FROM THE ACADEMY: COLLOQUIUM INTRODUCTION COLLOQUIUM INTRODUCTION In the light of evolution VIII: Darwinian thinking in the social sciences Brian Skyrmsa,1, John C. Aviseb,1, and Francisco J. Ayalab,1 norms are seen as a society’s way of coor- aDepartment of Logic and Philosophy of Science and bDepartment of Ecology and dinating on one of these equilibria. This, Evolutionary , University of California, Irvine, CA 92697 rather than enforcement of out-of-equilibrium behavior, is seen as the function of these Darwinian thinking in the social sciences was Evolutionary was initially norms. Fairness norms, however, may involve inaugurated by Darwin himself in The De- seen by some social scientists as just a way interpersonal comparisons of utility. This is scent of Man and Selection in Relation to to provide a low rationality foundation for true of egalitarian norms, and of the norms Sex high rationality equilibrium concepts. A Nash of proportionality suggested by Aristotle. In (8, 9). Despite various misappropriations ’ of the Darwinian label, true Darwinian think- equilibrium is a rest point of a large class of Binmore s account, the standards for such ing continued in the social sciences. However, adaptive dynamics. If the dynamics converges interpersonal comparisons themselves evolve. toarestpoint,itconvergestoNash.Thismay Komarova(22),inanarticlethatcould with the advent of Social Dynamics (10) there has been an explosion of Darwin- nothappeninallgames,butitmayhappen equally well be in ,discusses ian analysis in the social sciences. And it has in large classes of games, depending on the the speed evolution of complex phenotypes in led to reciprocal contributions from the social dynamics in question. asexual populations. Initiation of new com- sciences and mathematics to biology. The However, another idea from evolutionary plex phenotypes may require changes in mul- theory of games had been created (11) as a biology shows that the dynamical point of tiple genes. Accumulation of all of the req- theory of interaction among rational agents view can be more subversive. That is the idea uisite mutations may require crossing a with a (tacit) presumption of common knowl- of population structure. Interactions are not fitness valley, which could take a very long edge of rationality. Evolutionary game theory always best modeled as random encounters time. Various factors may affect this time. broadened the scope of game theory by re- in a large population. There may be correla- For instance, spatial interaction (as mod- moving the rationality assumption, and re- tion, positive or negative, between the strat- eled by a contact process) can make a sig- – placing it with an adaptive dynamics of dif- egies that interact. Hamilton (17 19) and nificant difference. Social interaction among ferential reproduction. Social scientists soon Price (20) clearly saw that such correlations cells, modeled using division-of-labor games, saw that a whole array of other broadly lieatthebasisofevolutionaryexplanationsof can also accelerate the evolution of complex adaptive dynamics of imitation, social learn- both altruism and spite. Sources of such cor- phenotypes. There are applications to bio- ing, inductive reasoning with best response, relationforhumansarecentralareasofcon- film formation and to evolution of cancer. and so forth, were relevant and could be ana- cern for social sciences: interaction on a Christakis and Fowler (23) investigate how lyzed with similar tools (12). social network, homophily or heterophily in the formation of human social ties is sensitive Evolutionary game theory was brought to network formation, reputation and partner to genetic differences. Sociologists have long the fore by Maynard Smith (10), but its ori- choice in repeated interactions, formation known that humans are more likely than and dissolution of groups for collective ac- chance to make friends with others who re- gins go back earlier, even to Darwin himself — (8, 9). The idea of an evolutionarily stable tion, and honest and dishonest signaling. semble them phenotypically that the pro- of Maynard Smith and Price (13) All are discussed in this colloquium. cess of friendship formation is homophilic. derives from the “unbeatable strategy” that Evolution of Social Norms There are various possible reasons for this: Onebeingthatthereissometendencyto Hamilton used in his analysis of sex ratios It is a commonplace in the social sciences ’ prefer genetically similar individuals, and that (14). Hamilton sanalysisisexplicitlygame that the values that are manifest in individual — genetic similarity is correlated with pheno- theoretic. The payoff in expected grandchil- decisions may reflect norms of the society — typic similarity, so that the phenotype serves dren of a sex ratio strategy depends of the in which the individual lives. Social norms strategies of all of the rest of the population. themselves evolve. They exhibit both com- This is a playing-the-field game in Maynard monalities and differences across cultures. This paper serves as an introduction to this PNAS supplement, Smith’sterminology.Hamiltonbuildsonthe This suggests that both biological and cul- which resulted from the Arthur M. Sackler Colloquium of the Na- reasoning of Fisher (15) in explaining human — — tional Academy of Sciences, “In the Light of Evolution VIII: tural evolution and coevolution play a Darwinian Thinking in the Social Sciences,” held January 10–11, sex ratios. Under reasonable assumptions, dif- role in their explanation. 