Evolutionary Game Theory

Five Principles for the Unification of the Behavioral Sciences Herbert Gintis Santa Fe Institute Central European University University of Siena

Princeton University Press 2009

Princeton University Press 2011

Evolutionary Game Theory Princeton University Press 2009

Disarray of the Behavioral Sciences The behavioral sciences (economics, sociology, anthropology, psychology, political science, biology) are in profound disarray, with incompatible models of human behavior. We now have the analytical and empirical basis for beginning to construct an integrated behavioral science.

Four Incompatible Models of Human Choice and Strategic Interaction Economics: Homo economicus, the self-regarding maximizer with unlimited and costless information processing capacity, who acts prosocially when the incentives align with selfish motives. Sociology: Homo sociologicus, the compliant actor socialized to fill social roles. Biology: The fitness maximizer whose prosociality is based on inclusive fitness (kin altruism) and selfinterested reciprocity (reciprocal altruism). Cognitive/Social Psychology: The irrational and illogical decision-maker.

Four Incompatible Models of Human Choice and Strategic Interaction The evidence for the existence and content of these four models comes from --what is taught in introductory graduate textbooks in the discipline. --what can be assumed in a disciplinary journal article without comment or defense. --what professionals in the discipline share in common, specialized fields being more nuanced.

Four Incompatible Models of Human Choice and Strategic Interaction At least three of these four models must be wrong. In fact they are all wrong, although all include fundamental insights that must be incorporated into a unified basic model of human choice and strategic interaction.

Five Principles for the Unification of the Behavioral Sciences 1. Gene-culture Coevolution (biology) 2. Socio-psychological Theory of Norms (sociology, cognitive psychology, social psychology) 3. Game Theory (economics, biology) Classical, Epistemic, Behavioral, and Evolutionary 4. The rational actor model, a.k.a. the Beliefs, Preferences, and Constraints (BPC) Model (economics, decision theory, biology). 5. Complexity Theory

Gene-culture Coevolution Individual fitness in humans depends on the structure of social life. Because culture is limited and facilitated by human genetic propensities, human cognitive, affective, and moral capacities are the product of an evolutionary dynamic involving the interaction of genes and culture (Cavalli-sforza and Feldman 1982; Boyd and Richerson 1985, 2004). This coevolutionary process has endowed us with preferences that go beyond the self-regarding concerns emphasized in traditional economic and biological theory.

Gene-culture Coevolution Gene-culture coevolution explains why we have a social epistemology facilitating the sharing of intentionality across minds, as well as why we have such non-self-regarding values as a taste for cooperation, fairness, and retribution, the capacity to empathize, and the ability to value character virtues (e.g., honesty)

Gene-culture Coevolution

g : genetic structure c : cultural structure e : environmental structure Note that it is plausible to add a link from c to e, so H(e)  H(e,c), or even from g to c, as in niche construction theory, so H(e)  H(g,c,e).

Evolutionary Game Theory Evolutionary game theory provides the analytical apparatus for building a dynamic model of changing gene frequencies and the distribution of cultural forms. Genes and culture obey similar dynamic laws, often captured by the replicator dynamic of evolutionary game theory. The analogy is not perfect, however, so cultural dynamics must be supplemented by several structural principles in addition to the “imitation” mechanism at the heart of the replicator dynamics.

Evolutionary Game Theory Important additions are the notion of vertical, horizontal, and biased transmission (Cavalli-sforza and Feldman, Boyd and Richerson) and the cognitive approach to norm-following (Castelfranchi and Conte).

The Socio-psychological Theory of Norms All social species have a division of labor, individuals being prepared for particular roles by nutritional and genetic differences. Human society has a division of labor characterized by dozens of specialized roles, appropriate behavior within which is given by social norms and individuals are prepared to fill these roles through a process of socialization.

The Socio-psychological Theory of Norms Socialization transmits moral values by leading individuals to internalize norms, through which they are induced to conform to the duties and obligations of the role-positions they will occupy. There are important limits to socialization, so additional causal processes are involved in cultural dynamics (Gintis, Journal of Theoretical Biology 2003).

