1J. McKenzie Alexander's The Structural Evolution of Morality falls within the scope of the growing evolutionary literature about morality and justice. As the title of the book clearly indicates, the author (a philosopher currently based at the London School of Economics) intends to give an evolutionary explanation of our moral behavior with the use of the most recent tools of evolutionary modeling. J.M. Alexander’s book asks several interesting questions, both methodological and substantive, that apply to the more general “evolutionary generalism” research program. Rather than providing a detailed summary of the book, we will focus on a methodological issue, namely: Are highly abstract evolutionary models relevant for explaining fairness principles and moral behavior?
2Alexander's book is made of seven chapters, two of them (the first and the last) being methodological and philosophical. The five other chapters present several evolutionary models of different basic moral behaviors: cooperation, trust, fairness and retribution. The main thesis is stated in the very first lines of the book: “The central claim of this book is that morality provides a set of heuristics that, when followed, serves to produce the best expected outcome, for each of us, over the course of our lives, given the constraints placed by other people” (p. vii). The book argues that morality stems from the actions of boundedly rational individuals. Despite the fact that individuals do not have the cognitive ability to reach the best outcome both in parametric and strategic situations, actually it appears that they are able to coordinate on near-optimal outcomes. The author presents morality as the main device for allowing some kind of “ecological rationality”.
3Alexander does not endorse the rationalistic approach of justice and morality (actually he does not explicitly discuss it). He also rejects the standard concept of rationality held by economists through the expected utility theory. In the introduction, he devotes several pages to describe the axioms of the expected utility theory and expresses a major concern with the assumption that there is a complete and transitive set of preferences (p. 17). This leads the author to reject the classical game theory framework, which is deemed incompatible with the assumption of bounded rationality. The evolutionary version of game theory seems to be a natural alternative since it allows studying the behavior of a boundedly rational individual in the pervasive context of repeated strategic interactions. In biology, evolutionary game theory relies on the assumption that strategies and behaviors are selected according to their associated expected fitness. However, the author's argument is independent of any biological assumption or mechanism. Instead, morality is the product of cultural evolution.
4In this framework, moral norms are equated with patterns of behaviors that have evolved because they allow individuals to solve frequent problems in strategic interactions. Already, we see that equating an optimal behavior in a strategic setting with a moral behavior is problematic, but we temporarily avoid this issue. Evolutionary game-theoretic models use a dynamic aggregative (or continuous) law for describing the evolution of the state of the population over time. Most of them use the replicator dynamics. The replicator dynamics is part of a subset of qualitative adaptive dynamics where a living strategy increases its proportion in the population when it has a higher fitness than the average fitness of the population. The replicator dynamics has a straightforward cultural interpretation once we substitute utility for fitness and once we assume that the higher the utility is, the more likely the strategy associated to it will be used in the population. As an aside, let us remark that this interpretation is not as straightforward as it seems to be because if payoffs are measured in utility, the link between payoffs and replication is empirically controversial. More important problems arise in spatially-structured models with imitate-the-best dynamics since it requires individuals to make interpersonal comparisons of utility. It is beyond the scope of this review to discuss further this crucial methodological issue.
5However, this approach is not entirely appropriate. The problem with aggregative dynamics is that it eschews the individual level; namely, it only describes the global state of the population and its evolution. This is even more problematic in the study of cultural rather biological phenomena since it is clear that human interactions are not purely random. Our environment is made of organizations and institutions that shape our interactions. In other words, human interactions are non-random. This is clearly the main message of Alexander's book: Morality is not the product of a blind cultural evolution but rather of structured interactions. The book is largely devoted to showing that structure matters in the evolution of morality. To defend this claim, the author contrasts aggregative dynamics with discrete dynamics used in agent-based models of social networks. Contrary to models based on the replicator dynamics, agent-based models do not assume infinite population and they assume spatially constrained interactions. Following his initial claim that individuals are boundedly rational, the author proposes several heuristic rules that help the individuals to choose their strategy in their interactions (pp. 39-42): (i) imitate the best neighbor, (ii) imitate with probability proportional to success, (iii) imitate best average payoff, and (iv) best response.