2014, at the Arnold and Mabel Beckman Center of the National ferential reproduction drives the sex ratio to- In “Bargaining and fairness” (21) Binmore Academies of Sciences and Engineering in Irvine, CA. The complete ward that empirically observed. Fisher quotes discusses the evolution of fairness norms as program and audio files of most presentations are available on the Darwin from the second edition of the De- devices for selection between multiple equi- NAS website at www.nasonline.org/ILE-Darwinian-Thinking. This is scent of Man the eighth in a series of Colloquia under the umbrella title “In the (9), where Darwin says that the libria. Consideration of equilibrium outcomes Light of Evolution” (see Box 1). Papers from previous Colloquia in problem was unsolved. However, in the first in repeated interactions leads one to a bar- the series appear in refs. 1–7. ’ edition (8), Darwin gives essentially Fisher s gaining game. This game has an infinite Author contributions: B.S., J.C.A., and F.J.A. wrote the paper. argument (16). Darwin had, in a way, invented number of Pareto optimal equilibria. Achiev- The authors declare no conflict of interest. evolutionary game theory. The idea had to ing an efficient social contract requires se- 1To whom correspondence may be addressed. Email: fjayala@uci. wait until the present to come full flower. lection among these equilibria, and fairness edu, [email protected], or [email protected].

www.pnas.org/cgi/doi/10.1073/pnas.1411483111 PNAS | July 22, 2014 | vol. 111 | suppl. 3 | 10781–10784 Downloaded by guest on September 30, 2021 response dynamics and adaptive dynamics Box 1. In the light of evolution. In 1973, Dobzhansky (37) penned a short com- are discussed. There is a generalization to mentary titled, “Nothing in biology makes sense except in the light of evolution.” Most games with a continuum of strategies. There scientists agree that evolution provides the unifying framework for interpreting bi- is a generalization to population games— ological phenomena that otherwise can often seem unrelated and perhaps un- Maynard Smith’s playing-the-field games, intelligible. Given the central position of evolutionary thought in biology, it is sadly which include the sex ratio game of Fisher and ironic that evolutionary perspectives outside the sciences have often been neglected, Darwin. There is a generalization to multi- misunderstood, or purposefully misrepresented. Biodiversity—the genetic variety of player games. Asymmetric games are treated, life—is an exuberant product of the evolutionary past, a vast human-supportive re- including those in which players have several source (aesthetic, intellectual, and material) of the present, and a rich legacy to cherish moves in an extensive form game. and preserve for the future. Two challenges, as well as opportunities, for 21st-century The next two papers deal with dynamics of science are to gain deeper insights into the evolutionary processes that foster biotic individual learning. Erev and Roth (26) argue diversity and to translate that understanding into workable solutions for the regional that human learning dynamics provides a and global crises that biodiversity currently faces. A grasp of evolutionary principles good explanation of many of the experimen- and processes is important in other societal arenas as well, such as education, medi- tal findings of behavioral economics. Subjects cine, sociology, and other applied fields including agriculture, pharmacology, and who learn by reinforcement will, in some biotechnology. The ramifications of evolutionary thought extend into learned realms kinds of games, rapidly learn to play a Nash traditionally reserved for philosophy and religion. The central goal of the “In the Light equilibrium. However, in other kinds of of Evolution” (ILE) series is to promote the evolutionary sciences through state-of-the- games, including the infamous ultimatum- art colloquia and their published proceedings. Each installment will explore evolu- bargaining game, learning to play Nash tionary perspectives on a particular biological topic that is scientifically intriguing but may be difficult. Experiments are in accord also has special relevance to contemporary societal issues or