The title of the target article advances the bold claim that: “Cultural group selection [CGS] plays an essential role in explaining human cooperation.” By the end of the target article, the argument has been watered down: “evidence … [justifies] taking the CGS hypothesis seriously as [one] basic explanation” (sect. 7, para. 2, emphasis added). That is, vaguely characterized non-quantitative facts about humans do not prove the impossibility of CGS influencing the distribution of cooperative behavior, however minutely. We agree it is possible. Richerson et al. similarly retreat from CGS being an essential explanation for human cooperation, substituting the claim that it is an explanation for “our species' highly unusual ability to create large societies with widespread cooperation” (sect. 7, para. 2, emphasis added).
Because “large societies” are a recent evolutionary novelty, it follows that selection in ancestral small-scale societies was what predominantly forged the (genetically) evolved mechanisms that make cooperation in modern large-scale societies possible. Indeed, the context in which something evolves (e.g., cooperative mechanisms in foragers living in small-scale societies) does not predict its capacity to scale (e.g., cooperation in mass societies). Vision that evolved to see things closer than 3 miles can see for light-years; language capacities that evolved to allow our ancestors to speak to hundreds now allow us to speak to millions.
CGS proponents find the existence of cooperation in mass societies a self-evident evolutionary puzzle because the numbers involved evoke the impression that selection is not at equilibrium (which it need not be). In contrast, we think researchers need to carefully characterize the computational architectures of our evolved array of neural adaptations for exchange, delayed implicit reciprocity, risk-pooling, alliances, coalitions, coordination (such as theory of mind), bargaining, aggression, mateship, parenting, kin selection, partner choice, reputation, externality-management, social learning, and so on, together with their interactions. Only then can you know whether any puzzling residue of “group-beneficial behavior” in modern societies remains, requiring further explanation.
The dazzlingly extended forms of modern cooperation we see today (Adam Smith's division of labor supporting globe-spanning trade) appear differentially built out of adaptations for small-scale sociality that modularly scale, such as exchange – rather than the marginal benevolence of Smith's butcher, brewer, and baker. Evidence indicates that political attitudes toward welfare and redistribution reflect a specialized forager psychology of sharing for variance reduction (Petersen et al. Reference Petersen, Sznycer, Cosmides and Tooby2012) and resource-conflict (Petersen et al. Reference Petersen, Sznycer, Sell, Cosmides and Tooby2013). Societies that attempted to harness general benevolence to organize institutions and production – the USSR, East Germany, China, Cambodia, North Korea, Cuba – were spectacular cooperative failures. That they functioned at all depended on other scalable small-scale specializations – aggressive threats (conditional punishment), hierarchy, dominance, coalitions, and so forth.
Even ancestral foragers had institutions (enduring coordination with different roles). We do not understand why individually selected psychological adaptations for cooperation, coordination, coalitions, theory of mind, metarepresentations (i.e., x is a rule), intelligent instrumental reasoning (that allows locally contingent tailoring of actions to goals), social learning, a social psychology that understands and deploys incentives, hierarchies (and so on) are considered inadequate to explain institutions, then or now. It is puzzling why the authors believe that modern institutions cannot be far better explained without recourse to CGS, by the combined operation of these neural adaptations in dense, persisting social networks of intelligent, cultural agents (Boyer & Petersen Reference Boyer and Petersen2011). When the interlinked cognitive niche adaptations (Tooby & DeVore Reference Tooby, DeVore and Kinzey1987; see also, Pinker Reference Pinker2010) such as intelligence, language, and culture are added, it is difficult to see any obvious cooperative anomalies.