6The bulk of the book consists in five chapters where the author studies the evolution of different moral behaviors through several models of generic games: the prisoner's dilemma, the ultimatum game, the Nash demand game and the stag hunt. All chapters are organized in the same way. First, the evolution of moral behaviors such as reciprocity or fairness is analyzed through the replicator dynamics. Then, the evolution of the same behavior is investigated through agent-based models in using several spatial structures: Lattice models, small-world networks, bounded-degree networks and dynamic networks (where both strategies and the structure of interactions evolve). The same operation is repeated in each chapter for a specific moral behavior. A striking and presumably robust result appears: The tendency of moral behaviors to evolve under the replicator dynamics is reinforced once the structural dimension of interactions is taken into account. That is, models with a spatial dimension demonstrate a strong tendency to allow the evolution of cooperative or fair behaviors. The ultimatum game seems to be the only exception to what is an otherwise universal rule. In fact, Alexander's simulations show that local interactions seem to favor more self-interested behaviors.
7Alexander's book is engaging and generates some questions. Most of them are not specific to this work but they apply more generally to most of the research using evolutionary game theory to analyze the foundations of justice and morality, notably Skyrms (1996) but also Axelrod (1984). The exploration of spatially structured models is the real original contribution of the book since its arguments on the replicator dynamics are well known as well as the results based on it. However, the results of this exploration as well as their interpretation are not totally convincing. The final chapter lucidly discusses the main drawbacks of the approach adopted in the previous chapters.
8As the author himself notes, his account of the evolution of morality is focused on observable behaviors. He does not make any assumption on the underlying motivation leading individuals to behave morally. In Alexander's models, moral behavior is only a consequence of a local optimization grounded on a specific heuristic. However, one can argue that such explanation is only partial since knowing the psychological mechanism that leads individuals to act morally may be worthwhile. In other words, the models offered by Alexander, or also by Skyrms (1996), only provide a plausible explanation of morality in terms of ultimate causes. They ignore the proximate causes of morality. D'Arms et al. (1998) call this approach “evolutionary generalism” to contrast it with the “evolutionary particularism” of evolutionary psychology. Evolutionary particularism seeks to reveal what are the specific adaptive mechanisms that evolved through natural selection to generate a particular behavior. Evolutionary generalism is not committed to providing an explanation of the evolution of the (psychological) adaptive mechanisms. If we take for granted that both approaches have, as explanandum, some kind of behavior b solving a problem p in a situation S, then we can contrast both evolutionary explanations with their explanans as follows (see also Alexander, 2000):
According to evolutionary particularism, the explanans will consist in a specific mechanism M through which a given p in S will generate the behavior b. Accordingly, M is psychological. Then, evolutionary particularism consists in identifying a particular mechanism for a specific behavior and explaining its evolution.
Conversely, evolutionary generalism does not search for M. It aims at explaining the evolution of the behavior b according to its adaptive properties A measured by some particular criterion (fitness, utility). Evolutionary generalism does not relate A to a specific problem p but rather to a general class of problems P that are sufficiently similar to appeal to the same adaptive properties.
9Evolutionary particularism and evolutionary generalism are clearly different evolutionary accounts of morality. But they are not necessarily antagonistic. In fact, in the final chapter Alexander argues for a narrower association of evolutionary game theory with the theory of bounded rationality and with works that bridge the gap between economics and psychology. However, while he is lucid about the explanatory power of his approach, the author downplays some of the problems associated with evolutionary generalism. Undoubtedly, the main problem is related to the representativeness of the models. It is not obvious that they refer to any relevant situation in the real world. In fact, Alexander, as well as authors in the evolutionary generalism approach, such as Brian Skyrms or Robert Axelrod, starts from generic and simple situations. The deep exploration of these models is clearly insightful but it is uncertain how far its results can be extended. Economists are used to make and to hear arguments relying on very abstract thinking grounded on simple models. However, the usefulness of these models depends on their credibility, that is the way they relate to the real world. For example, while it is standard to study the evolution of the cooperation through the lens of the prisoner's dilemma (as most evolutionary biologists do), it is less clear that the evolution of fairness is appropriately captured through the simple Nash demand game. Moreover, because evolutionary generalism ignores proximate mechanisms, the explanations it provides are at best conjectural. The evolution of morality has probably been shaped by psychological and institutional factors: Evolutionary generalism does not help to reflect on them.
10Nevertheless, Alexander's book remains an important contribution to enhance our understanding of morality through evolutionary lens. It calls for more theoretical and empirical research, while pointing to the necessity for granting more attention to the methodological issues discussed above.