challenges. Individually with these theoretical predictions. Learning and collectively, the ILE series aims to interpret phenomena in various areas of biology from experience tends to underweight rare through the lens of evolution, address some of the most intellectually engaging as well events because there is less experience with as pragmatically important societal issues of our times, and foster a greater appreci- them. This is in contrast to learning by ation of evolutionary biology as a consolidating foundation for the life sciences. description, which tends to overweight rare events. One large loss early on may keep subjects from exploring alternatives suffi- Social Dynamics as an indication of genotype. The authors test ciently to discover optimal actions. In this hypothesis statistically. There are possible Sometimes societies and cultures change general, it is easy to learn the optimal ac- confounding factors. For instance, individuals rapidly.Sometimestheymoreorlessbehave tion when it gives the best payoff both on livinginthesamelocationmaybegenetically as if they are approximately in equilibrium, average and most of the time. These prin- similar. Using a large dataset, the authors are just as some species continue changing, ciples are applied to , able to control for these factors. They estab- whereas others remain the same. Darwinian such that the desired behavior is easily and lish that, overall, humans in their study do approaches to the social sciences need to quickly learnable. Empirical case studies have a tendency to prefer friends who are study dynamic processes of social evolution. are given. genetically similar. Further analysis shows The explanatory significance of equilibrium Fudenberg and Levine (27) also discuss that for some sets of genes, for instance those depends on the underlying dynamics. Is the learning dynamics. In contrast with the pre- involved in immune response, the tendencies equilibrium a rest point of a plausibly op- viouspaper(26),thisisatheoreticalin- runintheoppositedirection. erative dynamics? Is it a local attractor? Are vestigation whose focus is on the long run. One way of sustaining cooperation in there many rest points, posing an equilib- Psychological studies show that in learning, interactions where cooperation seems prob- rium selection problem? Is there any reason recent trials count more than older ones. One lematic—such as prisoner’s dilemma or public to believe that the dynamics will lead to a rest model of such recency bias has recursive goods provision games—emerges when the point at all, rather than cycling or exhibiting discounting of the past. The distant past is interactions are repeated. Cooperative equi- even more complicated behavior? The dy- always there, but with small weight. A dif- libria can be sustained in con- namics here range from models of genetic ferent kind of model postulates a limited texts by rewards and punishments. Kandori and cultural evolution to those of individual memory capacity. More recent experience is and Obayashi (24) provide a detailed case learning. more likely to be found in memory. The study in human cooperation in a setting close Cressman and Tao (25) give an overview authors consider a model of each kind, and to theoretical models of overlapping gen- of deterministic evolutionary dynamics, be- show how they are related. In certain con- erations of repeated games. The case under ginning with the replicator dynamics. Repli- ditions, for instance with a long enough examination is a community union, a kind of cator dynamics was introduced in a biological limited memory, the models give related as- labor union unique to Japan. Individuals may setting as a simple model of differential re- ymptotic results. This includes a kind of join or exit at any time, so that typically there production, but it also has a social inter- universal consistency, which generalizes that are overlapping generations of membership. pretation as a simple model of differential used in their previous work. Convergence to Existing theory called for perfect monitoring imitation. It was originally formulated in a strict , provided that there of behavior, or alternatively, construction and large one-population setting, where fitness is such in the game, is proved. dissemination of reputations, to support a co- accrues from random interactions between The next paper returns to evolutionary operative equilibrium. This case study led to pairs of individuals playing symmetric (ma- dynamics, but looks at the interplay of mul- the formulation of an overlapping generations trix) games with a finite number of strate- tiple dynamics. Creanza and Feldman (28) model and to the discovery of a new equi- gies. This paper also treats generalizations, present a dynamical model of cultural niche librium that fits the observations. extensions, and alternatives. Alternative best- construction. There are three traits: a focal

