The Evolution of Reputation-Based Cooperation in Regular Networks

2019

Explaining the emergence of cooperation remains an open challenge in theoretical ecology and human evolution. Indirect Reciprocity (IR) constitutes one of the most elaborate mechanisms that provide a way out of what is known as the cooperation dilemma, relying on concepts such as social norms and reputations. In the last decades, several social norms have been identified as active promoters of cooperation in reputation-based settings. However, such norms often rely on the assumption that every action is observed by every member of a population, an assumption that cannot always be translated into real scenarios. A consequence of lifting this assumption is that reputations cease to be absolute and objective, i.e., independent opinions on agents’ reputations will tend to diverge throughout a population, defeating the advantages offered by IR. Here we propose an Evolutionary Game Theory computational model with a ternary reputation system, configurable observability of interactions and ...

Cooperation can be supported by indirect reciprocity via reputation. Thanks to gossip, reputations are built and circulated and humans can identify defectors and ostracise them. However, the evolutionary stability of gossip is allegedly undermined by the fact that it is more error-prone that direct observation, whereas ostracism could be ineffective if the partner selection mechanism is not robust. The aim of this work is to investigate the conditions under which the combination of gossip and ostracism might support cooperation in groups of different sizes. We are also interested in exploring the extent to which errors in transmission might undermine the reliability of gossip as a mechanism for identifying defectors. Our results show that a large quantity of gossip is necessary to support cooperation, and that group structure can mitigate the effects of errors in transmission. Cooperation among individuals is essential for their survival, in human and animal societies. Human beings are an intrinsically social species and most of our evolutionary success can be attributed to our highly developed ability to cooperate with each other. This ability is especially important in groups, where individuals need to coordinate their actions in order to achieve personal benefits that cannot be obtained without cooperation. However, those who do not contribute but reap the collective benefits are better off than cooperators 1. In models of indirect reciprocity 2–4 , cooperation can thrive when information about others is acquired either via direct observation, or via " image score " , a reliable and publicly visible indication of one's past cooperative behaviour 5. When modelled as simple scores, pro-social reputations are evolutionary stable only if they track behaviour with the same accuracy as direct experience 6. Cooperation becomes fragile when errors are possible , that is, when there is an even small probability for an individual to record a good partner as a bad one or vice-versa 7. Image score is effective in supporting group cooperation, but only when group size of individuals playing a Public Goods Game (PGG) does not exceed four 8. When group size increases, there is a concomitant decrease in the frequency of cooperation, showing that indirect reciprocity, even when supported by an image score mechanism , is not effective in large groups. The authors explain this decline in cooperation as due to the difficulty of observing reputations of many individuals in large communities. However, when agents are placed on a bipartite graph and they can actively select their group members, image score becomes effective in sustaining cooperation, even for groups of 20 individuals 9. Notwithstanding its effectiveness in supporting cooperation in models of indirect reciprocity, image score is limited by its reliance on direct observation. Thanks to language, humans are able to overcome this limitation and can exchange information about each other, thus isolating defectors and selecting cooperative partners 10,11. Thanks to gossip, we can map our social group 12 , learn about its rules 13 , and enforce social norms 14 , among other things. Gossip is crucial to make information about known cheaters travel within the network, thus allowing for identification of defectors, and it has a strong influence on the behaviour of participants in an economic experiment, even when they can rely on direct observation of others' actions 15. Gossip is also relatively more effective than punishment in promoting cooperation across a four-round PGG, it increases participants' gains and also efficiency, whereas punishment significantly decreases participants' earnings 16. In a computational study, gossipers who could actively select their group members and avoid ill-reputed agents are able to outperform free-riders and punishers in groups of 25 agents, whereas in smaller groups the combination of gossiping and material punishment is more successful in increasing cooperation levels 9. However, the evolutionary stability of gossip in supporting cooperation has been questioned. According to Nowak and Sigmund 17 , one of the main limitations of gossip consists in its being unreliable, while Ohtsuki, Iwasa

Physical Review E, 2008

We investigate the cooperation dynamics attributed to the interplay between the evolution of individual strategies and evolution of individual partnerships. We focus on the effect of reputation on an individual's partner-switching process. We assume that individuals can either change their strategies by imitating their partners or adjust their partnerships based on local information about reputations. We manipulate the partner switching in two ways; that is, individuals can switch from the lowest reputation partners, either to their partners' partners who have the highest reputation ͑i.e., ordering in partnership͒ or to others randomly chosen from the entire population ͑i.e., randomness in partnership͒. We show that when individuals are able to alter their behavioral strategies and their social interaction partnerships on the basis of reputation, cooperation can prevail. We find that the larger temptation to defect and the denser the partner network, the more frequently individuals need to shift their partnerships in order for cooperation to thrive. Furthermore, an increasing tendency of switching to partners' partners is more likely to lead to a higher level of cooperation. We show that when reputation is absent in such partner-switching processes, cooperation is much less favored than that of the reputation involved. Moreover, we investigate the effect of discounting an individual's reputation on the evolution of cooperation. Our results highlight the importance of the consideration of reputation ͑indirect reciprocity͒ on the promotion of cooperation when individuals can adjust their partnerships.