The Socio-psychological Theory of Norms This concepts of socialization and the internalization of norms goes back to Durkheim (1902), but was later developed by Parsons, Goffman, and many others. The socio-psychological theory of norms supplies mechanisms missing from game theory that promote coordinated behavior.

The Rational Actor Model Evolutionary principles suggest that individual decision making can be modeled as optimizing a preference function subject to subjective beliefs and objective constraints. Natural selection leads the content of preferences to reflect biological fitness although the isomorphism between fitness and utility disappears outside the environment in which the preferences evolved. This is why humans do not maximize fitness in the postPleistocene era.

The Rational Actor Model The principle of expected utility extends this optimization to stochastic outcomes. The resulting model is called the rational actor model or the beliefs, preferences, and constraints (BPC) model.

The Rational Actor Model Four caveats are in order. First, individuals do not consciously maximize something called utility, or anything else. Second, individual choices, even if they are selfregarding (e.g., personal consumption) are not necessarily welfare-enhancing. Third, preferences must be stable across time to be theoretically useful, but preferences are ineluctably a function of an individual's current state.

The Rational Actor Model Fourth, beliefs are the Achilles’ heel of the BPC model, because the model treats beliefs as subjective, whereas individual beliefs are in fact constituted through a social network of interdependent beliefs.

Game Theory In the BPC model, choices give rise to probability distributions over outcomes, the expected values of which are the payoffs to the choice from which they arose. Game theory extends this analysis to cases where there are multiple decision makers. In the language of game theory, players are endowed with strategies, and have certain information, and for each array of choices by the players, the game specifies a distribution of payoffs to the players. Game theory predicts the behavior of the players by assuming each maximizes a preference function subject to available information, personal beliefs, and exogenous constraints.

Game Theory Game theory is a logical extension of evolutionary theory. Game theory is population biology with frequency dependent payoffs. Game theory has become the basic framework for modeling animal behavior (Maynard Smith 1982; Alcock 1993; Krebs and Davies 1997). Evolutionary and behavioral game theory do not require the formidable information processing capacities of classical game theory, so disciplines that recognize that cognition is scarce and costly can make use of contemporary gametheoretic models.

Game Theory In particular, agents may use by rule-of-thumb heuristics rather than maximization techniques (Gigerenzer and Selten 2001) and of course we need not assume individuals are self-regarding.

Society as Complex Adaptive System The behavioral sciences advance not only by developing analytical and quantitative models, but by accumulating historical, descriptive and ethnographic evidence that pays heed to the detailed complexities of life. Historical contingency is a primary focus for many students of sociology, anthropology, ecology, biology, politics, and even economics. By contrast, the natural sciences have found little use for narrative along side analytical modeling.

Society as Complex Adaptive System The reason for this contrast between the natural and the behavioral sciences is that living systems are generally complex, dynamic adaptive systems with emergent properties that cannot be fully captured in analytical models that attend only to the local interactions. The deductive methods of game theory, the BPC model, and even gene-culture coevolutionary theory must therefore be complemented by the work of behavioral scientists who adhere to a more historical and interpretive traditions.

Complex Systems A complex system consists of a large population of similar entities in our case, human individuals who interact through regularized channels (e.g., networks, markets, social institutions) with significant stochastic elements, without a system of centralized organization and control. A complex system is adaptive if undergoes an evolutionary (genetic or cultural) process of reproduction, mutation, and selection.

Complex Systems To characterize a system as complex adaptive does not explain its operation, and does not solve any problems. However, it suggests that certain modeling tools are likely to be effective that have little use in a noncomplex system. In particular, the traditional mathematical methods of physics and chemistry must be supplemented by other modeling tools, such as agent-based simulation, network theory, and narrative thick description.

Five Principles for the Unification of the Behavioral Sciences Redux 1. Gene-culture Coevolution (biology) 2. Socio-psychological Theory of Norms (sociology, cognitive psychology, social psychology) 3. Classical, Epistemic, Behavioral, and Evolutionary Game Theory (economics, biology) 4. The rational actor model, a.k.a. the Beliefs, Preferences, and Constraints (BPC) Model (economics, decision theory, biology). 5. Complexity Theory