For a decade or so Samuel Bowles and Herbert Gintis have played a central role in the debate on the evolution of cooperation. Building on the literature in experimental economics, they have proposed a number of influential evolutionary models aimed at explaining the ‘paradox’ of human cooperation. The paradox is well known. Cooperation in humans is unique in that it can be sustained among a large number of unrelated individuals, who often contribute to public goods and punish free-riders for seemingly disinterested reasons. The paradox stems from the fact that cooperation brings substantial benefits for the group, but is often costly to the individual. Uncooperative group members are then expected to have a fitness advantage over cooperative ones, which should prevent cooperation from evolving. A Cooperative Species offers a comprehensive review of potential game theoretic answers to this paradox, as well as an updated version of arguments and models presented by Bowles and Gintis over the years.
The first two chapters set the stage for the rest of the book. Bowles and Gintis explain why the evolution of cooperation is problematic and introduce the reader to key concepts such as altruism, cooperation, social preferences, social dilemmas and culture. The subsequent chapters offer a useful review of recent research on cooperation. Chapter 3 is organized around ten ‘key findings’ in experimental economics. Bowles and Gintis, for instance, review evidence indicating that strong reciprocity is common, that free-riders undermine cooperation but punishment helps sustain it, and that culture, institutions and group membership matter to cooperation.
Chapter 4 and 5 introduces the reader to a number of mathematical models that have been used, over the years, to account for the evolution of cooperation and altruism. The authors describe the conditions under which evolutionary mechanisms – such as inclusive fitness, direct reciprocity, indirect reciprocity, multi-level selection, costly signalling, or positive assortment – can favour the evolution of other-regarding preferences. Readers less fond of evolutionary game theory and theoretical biology will find the mathematical apparatus heavy and may be sceptical of the relevance of some of the material and issues presented. Chapter 5, for instance, is entirely dedicated to the so-called ‘folk theorem’ in game theory, whose relevance for the broader argument is not made entirely clear.
Overall, Bowles and Gintis do a good job at spelling out the strengths and weaknesses of the different evolutionary mechanisms that they describe. Even those familiar with this literature will learn new relevant nuances. The authors present their preference for multi-level selection, but also emphasize the very particular conditions that must be in place for this mechanism to be effective, namely, the need for cooperation to generate substantial benefits for the group and limited costs to the individual.
Chapter 6 aims at presenting the genetic, archaeological and ethnographic evidence supporting the multi-level selection hypothesis. First, Bowles and Gintis review ethnographic evidence on group size and genetic relatedness among historic hunter-gatherers. Extrapolating to the ‘Late Pleistocene and early Holocene’, they argue that our ancestors were living in bands that were too large and whose members were insufficiently related to one another for altruism and cooperation to have evolved primarily through kin selection and direct reciprocity (I come back below to Bowles and Gintis’ focus on the ‘Late Pleistocene and early Holocene’). Crucially for their argument, they also review evidence of ‘prehistoric warfare’ – showing that mortality due to lethal intergroup conflict was high enough among hunter-gatherers to have exerted a significant selection pressure – and on the role of ostracism, ridicule, humiliation and gossiping as means to achieve ‘social levelling’ among foragers.
Chapters 7 to 11 mostly build on models and simulations published by the authors over the last decade. They explore different facets of the evolution of cooperation and altruism: the evolution of institutions, parochialism, strong reciprocity, socialization and social emotions. The strength of those chapters is to demonstrate the sophistication and flexibility of existing mathematical models. Their weakness is to fail to explain clearly how the different questions actually fit into a comprehensive theory of the evolution of cooperation. Although the central argument is clear – that multilevel selection played a crucial role – the role of other mechanisms is less so. Granted, the objective of Bowles and Gintis is more to explain how cooperation could have evolved than to say how it did evolve. Nevertheless, their work raises questions about how game theoretical models can be anchored in archaeological data and evidence on how the mind works.