Proceedings of the First International Conference on …, 2008

A common finding is that cooperation is more likely if interactions are embedded in cohesive social networks, which is often explained by reputation mechanisms. An underlying assumption in these explanations is that networks are exogenously imposed on the actors. We relax this assumption and study a model in which actors play dyadic Prisoner's Dilemmas while also choosing their interaction partners, such that behavior and social networks co-evolve. Can cohesive networks and cooperation naturally evolve, is an exogenously imposed cohesive network a condition for cooperation, or are cohesive networks a result of high levels of cooperation? We propose formal model in which actors are modeled as boundedly rational and base their decisions to cooperate with a given partner on expectations of the partner's behavior. At the same time, they build or dissolve interactions based on the expected utility of these interactions. Actors form beliefs by learning from their own experience as well as from third-party information obtained via the network. We derive basic analytical results on stable states in this model, and apply computer simulations to study the dynamics of the process in detail. Results suggest that the spread of reputation does not always foster cooperation, and that network cohesiveness is more likely to be a consequence of cooperation than a cause.

Evolution, 2013

The threat of punishment usually promotes cooperation. However, punishing itself is costly, 22 rare in non-human animals, and humans who punish often finish with low payoffs in 23 economic experiments. The evolution of punishment has therefore been unclear. Recent 24 theoretical developments suggest that punishment has evolved in the context of reputation 25 games. We tested this idea in a simple helping game with observers and with punishment and 26 punishment reputation (experimentally controlling for other possible reputational effects). We 27 show that punishers fully compensate their costs as they receive help more often. The more 28 likely defection is punished within a group, the higher the level of within-group cooperation. 29 These beneficial effects perish if the punishment reputation is removed. We conclude that 30 reputation is key to the evolution of punishment.

ABSTRACT The standard method when analyzing the problem of cooperation using evolutionary game theory, is to assume that people are randomly matched against each other in repeated games. In this paper we discuss the implications of allowing agents to have preferences over possible opponents. We model reputation as a noisy observation of actual propensity to cooperate and illustrate how reputation based choice of opponents can explain both the emergence and deterioration of cooperation.

Journal of Theoretical Biology, 2005

Theory of indirect reciprocity is important in explaining cooperation between humans. Since a partner of a social interaction often changes, an individual should assess his partner by using social information such as reputation and make decisions whether to help him or not. To those who have 'good' social reputation does a player give aid as reciprocation, whereas he has to refuse to help those who have 'bad' reputation. Otherwise benefits of altruism is easily exploited by them. Little has been known, however, about the definition of 'goodness' in reputation. What kind of actions are and should be regarded as good and what kind of actions bad? And what sort of goodness enables sustaining exchange of altruism? We herein challenge this question with an evolutionary perspective. We generalize social reputation as 'Honor-score' (H-score) and examine the conditions under which individuals in a group stably maintain cooperative relationships based on indirect reciprocity. We examine the condition for evolutionarily stable strategies (ESSs) over 4096 possible cases exhaustively. Mathematical analysis reveals that only eight cases called 'leading eight' are crucial to the evolution of indirect reciprocity. Each in the leading eight shares two common characteristics: (i) cooperation with good persons is regarded as good while defection against them is regarded as bad, and (ii) defection against bad persons should be regarded as a good behavior because it works as sanction. Our results give one solution to the definition of goodness from an evolutionary viewpoint. In addition, we believe that the formalism of reputation dynamics gives general insights into the way social information is generated, handled, and transmitted in animal societies.

PloS one, 2012

In the well-mixed prisoner's dilemma game, individuals are typically assumed to have no choice about whether to interact with other individuals in the population. In this paper, we instead consider reputation-based conditional interaction and its consequences for the evolution of cooperation. Each individual has a tolerance range, and only interacts with other individuals whose reputation lies within its tolerance range in a chosen sample of the population. Reputation contains information about the number of interaction partners an individual has just cooperated with. We find that the introduction of conditional interaction promotes cooperation in well-mixed populations, and there exist moderate tolerance ranges for which this effect is maximized. For a given tolerance range, there is a critical cost-to-benefit ratio below which cooperation can be promoted. Interestingly, we find that if cooperation evolves, different cooperators' interaction clusters are typically maintaine...

2019

Cooperation among selfish agents can be promoted by allowing agents to condition behavior on reputation. Social norms -- dictating how agents update the reputations of others -- are central in determining whether this mechanism is effective. In particular, norms that reward justified defection have been shown to promote cooperation. A major limitation of existing models is that they assume all agents adopt a uniform norm, in a top down fashion. Here we show that when agents can spontaneously adopt novel norms, a learning process will see them drift towards socially undesirable outcomes. We present a model where agents can choose both how to react to reputations and how to assign the reputations of others -- making social norms emergent. In this scenario cooperation can only be achieved when the space of norms is severely restricted. In the real world, reputation systems have a mixed record. This is often attributed to the costly nature of assigning reputations, and the ability of ag...

SSRN Electronic Journal, 2000

In dyadic models of indirect reciprocity, the receivers' history of giving has a significant impact on the donor's decision. When the interaction involves more than two agents things become more complicated, and in large groups cooperation can hardly emerge. In this work we use a Public Goods Game to investigate whether publicly available reputation scores may support the evolution of cooperation and whether this is affected by the kind of network structure adopted. Moreover, if agents interact on a bipartite graph with partner selection cooperation can thrive in large groups and in a small amount of time.