Cooperation may depend on more than whether people help or refuse to help. A new theoretical study led by Kobe University suggests that the way a person is treated can also reshape how observers judge that person—and that this overlooked process may help large societies maintain cooperation. The finding offers a simple explanation for a long-standing problem in evolutionary and social science: why individuals continue to pay costs to help others when communities become too large for everyone to know one another personally. It may also shed light on the social dynamics behind the bystander effect, exclusion, and bullying.
Cooperation is essential to human life, from sharing resources and caring for vulnerable people to participating in institutions that benefit entire communities. Yet from an evolutionary perspective, cooperation is puzzling. Helping another person can be costly, while the immediate benefit may go to someone else. Researchers have therefore spent decades studying indirect reciprocity, the process through which people gain or lose reputations based on how they treat others. A person known to be helpful may receive help in the future, even from individuals they have never assisted directly. In this way, reputation can transform isolated acts of generosity into a stable social system.
Most mathematical models of indirect reciprocity focus on pairwise interactions. One person, known as the donor, decides whether to help another person, known as the recipient. Observers then evaluate the donor’s action and update the donor’s reputation. If the donor helps someone considered deserving, the donor may be judged “good.” If the donor refuses to help a deserving person, the donor may be viewed negatively. These judgments influence future decisions, creating a feedback loop in which reputations guide behavior and behavior reinforces reputations.
The problem becomes much more complicated as communities grow. In a small pair-based network, people may be able to keep track of who helped whom and why. In larger groups, however, individuals observe different encounters, interpret them differently, or rely on incomplete information. Two observers can therefore assign different reputations to the same person. Such private disagreements can destabilize cooperation: someone judged trustworthy by one observer may be denied help by another, while a person considered unworthy by one group may still receive support from someone else. Earlier models have often addressed this problem by adding mechanisms such as gossip, reputation-sharing systems, or public institutions that synchronize social judgments.
The Kobe University team, working with collaborators from RIKEN and the Interdisciplinary Transformation University, examined whether cooperation could remain stable through a simpler mechanism. Instead of allowing observers to update only the donor’s reputation, their model allowed them to update their view of both people involved in the interaction. When Alice observes Bob helping Charlie, for example, she does not merely decide whether Bob acted well. She also learns something about Charlie from the fact that Charlie was helped. Alice might conclude that Charlie is regarded as deserving, or she might simply use Bob’s action as evidence about Charlie’s social standing. The interaction thus becomes a source of information about the recipient as well as the donor.
This dual update produces a powerful form of opinion alignment. The researchers identified a recipient-evaluation rule called recipient image scoring, or RIS, that was particularly effective. Under RIS, an observer evaluates a recipient positively if that person receives help and negatively if that person does not. The rule does not require observers to exchange messages, compare notes, or consult a central authority. Instead, people gradually revise their private opinions by watching how others are treated. If several observers see a person receive assistance, their assessments are more likely to converge. If the person is repeatedly denied assistance, observers may also move toward a shared negative evaluation.
The model suggests that this apparently simple rule can prevent disagreements from multiplying through a population. Without recipient evaluation, observers may disagree about whether a person deserves help, and those disagreements can influence how they judge later donors. With dual reputation updates, the treatment of recipients supplies an additional pathway through which information travels. A single interaction communicates not only that a donor chose to help or refuse, but also that the recipient was treated in a particular way. In technical terms, the model links the state of the donor’s reputation to the observer’s estimate of the recipient’s reputation, reducing divergence among private assessments and improving the conditions for stable cooperation.
To test how broadly the effect applied, the researchers used Fugaku, the supercomputer located in Kobe that was the world’s fastest supercomputer when it began regular operation in 2021. They examined more than one million combinations of social norms—rules specifying how people should act toward individuals with different reputations and how observers should evaluate those actions. The search identified 61 norms that produced both high levels of cooperation and evolutionary stability. Every one of these successful norms included some form of dual reputation updating, indicating that recipient evaluation was not merely a special feature of RIS. It appeared repeatedly among the most successful solutions in the model.
The results do not mean that gossip, public records, or institutions are unnecessary. Those mechanisms may still help societies coordinate judgments, especially when interactions are hidden or information is unreliable. The researchers instead propose that multiple systems may work together, each reducing disagreement in a different way. Their findings also point to a darker implication. People may judge someone not only by what that person does, but by how others treat them. That tendency can support cooperation when assistance signals social worth, but it can also reinforce exclusion, stigma, or bullying when observers imitate hostile treatment without independent evidence. Whether people consistently form reputations in this way remains an empirical question, but the model suggests that observing social treatment may be a fundamental and double-edged component of human judgment.
Subject of Research: People
Article Title: Indirect reciprocity with dual private assessment
News Publication Date: 24-Aug-2026
Web References: https://doi.org/10.1073/pnas.2624656123
References: Proceedings of the National Academy of Sciences; DOI: 10.1073/pnas.2624656123
Image Credits: Kobe University
Keywords: cooperation, reputation, indirect reciprocity, social norms, recipient image scoring, evolutionary stability, private assessment, social psychology, bystander effect, group behavior, prosocial behavior, computational modeling
