Loading field map
Standard economic theory long assumed that people care only about their own material payoffs. Yet from the earliest laboratory experiments, participants routinely gave money to strangers, punished unfair behavior at a cost to themselves, and cooperated in one-shot interactions where defection would have been the rational choice. These anomalies—acts that could not be explained by narrow self-interest—pressed economists to ask a deeper question: what do people actually care about when they interact with others? The answer has unfolded through a sequence of formal models, each proposing a different structure for social preferences and each responding to the limitations of its predecessors.
The first systematic attempt to model other-regarding motives was the altruism framework, introduced in the early 1970s. Its core idea was simple: an individual's utility depends not only on her own consumption but also on the consumption of others, weighted by a positive parameter. In its purest form, altruism treats others' payoffs as direct substitutes for one's own—the more others have, the better off the altruist feels. This framework captured charitable giving and some forms of cooperation, but it ran into a serious empirical problem. In laboratory games such as the ultimatum game, responders frequently rejected low offers even though rejection left both parties with nothing. A purely altruistic responder would have accepted any positive offer, since a small gain for herself and a larger gain for the proposer would have increased total welfare. The altruism model could not explain why people would sacrifice their own payoff to reduce someone else's. That failure pointed toward a different kind of motive: people care not just about outcomes but about how those outcomes came about.
In the early 1990s, a new wave of models shifted attention from outcomes to intentions. The reciprocity framework, formalized most influentially by Matthew Rabin in 1993, proposed that people want to be kind to those who have been kind to them and to punish those who have been unkind—even at a material cost. The key innovation was to embed these motives in psychological game theory, where players' utilities depend on their beliefs about others' intentions. A low offer in the ultimatum game is not merely a low payoff; it signals hostile intent, and the responder's rejection is a reciprocal act of punishment. This framework elegantly explained why the same material outcome could be accepted or rejected depending on the context that produced it. Yet reciprocity models were mathematically complex and required strong assumptions about how players form and update beliefs. They also struggled to explain behavior in games where intentions were ambiguous or where multiple players interacted simultaneously. The field needed a model that could match the empirical patterns while remaining tractable enough for applied work.
The inequity aversion framework, developed by Ernst Fehr and Klaus Schmidt in 1999, offered a return to outcome-based parsimony. Its central claim was that people dislike unequal outcomes—they experience disutility when their own payoff falls below a reference point (disadvantageous inequity) and, for some individuals, when their payoff exceeds others' (advantageous inequity). The model captured a wide range of experimental regularities with just two parameters: one for envy and one for guilt. It could explain costly punishment in the ultimatum game, cooperation in public goods games when punishment was available, and the puzzling fact that people sometimes reject offers that benefit them. Because inequity aversion was simple and could be plugged directly into standard game-theoretic models, it quickly became the workhorse of applied behavioral economics. But its very simplicity was also its limitation. The model treated all deviations from equality as equally aversive regardless of how they arose, yet experiments showed that people react very differently to the same unequal outcome depending on whether it was produced by fair or unfair procedures. Inequity aversion could not distinguish between a low offer that resulted from a random draw and one that resulted from deliberate selfishness.
Around the same time, a separate line of research was documenting a robust behavioral regularity in public goods games: many people cooperate if they believe others will cooperate, but defect if they believe others will defect. This pattern, labeled conditional cooperation, was formalized in models from 2001 onward. Unlike inequity aversion, which posits a fixed distaste for inequality, conditional cooperation makes cooperation depend on expectations about others' behavior. A conditional cooperator contributes to a public good not because she dislikes inequality per se, but because she wants to match the cooperation she expects from others. This framework captured the strong correlation between beliefs and contributions observed in experiments, and it explained why cooperation often unravels when participants learn that others have defected. However, conditional cooperation is often described as a behavioral pattern rather than a fully specified theory of underlying motives. It is compatible with several deeper mechanisms—reciprocity, inequity aversion, or even a desire to conform—and the framework itself does not commit to a single psychological driver. Its strength lies in its predictive power for aggregate behavior in repeated interactions, but it leaves open the question of what ultimately motivates the matching.
The most recent major framework, guilt aversion, emerged in the mid-2000s and returned to the belief-dependent approach of reciprocity, but with a different psychological mechanism. In guilt aversion, people experience disutility when they let others down—specifically, when they fall short of what others expect of them. The model was developed to explain the powerful effect of communication in economic games. When players can make promises before interacting, cooperation rates rise dramatically, and the guilt aversion framework accounts for this by modeling the promise as raising the promisee's expectations; breaking the promise then generates guilt. This framework differs from reciprocity in a crucial way: reciprocity focuses on responding to others' past actions, while guilt aversion focuses on meeting others' forward-looking expectations. In experiments with pre-play communication, guilt aversion often outperforms both inequity aversion and reciprocity, especially when promises are explicit and beliefs are measured directly. Yet guilt aversion shares the complexity of earlier belief-dependent models—it requires tracking players' beliefs about others' beliefs, which can multiply the theoretical degrees of freedom.
Today, no single framework has won the field. Instead, social preferences research has settled into a productive tension between outcome-based models (altruism, inequity aversion) and intention- or belief-based models (reciprocity, guilt aversion). Conditional cooperation sits between them, often used as a reduced-form description that can be generated by either family. The leading frameworks agree on one fundamental point: narrow self-interest is insufficient to explain behavior in strategic interactions. They disagree, however, on what the correct supplementary motive is. Outcome-based models argue that parsimony and tractability should take priority—they can explain a broad range of data with few parameters, and they are easy to embed in macroeconomic and policy models. Intention-based models counter that parsimony is worthless if it mispredicts behavior in contexts where intentions matter, such as negotiations, contracts, and political bargaining. The practical division of labor reflects this disagreement: inequity aversion dominates applied work in labor economics, public finance, and organizational economics, where the goal is to predict aggregate responses to policy changes. Reciprocity and guilt aversion are more common in experimental studies of communication, bargaining, and contract design, where the fine-grained structure of beliefs and intentions is essential. The coexistence of these frameworks is not a sign of failure but of a mature subfield that has learned to match its models to the phenomena they are asked to explain.