Gene-culture coevolution proponents claim to see overwhelming evidence of group-beneficial, individually costly behaviors in large societies that cannot be explained by (their computationally impoverished models of) reciprocity. For example, many results are interpreted as showing prosocial, other-regarding preferences purportedly inconsistent with individual selection, including a taste for fairness, excess generosity, and a failure to uniformly act with short-run selfishness. These preferences, together with a taste for altruistically punishing fairness norm-violators, are believed to work together to make people sacrifice their individual interests for the benefit of the group – which then helps groups in intergroup competition. However, the supporting experimental findings typically involve constrained choices that conflate hypotheses, rather than test them cleanly. When these defects are removed from experimental designs, supporting results collapse. For example, when subjects have the added choice of taking from others as well as giving, they no longer give in dictator games (List Reference List2007). Young children, purportedly averse to unfair divisions, will choose to pay a cost to reduce the welfare of others when given the chance (Sheskin et al. Reference Sheskin, Wynn and Bloom2014). In previous experimental designs testing for third-party punishment of unfair dividers, the only choices available were to punish or not. When this demand-characteristic is removed by adding the option of rewarding unfairness, average “altruistic” punishment approaches zero (Pedersen et al. Reference Pedersen, Kurzban and McCullough2013). Where partners can defect on both the subject and third parties, subjects punish those who defect on them personally (Krasnow et al. Reference Krasnow, Cosmides, Pedersen and Tooby2012). Moreover, they only punish those they subsequently choose to interact with, not those who could only harm others. This indicates that punishment is a tool of negotiation, and not primarily designed to altruistically uphold group norms.
Finally, the models typically used to represent individually selected strategies of social interaction (“reciprocity”) are strikingly impoverished (e.g., Cooperate, Defect), are largely free of social ecology and computation, and rest on many implausible assumptions. They need to be replaced with an adaptationist game theory. For example, to determine whether to cooperate, agents need to know whether an interaction is one-shot or repeated. When this aspect of real-world cooperation is added to simulations of reciprocity, then “excess” generosity reliably coevolves with reciprocity, eliminating observed generosity as an evolutionary puzzle (Delton et al. Reference Delton, Krasnow, Cosmides and Tooby2011). The rapid cultural dynamics of moral norms (think French Revolution) seem far better explained by rivalrous agents opportunistically seeking allies to jockey for self-advantageous norms (Tooby & Cosmides Reference Tooby, Cosmides and Hogh-Olesen2010), than by the slow accumulation of group-benefiting norms through some groups doing better than others. In 1789, institutions that developed over a thousand years were swept away in a thousand days. Thus, we may find that the code and open parameters of evolved programs underlie the combinatorial rules and building blocks of institutional cooperation.
The title of the target article advances the bold claim that: “Cultural group selection [CGS] plays an essential role in explaining human cooperation.” By the end of the target article, the argument has been watered down: “evidence … [justifies] taking the CGS hypothesis seriously as [one] basic explanation” (sect. 7, para. 2, emphasis added). That is, vaguely characterized non-quantitative facts about humans do not prove the impossibility of CGS influencing the distribution of cooperative behavior, however minutely. We agree it is possible. Richerson et al. similarly retreat from CGS being an essential explanation for human cooperation, substituting the claim that it is an explanation for “our species' highly unusual ability to create large societies with widespread cooperation” (sect. 7, para. 2, emphasis added).
Because “large societies” are a recent evolutionary novelty, it follows that selection in ancestral small-scale societies was what predominantly forged the (genetically) evolved mechanisms that make cooperation in modern large-scale societies possible. Indeed, the context in which something evolves (e.g., cooperative mechanisms in foragers living in small-scale societies) does not predict its capacity to scale (e.g., cooperation in mass societies). Vision that evolved to see things closer than 3 miles can see for light-years; language capacities that evolved to allow our ancestors to speak to hundreds now allow us to speak to millions.
CGS proponents find the existence of cooperation in mass societies a self-evident evolutionary puzzle because the numbers involved evoke the impression that selection is not at equilibrium (which it need not be). In contrast, we think researchers need to carefully characterize the computational architectures of our evolved array of neural adaptations for exchange, delayed implicit reciprocity, risk-pooling, alliances, coalitions, coordination (such as theory of mind), bargaining, aggression, mateship, parenting, kin selection, partner choice, reputation, externality-management, social learning, and so on, together with their interactions. Only then can you know whether any puzzling residue of “group-beneficial behavior” in modern societies remains, requiring further explanation.