10782 | www.pnas.org/cgi/doi/10.1073/pnas.1411483111 Skyrms et al. Downloaded by guest on September 30, 2021 ’ trait, a trait that can alter the selection pressure O Neill (31) gives a case study in cultural discussion of empirical phenomena calling for COLLOQUIUM INTRODUCTION on the focal trait, and a trait that affects evolution based on policy folklists. These are theoretical analysis. We have seen examples assortative mating. Extensive simulations are lists of supposed facts that appear to be rel- of all of these already in this colloquium. Here used to sweep the parameter space. The evant to social policies. They are picked up by we have application of a number of dynamics model is found to be capable of exhibiting media, and copied, sometimes modified, and of evolution and learning to signaling, appli- complex dynamical behavior. This includes recopied over generations. Often these lists cation of learning with a finite level of exper- cycles, stable polymorphisms, and simulta- have no true foundation in reality, but have imentation to spread of innovation through neous stability of oscillation and fixation. A features that enhance their reproductive a social network, and application of online number of applications of the model are success as memes. He discusses both experimentation to cultural differences in a discussed. These range from education and similarities to genetic evolution and rel- . contraception to agriculture and animal evant differences. Phenotypic plasticity, Huttegger et al. (34) discuss the dynamics domestication. self-repair, speciation, and predation all of signaling games. They consider replicator Levin (29) focuses on problems of collec- come under consideration. dynamics, replicator–mutator dynamics for tive action that are common to all social In “On the evolution of hoarding, risk- large populations, finite population dynamics species, from humans to bacteria. Central taking, and wealth distribution in nonhuman in the rare mutation limit, and individual concerns are (i) the creation of a public good and human populations” (32) Bergstrom reinforcement learning. These are applied to through individual contributions and (ii) analyzes optimal strategies for savings in games of common interest and opposed in- collective prudent management of common an environment with stochastic payoffs. terest and games where interests can be pool resources. These are really two sides of This is a problem for both humans and aligned by costly signaling. Throughout it is the same coin. Some species have solved some animals., Bergstrom uses squirrels shown that to predict the outcome of play, it these problems, but humans are still strug- stocking nuts to survive winter as his focal is not sufficient to rely on classical equilib- gling with them on a global scale. The paper example. However, storing additional nuts rium concepts. In signaling games where focuses on three central issues. The first is the exposes the squirrels to an additional pre- perfect signaling is the unique evolutionarily nature of discounting the future. The second dation risk, therefore one should not stock stable strategy and the unique strict Nash is prosociality—the extent to which individ- too many nuts. However, because winters equilibrium, equilibrium may or may not uals value the welfare of others. The final one may be long or short, the squirrels have to evolve and may or may not be generated by is the nature of collective decision making. take their chances. The optimal gene may be reinforcement learning. In games where no- The discussion is complex. Some of the fac- one that plays a mixed strategy, producing information-transfer is the unique Nash tors that make evolution of prosociality pos- more or less conservative phenotypes. The equilibrium, replicator dynamics may never sible, such as local interaction in space and optimal strategy is analyzed first for long reach equilibrium, but instead result in repeated interaction in time, can also make and short winters, and then for variable chaotic dynamics, where signals always bear possible the evolution of spite. For humans, length winters. Finally it is shown how the some information. In finite populations, the evolution of social norms is seen as playing analysis is impacted by the presence of a population may spend a substantial pro- a critical role. redistribution mechanism. portion of the time signaling, even if sig- Morgenstern et al. (33) approach empirical naling is not a Nash equilibrium at all. Special Sciences phenomena of color constancy and color Kreindler and Young (35) study spread of Various sciences are, of course, represented contrast, which are instances of the inverse innovations through social networks. Indi- throughout this colloquium. This section in- problem in optics. The perceiver makes in- viduals play a 2 × 2 coordination game