I agree with Richerson et al.'s stimulating article on its most important point: Human coalitions and institutions have a long cultural history. Their evolution may be driven by three different mechanisms (sect. 2.1.): the differential growth of biological groups; migrations from one group to another; and imitation of norms and other institutional arrangements. The authors' version of cultural group selection (CGS) makes use of all three mechanisms, presenting them as three complementary facets of a unified theory. I will argue that they have little in common (see Fig. 1).
Figure 1
(Morin). Three processes that, for Richerson et al., qualify as “cultural group selection,” and my interpretation of the assumptions and consequences behind each.
In the first form of CGS distinguished in the target article, groups are defined by three overlapping properties: (i) by descent (you are a group member if you are descended from group members); (ii) by preferential cooperation (group members tend to help one another rather than non-members); and (iii) by cultural links (they make use of the same institutional innovations). In this scenario, you cannot vote with your feet by going to a new group and becoming part of it; neither would you help someone who does not share your values, or fight someone who does. When the last two criteria are relaxed, two new types of CGS result. Type 2 CGS is migration, when individuals vote with their feet to join the societies that offer them the best deal (Boyd & Richerson Reference Boyd and Richerson2009). Here, groups are defined by culture and preferential cooperation alone. I am a member of the group of people whose norms I share and whose members I choose to cooperate with – even if I was not born there. Finally, in the third form of CGS (selective imitation) preferential cooperation must also be dropped, as is obvious from Richerson et al.'s logic and examples. Group members are defined by their common adoption of some institutional innovation – that is all: for example, when American firms copy the institutions of German research and development, they do so strictly for their own benefit, not as a way to ally themselves with German industry.
One can use CGS to predict, or explain, two very different things: (1) the evolution of altruistic behavior, and (2) equilibrium selection (André & Morin Reference André and Morin2011; Boyd & Richerson Reference Boyd and Richerson1990). Behavioral altruism is an uncompensated sacrifice from the individual to the group, with fitness consequences. Equilibrium selection is a process that favors the most stable forms of cooperation, including non-altruistic, mutually beneficial cooperation. Migration and imitation are more likely to work toward equilibrium selection: We can expect people to join the societies, or adopt the institutions, that give them a better deal. Type 1 CGS is more conducive to true altruism, because it short-cuts adaptive individual decisions. Much of the appeal of CGS is due to Type 1 CGS, and to the promise of finding a scientific rationale for altruism, as a way of backing attacks against standard rational choice theory (Fehr & Fischbacher Reference Fehr and Fischbacher2003; Sober & Wilson Reference Sober and Wilson1999). Yet, much of its plausibility derives from models of the second type, where mutually profitable forms of cooperation spread and no genuine altruism results. The authors' last two types of CGS, thus, do not seem capable of delivering what CGS is widely thought to explain, namely, altruism (André & Morin Reference André and Morin2011; Morin Reference Morin2014). Tellingly, the authors refrain from making any specific prediction regarding behavioral altruism (sect. 6.2.).
Both Type 2 and Type 3 CGS (migration and imitation) are “driven by human choices” (sect. 2, para. 1), as opposed to the relatively autonomous demographic dynamics that prevail in Type 1 CGS (birth-rates, wars, etc.). Types 2 and 3 of CGS must, at least occasionally, create group-level selection pressures that are fully aligned with individual-level selection pressures, and thus undistinguishable from them (unless we assume that human decision-making is so maladaptive as to decreases individual fitness systematically). People will vote with their feet: they will join the societies that promise them the best deal. They are likely to adopt institutional innovations that serve everyone's interests. Indeed, most of the examples cited by Richerson et al. concern mutually beneficial forms of cooperation. One can still use groups as units of accounting, but the causality resides entirely in individual choices. Contrary to what is stated in section 4.1, CGS is not, then, characterized by group-level selection pressures counteracting individual-level pressures, pushing individuals to sacrifice their own fitness for the group.
The disunity of CGS is deepened by the authors' inclusion of Type 3 CGS. Cultural diffusion need involve no demographic change, and no change in the composition of coalitions. The manner of cooperation is all that changes. When we use money, majority voting, or the academic practice of research seminars, we are not ipso facto allying ourselves with most of the people who use those practices. A state that decides to allow the legal use of class actions is not forming a coalition with the other states that allow it, or against those that do not. Here, we have sheer equilibrium selection with no differential growth, and no competition between coalitions. Group selection without all this is Hamlet without the prince.
What is wrong with heterogeneous theories? Science needs lumpers as well as splitters; but to test three very different hypotheses as if they were one, is to test only their weakest common denominator. Here, we are left with the uncontroversial view that coalitions or institutions vary and evolve. Richerson et al.'s more substantive suggestion is that useful institutions can only spread through some form of group selection – but to make this point, they dilute the concept of CGS to such an extent that it loses coherence and explanatory heft.