The dazzlingly extended forms of modern cooperation we see today (Adam Smith's division of labor supporting globe-spanning trade) appear differentially built out of adaptations for small-scale sociality that modularly scale, such as exchange – rather than the marginal benevolence of Smith's butcher, brewer, and baker. Evidence indicates that political attitudes toward welfare and redistribution reflect a specialized forager psychology of sharing for variance reduction (Petersen et al. Reference Petersen, Sznycer, Cosmides and Tooby2012) and resource-conflict (Petersen et al. Reference Petersen, Sznycer, Sell, Cosmides and Tooby2013). Societies that attempted to harness general benevolence to organize institutions and production – the USSR, East Germany, China, Cambodia, North Korea, Cuba – were spectacular cooperative failures. That they functioned at all depended on other scalable small-scale specializations – aggressive threats (conditional punishment), hierarchy, dominance, coalitions, and so forth.
Even ancestral foragers had institutions (enduring coordination with different roles). We do not understand why individually selected psychological adaptations for cooperation, coordination, coalitions, theory of mind, metarepresentations (i.e., x is a rule), intelligent instrumental reasoning (that allows locally contingent tailoring of actions to goals), social learning, a social psychology that understands and deploys incentives, hierarchies (and so on) are considered inadequate to explain institutions, then or now. It is puzzling why the authors believe that modern institutions cannot be far better explained without recourse to CGS, by the combined operation of these neural adaptations in dense, persisting social networks of intelligent, cultural agents (Boyer & Petersen Reference Boyer and Petersen2011). When the interlinked cognitive niche adaptations (Tooby & DeVore Reference Tooby, DeVore and Kinzey1987; see also, Pinker Reference Pinker2010) such as intelligence, language, and culture are added, it is difficult to see any obvious cooperative anomalies.
Gene-culture coevolution proponents claim to see overwhelming evidence of group-beneficial, individually costly behaviors in large societies that cannot be explained by (their computationally impoverished models of) reciprocity. For example, many results are interpreted as showing prosocial, other-regarding preferences purportedly inconsistent with individual selection, including a taste for fairness, excess generosity, and a failure to uniformly act with short-run selfishness. These preferences, together with a taste for altruistically punishing fairness norm-violators, are believed to work together to make people sacrifice their individual interests for the benefit of the group – which then helps groups in intergroup competition. However, the supporting experimental findings typically involve constrained choices that conflate hypotheses, rather than test them cleanly. When these defects are removed from experimental designs, supporting results collapse. For example, when subjects have the added choice of taking from others as well as giving, they no longer give in dictator games (List Reference List2007). Young children, purportedly averse to unfair divisions, will choose to pay a cost to reduce the welfare of others when given the chance (Sheskin et al. Reference Sheskin, Wynn and Bloom2014). In previous experimental designs testing for third-party punishment of unfair dividers, the only choices available were to punish or not. When this demand-characteristic is removed by adding the option of rewarding unfairness, average “altruistic” punishment approaches zero (Pedersen et al. Reference Pedersen, Kurzban and McCullough2013). Where partners can defect on both the subject and third parties, subjects punish those who defect on them personally (Krasnow et al. Reference Krasnow, Cosmides, Pedersen and Tooby2012). Moreover, they only punish those they subsequently choose to interact with, not those who could only harm others. This indicates that punishment is a tool of negotiation, and not primarily designed to altruistically uphold group norms.
Finally, the models typically used to represent individually selected strategies of social interaction (“reciprocity”) are strikingly impoverished (e.g., Cooperate, Defect), are largely free of social ecology and computation, and rest on many implausible assumptions. They need to be replaced with an adaptationist game theory. For example, to determine whether to cooperate, agents need to know whether an interaction is one-shot or repeated. When this aspect of real-world cooperation is added to simulations of reciprocity, then “excess” generosity reliably coevolves with reciprocity, eliminating observed generosity as an evolutionary puzzle (Delton et al. Reference Delton, Krasnow, Cosmides and Tooby2011). The rapid cultural dynamics of moral norms (think French Revolution) seem far better explained by rivalrous agents opportunistically seeking allies to jockey for self-advantageous norms (Tooby & Cosmides Reference Tooby, Cosmides and Hogh-Olesen2010), than by the slow accumulation of group-benefiting norms through some groups doing better than others. In 1789, institutions that developed over a thousand years were swept away in a thousand days. Thus, we may find that the code and open parameters of evolved programs underlie the combinatorial rules and building blocks of institutional cooperation.