Chapter 7 is a case in point. The models and simulations presented aim at demonstrating the centrality of ‘group competition and culturally transmitted group differences in institutional structures’ to the evolution of cooperative behaviours. The central argument is that the selective extinction of less cooperative groups can be a powerful driver of the evolution of altruistic cooperation, but only given the right institutional structures. Bowles and Gintis thus emphasize the importance of ‘leveling institutions’, that is, institutions that ‘regulate competition among members’ and ‘attenuate within-group selective pressures operating against individually costly but group-beneficial behaviors’ (p. 113). Examples are food sharing beyond the family, information sharing, and reproductive levelling that prevents ‘dominant males from monopolizing reproduction’ (p. 112). By reducing within-group competition, levelling practices allow between-group selection to become fully effective.
Readers may find the argument question begging, however, in the sense that practices such as food sharing and reproductive levelling are generally understood as advanced forms of cooperation. In other words, they are the types of behaviours that a theory of the evolution of cooperation should seek to explain. But Bowles and Gintis do not consider them as such. They present levelling institutions not as cooperative outcomes, but as group-level traits resulting from ‘a cultural transmission process based on learned behaviors’. In their models, levelling institutions exist because ‘when new members of the population mature or immigrate, they adhere to the existing institutions [. . .] because this is a best response as long as others do the same’ (p. 112).
But is participation in food sharing and reproductive levelling really the best response ‘as long as others do the same’? If other players refrain from monopolizing food and females, isn't the best response rather to get as much food and females as one can? Bowles and Gintis suggest that ‘leveling institutions’ are not based on altruistic preferences. But why would males refrain from monopolizing females if no altruistic preferences were present in the group? An alternative interpretation of the models and simulations presented in chapter 7 would say that group selection can contribute to the evolution of cooperation, but only if cooperation is sufficiently well established to make some form of social levelling effective.
Another limitation to the argument stems from the way Bowles and Gintis connect their models with archaeological and ethnographic data. Throughout the book, they emphasize the need to understand the evolution of cooperation in the social and physical environments of the Late Pleistocene (i.e. approximately the period between 125 000 and 11 000 years ago). They argue that during this period average group size was too large for kin selection to have been effective and intergroup conflict was sufficiently frequent for selective group extinction to have been a significant selection pressure. I was somehow puzzled by this focus on the Late Pleistocene. If we accept large-game hunting, food sharing, reproductive levelling, cooperative breeding and group defence as proxies for altruistic cooperation, then strong evidence of extensive cooperation exists already for the Middle Pleistocene (with Homo heidelbergensis) and plausible evidence for the Early Pleistocene (with early Homo erectus) (Burkart et al. Reference Burkart, Hrdy and van Schaik2009; Dubreuil Reference Dubreuil2010a, Reference Dubreuil2010b). Bowles and Gintis's estimates for group size and intergroup conflict during the Late Pleistocene may well hold for these earlier periods.
In any case multilevel selection might have played a role for reasons that have nothing to do with lethal intergroup conflict. Bowles and Gintis agree with this idea. They note that the ‘differential survival of more cooperative groups need not have been the result of warfare’ and that the ‘tumultuous climate of the Late Pleistocene presented groups with extraordinary challenges quite apart from direct confrontations with other groups’ (p. 141). In other words, what really mattered for multi-level selection to have been effective was the fact that groups lived in social and physical environments in which cooperation was crucial to survival.
If this is the case, there may be a simple way for Bowles and Gintis to strengthen their argument in favour of multi-level selection. Rather than looking at Late Pleistocene human populations – which were probably composed of expert cooperators since hundred of thousands of years – they could turn their attention to hominid populations of the Late Pliocene and Early Pleistocene (about 2.5–1.5 million years ago), in which human-like forms of cooperation probably evolved for the first time. This period coincides with the evolution of the first members of the genus Homo, which occupied a significantly different niche than australopithecines (especially early Homo erectus).