I agree with Richerson et al.'s stimulating article on its most important point: Human coalitions and institutions have a long cultural history. Their evolution may be driven by three different mechanisms (sect. 2.1.): the differential growth of biological groups; migrations from one group to another; and imitation of norms and other institutional arrangements. The authors' version of cultural group selection (CGS) makes use of all three mechanisms, presenting them as three complementary facets of a unified theory. I will argue that they have little in common (see Fig. 1).
(Morin). Three processes that, for Richerson et al., qualify as “cultural group selection,” and my interpretation of the assumptions and consequences behind each.
In the first form of CGS distinguished in the target article, groups are defined by three overlapping properties: (i) by descent (you are a group member if you are descended from group members); (ii) by preferential cooperation (group members tend to help one another rather than non-members); and (iii) by cultural links (they make use of the same institutional innovations). In this scenario, you cannot vote with your feet by going to a new group and becoming part of it; neither would you help someone who does not share your values, or fight someone who does. When the last two criteria are relaxed, two new types of CGS result. Type 2 CGS is migration, when individuals vote with their feet to join the societies that offer them the best deal (Boyd & Richerson Reference Boyd and Richerson2009). Here, groups are defined by culture and preferential cooperation alone. I am a member of the group of people whose norms I share and whose members I choose to cooperate with – even if I was not born there. Finally, in the third form of CGS (selective imitation) preferential cooperation must also be dropped, as is obvious from Richerson et al.'s logic and examples. Group members are defined by their common adoption of some institutional innovation – that is all: for example, when American firms copy the institutions of German research and development, they do so strictly for their own benefit, not as a way to ally themselves with German industry.
One can use CGS to predict, or explain, two very different things: (1) the evolution of altruistic behavior, and (2) equilibrium selection (André & Morin Reference André and Morin2011; Boyd & Richerson Reference Boyd and Richerson1990). Behavioral altruism is an uncompensated sacrifice from the individual to the group, with fitness consequences. Equilibrium selection is a process that favors the most stable forms of cooperation, including non-altruistic, mutually beneficial cooperation. Migration and imitation are more likely to work toward equilibrium selection: We can expect people to join the societies, or adopt the institutions, that give them a better deal. Type 1 CGS is more conducive to true altruism, because it short-cuts adaptive individual decisions. Much of the appeal of CGS is due to Type 1 CGS, and to the promise of finding a scientific rationale for altruism, as a way of backing attacks against standard rational choice theory (Fehr & Fischbacher Reference Fehr and Fischbacher2003; Sober & Wilson Reference Sober and Wilson1999). Yet, much of its plausibility derives from models of the second type, where mutually profitable forms of cooperation spread and no genuine altruism results. The authors' last two types of CGS, thus, do not seem capable of delivering what CGS is widely thought to explain, namely, altruism (André & Morin Reference André and Morin2011; Morin Reference Morin2014). Tellingly, the authors refrain from making any specific prediction regarding behavioral altruism (sect. 6.2.).
Both Type 2 and Type 3 CGS (migration and imitation) are “driven by human choices” (sect. 2, para. 1), as opposed to the relatively autonomous demographic dynamics that prevail in Type 1 CGS (birth-rates, wars, etc.). Types 2 and 3 of CGS must, at least occasionally, create group-level selection pressures that are fully aligned with individual-level selection pressures, and thus undistinguishable from them (unless we assume that human decision-making is so maladaptive as to decreases individual fitness systematically). People will vote with their feet: they will join the societies that promise them the best deal. They are likely to adopt institutional innovations that serve everyone's interests. Indeed, most of the examples cited by Richerson et al. concern mutually beneficial forms of cooperation. One can still use groups as units of accounting, but the causality resides entirely in individual choices. Contrary to what is stated in section 4.1, CGS is not, then, characterized by group-level selection pressures counteracting individual-level pressures, pushing individuals to sacrifice their own fitness for the group.
The disunity of CGS is deepened by the authors' inclusion of Type 3 CGS. Cultural diffusion need involve no demographic change, and no change in the composition of coalitions. The manner of cooperation is all that changes. When we use money, majority voting, or the academic practice of research seminars, we are not ipso facto allying ourselves with most of the people who use those practices. A state that decides to allow the legal use of class actions is not forming a coalition with the other states that allow it, or against those that do not. Here, we have sheer equilibrium selection with no differential growth, and no competition between coalitions. Group selection without all this is Hamlet without the prince.
What is wrong with heterogeneous theories? Science needs lumpers as well as splitters; but to test three very different hypotheses as if they were one, is to test only their weakest common denominator. Here, we are left with the uncontroversial view that coalitions or institutions vary and evolve. Richerson et al.'s more substantive suggestion is that useful institutions can only spread through some form of group selection – but to make this point, they dilute the concept of CGS to such an extent that it loses coherence and explanatory heft.