with cludes papers with special interdisciplinary ferences to the physical state of the world neighbors on the network. There are two perspectives: biodemography, evolutionary based on stimuli reaching the retina that un- equilibria: Both play status quo and both play political science, bioeconomics, and neuro- derdetermine, and sometimes seem to mis- innovation. The innovation is assumed to be psychology. represent, the physical state. Somehow the risk dominant. Players change, one at a time, Wachter et al. (30) discuss evolutionary nervous system has evolved to use contextual by smoothed best response [e.g., logistic best factors that shape demographic schedules. information to compensate for this mis- response, softmax] to the play of their They review three main approaches, which interpretation, at least in cases impacting re- neighbors. Previous results in the literature have proceeded more or less independently. productive fitness. Color constancy is a case have focused on the effect of different net- Mutation accumulation theory postulates a in point. Colors originating from the same work topologies on the speed of diffusion. steady stream of mutations with age-specific physical surfaces appear the same under This paper derives topology-free results pro- effects on mortality. In a mutation-selection different illumination even though the viding that errors (or experiments) have a equilibrium, there will be more deleterious stimuli reaching the retina are different. large enough finite rate, and the payoff gain alleles that act at a later age. Stochastic vitality The authors see this as an evolutionary ad- from innovation is large enough. Upper theory postulates a heterogeneous population aptation. They investigate it via the dynam- bounds for expected waiting times are impacted by stochastic shocks. Robustness to ics of artificial neural networks. Simulated derived, first for regular networks and shocks changes over life history. Optimal life evolution leads their model neural networks to then for general networks. The model is history theory studies the balance between display color constancy and color contrast. A then reinterpreted as one where there is factors that influence reproductive success double-opponent system, similar to that found a strict best response, but the payoffs are over a lifetime. The authors prove a theorem in humans and some animals, evolves. subject to random shocks. about mutation accumulations that raises Jackson and Xing (36) present a compara- an explanatory challenge for biodemography, Applications tive study across two cultures, India and and discuss possible approaches for address- We may have applications of general theory to United States. The experiment is conducted ing it. They close by considering the prospects a specific theoretical problem, or applications online using Amazon’s Mechanical Turk. for combining the three different approaches. of theory to empirical phenomena, or a The game used was a coordination game—a

Skyrms et al. PNAS | July 22, 2014 | vol. 111 | suppl. 3 | 10783 Downloaded by guest on September 30, 2021 battle-of-the-sexes game augmented by an greater proportion of players than in the biology, mathematics, medicine, neuropsy- additional strategy. There are two asymmetric United States chose strategies that led, in chology, political science, sociology, and phi- pure equilibria from the battle-of-the-sexes equilibrium, to asymmetric payoffs. Strategies losophy—with many of the papers being the component: Player 1 gets a lot and player 2 a were presented to players as different colors. result of collaboration of authors from dif- little, or conversely, player 2 gets a lot whereas The effect of subtly suggesting strategies to ferent disciplines. These papers give us a rep- play 1 a little. If players coordinate on the players, by telling them that they are in a resentative picture of the ongoing Darwinian additional strategy, they both get equal pay- room corresponding to a particular strategy, thinking in the social sciences. offs, less than a lot and more than a little. is investigated. ACKNOWLEDGMENTS. The organizers and founding Average total winnings in most of the games The essays collected here cross many editors of the “In the Light of Evolution” series (J.C.A. and were US$1; some experiments were run with disciplinary lines, both between evolu- F.J.A.) are the academic grandson and son, respectively, higher payoffs with similar results. Overall, tionary theory and the social sciences, and of Theodosius Dobzhansky, to whose fond memory this series is dedicated. May Dobzhansky’s words and insights the majority of players chose the strategy that between individual social sciences. They rep- continue to inspire rational scientific inquiry into nature’s led to equal payoffs. However, in India a resent demography, economics, evolutionary marvelous operations.