Early Homo erectus, 1.8 million years ago, were active scavengers, travelling long distances to access carcasses with significant amount of meat and actively confronting other dangerous scavengers. They were living in relatively open environments and, in contrast with australopithecines, were probably not good at climbing. This last point is crucial because previous hominid species, just like apes today, were almost certainly using trees as refuges against predators, especially during the night. The evolution of an exclusive bipedalism indicates that groups of early Homo erectus were able to defend themselves effectively against predators, including large felidae. If intergroup violence was present among early members of the genus Homo, it is arguably predation that created the greatest pressure in favour of cooperation.
Even if predation pressure in Late Pliocene–Early Pleistocene environment was substantial and could have favoured the evolution of cooperation, Bowles and Gintis show that multi-level selection is unlikely in the absence of some ‘leveling practices’. But these practices, in turn, are likely to have required some form of altruistic cooperation, which takes us back to our starting point: how could the evolution of cooperation get started? The material presented in chapter 9 helps us advance our understanding. Bowles and Gintis discuss the evolution of ‘strong reciprocity’, a form of reciprocity in which players are ready to punish free-riders ‘altruistically’ (that is, at a net cost for them). Previous work has shown that, for strong reciprocity to evolve, ‘punishers must engage in enough punishment of defectors so that the induced cooperation more than offsets the cost of punishing’ (p. 149). The models and simulations presented in the chapter help understand the conditions under which this might become plausible. The key is that strong reciprocators must find a way to coordinate their punishment and make it conditional on the presence of a sufficient number of punishers. If punishers sanction free-riders only when they are sufficiently numerous to do so, punishment can become cost-effective.
Coordinated punishment is a welcomed improvement to Bowles and Gintis's framework. It also appears as a concession with respect to their previous work on ‘altruistic punishment’. If punishment can be made significantly cheaper, it might not be so ‘altruistic’ after all and its evolution would appear a lot less mysterious. But the mystery does not fully disappear: for coordination to be effective, information must be public, that is, all players must have access to honest signals about who is a free rider and who is a likely punisher. This implies either that ‘all members of a group see the same signals’ or that ‘they receive honest reports of the signals from those who did observe them’ (p. 165). The difficulty is that truthful signalling requires ‘ethical and other-regarding motives’, that is, ‘a system of information-sharing in which truth-telling was rewarded and lying punished’ (p. 165). But if information sharing is already a form of altruistic cooperation, we are back to our initial problem: how could cooperation get started?
The difficulty may not be insurmountable, however. To identify potential solutions, a close examination of the social and physical environments in which early forms of cooperation emerged in the human lineage – arguably at the boundary of the Pliocene and early Pleistocene – would be needed. As a first approximation, we can say that, in this period, groups of early humans moved into a niche characterized by some form of cooperative foraging (Tomasello Reference Tomasello2008). These groups were highly mobile and capable of group defence against predators and coordinated attacks on competitors. In all likelihood, they were composed of a relatively small number of individuals spending most of their time together, making monitoring and punishment of free-riders cost-effective. In this new ecological niche, cooperation would have increased substantially the probability of groups to survive, unleashing multi-level selection as an additional evolutionary driver.
As Bowles and Gintis emphasize, it may be more prudent at the moment to try to find out how cooperation might have evolved than how it did evolve. However, my intuition is that by focusing on what is known of early members of the genus Homo – rather than on Late Pleistocene Homo sapiens population – we could produce a limited range of scenarios about how human-like cooperation really evolved. Some parameters would be difficult to estimate (e.g. group size, predation pressure), but future archaeological research may help produce more reliable estimates.
Overall, A Cooperative Species does not give the fully integrated picture advertised in the book summary and leaves the reader with the conviction that additional work needs to be done to connect theoretical models with what we know of the evolution of human mind and behaviour. Nevertheless, Bowles and Gintis go a long way in building these connections, demonstrate once again the sophistication of existing game theoretic models and the benefits that result from seriously engaging with them.
