Abstract: More than half a century has now elapsed since coalition or alliance formation theory (CAFT) was first developed. During that time, researchers have amassed a vast amount of detailed and high-quality data on coalitions or alliances among primates and other animals. But models have not kept pace, and more relevant theory is needed. In particular, even though CAFT is primarily an exercise in polyadic game theory, game theorists have devoted relatively little attention to questions that motivate field research, and much remains largely unexplored. The state of the art is both a challenge and an opportunity. In this review we describe a variety of game-theoretic and related modelling approaches that have much untapped potential to address the questions that field biologists ask.

Abstract: Although many variants of the hawk-dove game predict the frequency at which group foraging animals should compete aggressively, none of them can explain why a large number of group foraging animals share food clumps without any overt aggression. One reason for this shortcoming is that hawk-dove games typically consider only a single contest, while most group foraging situations involve opponents that interact repeatedly over discovered food clumps. The present iterated hawk-dove game predicts that in situations that are analogous to a prisoner's dilemma, animals should share the resources without aggression, provided that the number of simultaneously available food clumps is sufficiently large and the number of competitors is relatively small. However, given that the expected gain of an aggressive animal is more variable than the gain expected by nonaggressive individuals, the predicted effect of the number of food items in a clump-clump richness-depends on whether only the mean or both the mean and variability associated with payoffs are considered. More precisely, the deterministic game predicts that aggression should increase with clump richness, whereas the stochastic risk-sensitive game predicts that the frequency of encounters resulting in aggression should peak at intermediate clump richnesses or decrease with increasing clump richness if animals show sensitivity to the variance or coefficient of variation, respectively.

Abstract: An altruistic individual has to gamble on cooperation to a stranger because it does not know whether the stranger is trustworthy before direct interaction. Nowak and Sigmund (Nature 393 (1998a) 573; J. Theor. Biol. 194 (1998b) 561) presented a new theoretical framework of indirect reciprocal altruism by image scoring game where all individuals are informed about a partner's behavior from its image score without direct interaction. Interestingly, in a simplified version of the image scoring game, the evolutionarily stability condition for altruism became a similar form of Hamilton's rule, i.e. inequality that the probability of getting correct information is more than the ratio of cost to benefit. Since the Hamilton's rule was derived by evolutionarily stable analysis, the evolutionary meaning of the probability of getting correct information has not been clearly examined in terms of kin and group selection. In this study, we applied covariance analysis to the two-score model for deriving the Hamilton's rule. We confirmed that the probability of getting correct information was proportional to the bias of altruistic interactions caused by using information about a partner's image score. The Hamilton's rule was dependent on the number of game bouts even though the information reduced the risk of cooperation to selfish one at the first encounter. In addition, we incorporated group structure to the two-score model to examine whether the probability of getting correct information affect selection for altruism by group selection. We calculated a Hamilton's rule of group selection by contextual analysis. Group selection is very effective when either the probability of getting correct information or that of future interaction, or both are low. The two Hamilton's rules derived by covariance and contextual analyses demonstrated the effects of information and group structure on the evolution of altruism. We inferred that information about a partner's behavior and group structure can produce flexible pathways for the evolution of altruism.

Abstract: Bekoff [J. Consci. Stud. 8 (2001) 81] argued that mammalian social play is a useful behavioral phenotype on which to concentrate in order to learn more about the evolution of fairness. Here, we build a game theoretical model designed to formalize some of the ideas laid out by Bekoff, and to examine whether `fair' strategies can in fact be evolutionarily stable. The models we present examine fairness at two different developmental stages during an individual's ontogeny, and hence we create four strategies--fair at time 1/fair at time 2, not fair at time 1/not fair at time 2, fair at time 1/not fair at time 2, not fair at time 1/fair at time 2. Our results suggest that when considering species where fairness can be expressed during two different developmental stages, acting fairly should be more common than never acting fairly. In addition, when no one strategy was evolutionarily stable, we found that all four strategies we model can coexist at evolutionary equilibrium. Even in the absence of an overwhelming database from which to test our model, the general predictions we make have significant implications for the evolution of fairness.

Abstract: To evaluate the role of individual recognition in the evolution of cooperation, we formulated and analyzed a genetic algorithm model (EvCo) for playing the Iterated Prisoner's Dilemma (IPD) game. Strategies compete against each other during each generation, and successful strategies contribute more of their attributes to the next generation. Each strategy is encoded on a `chromosome' that plays the IPD, responding to the sequences of most recent responses by the interacting individuals (chromosomes). The analysis reported in this paper considered different memory capabilities (one to five previous interactions), pairing continuities (pairs of individuals remain together for about one, two, five, or 1000 consecutive interactions), and types of individual recognition (recognition capability was maximal, nil, or allowed to evolve between these limits). Analysis of the results focused on the frequency of mutual cooperation in pairwise interactions (a good indicator of overall success in the IPD) and on the extent to which previous responses by the focal individual and its partner were associated with the partner's identity (individual recognition). Results indicated that a fixed, substantial amount of individual recognition could maintain high levels of mutual cooperation even at low pairing continuities, and a significant but limited capability for individual recognition evolved under selection. Recognition generally increased mutual cooperation more when the recent responses of individuals other than the current partner were ignored. Titrating recognition memory under selection using a fitness cost suggested that memory of the partner's previous responses was more valuable than memory of the focal's previous responses. The dynamics produced to date by EvCo are a step toward understanding the evolution of social networks, for which additional benefits associated with group interactions must be incorporated.