is that the program does not discard rationality wholesale but bounds and enriches it, seeking models that are both psychologically realistic and predictively precise. It matters for the design of markets, auctions, contracts, negotiations, and institutions, where assuming textbook rationality can lead designers badly astray.
Modeling limited reasoning
Classical theory assumes players iterate "I think that you think" endlessly until beliefs are mutually consistent. Behavioral game theory truncates that chain. Level-k and cognitive-hierarchy models start from a level-0 player who chooses without strategic thought, a level-1 player who best responds to level-0, and so on, with most people stopping after one or two steps. The p-beauty contest, where players pick a number and try to land near a fraction of the group average, is the workhorse test: the unique equilibrium is zero, yet first guesses cluster at values that imply only one or two rounds of reasoning. Camerer, Ho and Chong fit such data with a Poisson distribution of thinking steps averaging around 1.5, which explains why equilibrium predicts well in some games and badly in others.
Putting fairness into the payoffs
To explain why people reject money or share it, the program rewrites the utility function rather than the game. Fehr and Schmidt's inequity-aversion model adds two terms: a player loses utility when he earns less than others and, more weakly, when he earns more. With the right parameters that single specification reconciles selfish behavior in competitive markets with fair behavior in bargaining. A rival family of models makes the driver intentions rather than outcomes, so people reward kindness and punish hostility and the same split feels different depending on whether it was chosen or imposed. The two approaches diverge sharply in games where a fair outcome arrives by an unkind route, which is how experiments try to separate them. No single specification wins everywhere.
Adding noise to best response
Even players who reason well make mistakes, and a model that ignores this predicts crisp behavior that data never show. Quantal response equilibrium, introduced by McKelvey and Palfrey, keeps the equilibrium idea that beliefs are right on average but lets players pick better options more often rather than always. Choice becomes probabilistic: the costlier the error, the rarer it is, governed by a single precision parameter. As that parameter grows the model collapses back to Nash equilibrium; as it shrinks, play approaches random. This smoothing lets the theory match the spread of choices seen in experiments and rationalize systematic overbidding in auctions and over-entry into markets, where perfectly rational players would coordinate but noisy ones do not.
What the evidence shows, and where it strains
The core anomalies are among the most replicated results in economics, but their size is not universal. When Henrich and colleagues ran the ultimatum and related games across fifteen small-scale societies, offers and rejection rates varied widely and tracked local norms of exchange and sharing, undercutting any single fairness constant. Laboratory findings also face an external-validity challenge: Levitt and List argue that being watched, self-selected volunteers, and small stakes can inflate measured social preferences relative to anonymous, high-stakes markets. Behavioral game theory answers with field experiments and larger stakes, which shrink some effects without erasing them. The honest summary is that departures from self-interest and full reasoning are real and structured, but the right model and its parameters depend on the population and the setting.
Examples
It predicts the non-Nash but very human rejection of unfair offers in the ultimatum game.
Classical theory says nobody chips in for the office coffee fund. Behavioral game theory predicts what really happens: most people pay at first, then contributions decay as they watch the free-riders.
Auction designers who assume perfectly calculating bidders get caught out by the winner's curse; models that build in limited strategic reasoning predict the overpaying that actually occurs.
A firm setting a launch price assumes rivals reason one step further than they really do; modeling competitors as level-one thinkers, not flawless optimizers, forecasts the resulting price war more accurately.
In a gift-exchange labor setting, workers repay above-market wages with extra effort and coast when underpaid, so a model with reciprocity built in predicts productivity that pure self-interest misses.
First described in Colin Camerer (2003).
Key references
- Levitt, S. D., & List, J. A. (2007). What do laboratory experiments measuring social preferences reveal about the real world? Journal of Economic Perspectives, 21(2), 153-174. doi.org/10.1257/jep.21.2.153
- Camerer, C. F., Ho, T.-H., & Chong, J.-K. (2004). A cognitive hierarchy model of games. The Quarterly Journal of Economics, 119(3), 861-898. doi.org/10.1162/0033553041502225
- Henrich, J., Boyd, R., Bowles, S., Camerer, C., Fehr, E., Gintis, H., & McElreath, R. (2001). In search of Homo economicus: Behavioral experiments in 15 small-scale societies. American Economic Review, 91(2), 73-78. doi.org/10.1257/aer.91.2.73
- Fehr, E., & Schmidt, K. M. (1999). A theory of fairness, competition, and cooperation. The Quarterly Journal of Economics, 114(3), 817-868. doi.org/10.1162/003355399556151
- McKelvey, R. D., & Palfrey, T. R. (1995). Quantal response equilibria for normal form games. Games and Economic Behavior, 10(1), 6-38. doi.org/10.1006/game.1995.1023
- Nagel, R. (1995). Unraveling in guessing games: An experimental study. American Economic Review, 85(5), 1313-1326. www.jstor.org/stable/2950991