1 Avise JC, Ayala FJ (2007) From the Academy: Colloquium 12 Fudenberg D, Levine D (1998) The Theory of Learning in Games 26 Erev I, Roth AE (2014) Maximization, learning, and economic perspective: In the light of evolution I: Adaptation and (MIT Press, Cambridge, MA). behavior. Proc Natl Acad Sci USA 111:10818–10825. complex design. Proc Natl Acad Sci USA 104(Suppl 1): 13 Smith JM, Price GR (1973) The logic of animal conflict. Nature 27 Fudenberg D, Levine DK (2014) Regency, consistent learning, – 8563–8566. 246(5427):15 18. and Nash equilibrium. Proc Natl Acad Sci USA 111:10826–10829. 14 2 Avise JC, Hubbell SP, Ayala FJ (2008) Colloquium paper: In the light Hamilton WD (1967) Extraordinary sex ratios. A sex-ratio theory 28 Creanza N, Feldman MW (2014) Complexity in models of cultural of evolution II: Biodiversity and extinction. Proc Natl Acad Sci USA for sex linkage and inbreeding has new implications in cytogenetics niche construction with selection and homophily. Proc Natl Acad Sci – and entomology. Science 156(3774):477–488. 105(Suppl 1):11453 11457. USA 111:10830–10837. 3 15 Fisher RA (1930) The Genetical Theory of Natural Selection Avise JC, Ayala FJ (2009) In the light of evolution III: Two centuries 29 Levin SA (2014) Public goods in relation to competition, – (Clarendon, Oxford). of Darwin. Proc Natl Acad Sci USA 106(Suppl 1):9933 9938. cooperation, and spite. Proc Natl Acad Sci USA 111:10838–10845. 16 Edwards AW (1998) Natural selection and the sex ratio: Fisher’s 4 Avise JC, Ayala FJ (2010) Colloquium paper: In the light of 30 sources. Am Nat 151(6):564–569. Wachter KW, Steinsaltz D, Evans SN (2014) Evolutionary shaping evolution IV: The human condition. Proc Natl Acad Sci USA 17 Hamilton WD (1963) The evolution of altruistic behavior. Am Nat of demographic schedules. Proc Natl Acad Sci USA 107(Suppl 2):8897–8901. 97(896):354–356. 111:10846–10853. 5 Strassmann JE, Queller DC, Avise JC, Ayala FJ (2011) In the light 18 Hamilton WD (1964) The genetical evolution of social behaviour. 31 O’Neill B (2014) Policy folklists and evolutionary theory. Proc Natl of evolution V: Cooperation and conflict. Proc Natl Acad Sci USA I. J Theor Biol 7(1):1–16. Acad Sci USA 111:10854–10859. 108(Suppl 2):10787–10791. 19 Hamilton WD (1971) Man and Beast, eds Eisenberg JF, Dillon WS 32 Bergstrom TC (2014) On the evolution of hoarding, risk-taking, 6 Striedter GF, Avise JC, Ayala FJ (2012) In the light of evolution VI: (Smithsonian Institution Press, Washington), pp 59–91. and wealth distribution in nonhuman and human populations. Proc Brain and behavior. Proc Natl Acad Sci USA 109(Suppl): 20 Price GR (1970) Selection and covariance. Nature 227(5257): Natl Acad Sci USA 111:10860–10867. 10607–10611. – 520 521. 33 Morgenstern Y, Rostami M, Purves D (2014) Properties of 7 Cela-Conde CJ, Gutiérrez Lombardo R, Avise JC, Ayala FJ (2013) In 21 Binmore K (2014) Bargaining and fairness. Proc Natl Acad Sci artificial networks evolved to contend with natural spectra. Proc Natl the light of evolution VII: The human mental machinery. Proc Natl USA 111:10785–10788. – Acad Sci USA 111:10868–10872. Acad Sci USA 110(Suppl 2):10339 10342. 22 Komarova NL (2014) Spatial interactions and cooperation 8 34 Huttegger S, Skyrms B, Tarrès P, Wagner E (2014) Some dynamics Darwin C (1871) The Descent of Man and Selection in Relation to can change the speed of evolution of complex phenotypes. Proc of signaling games. Proc Natl Acad Sci USA 111:10873–10880. Sex (John Murray, London). Natl Acad Sci USA 111:10789–10795. 35 Kreindler GE, Young HP (2014) Rapid innovation diffusion in 9 Darwin C (1874) The Descent of Man and Selection in Relation to 23 Christakis NA, Fowler JH (2014) Friendship and natural selection. – Sex (John Murray, London), 2nd Ed. Proc Natl Acad Sci USA 111:10796–10801. social networks. Proc Natl Acad Sci USA 111:10881 10888. 36 10 Maynard Smith J (1982) Evolution and the Theory of Games 24 Kandori M, Obayashi S (2014) Labor union members play an Jackson MO, Xing Y (2014) Culture-dependent strategies in (Cambridge Univ Press, Cambridge, UK). OLG repeated game. Proc Natl Acad Sci USA 111:10802–10809. coordination games. Proc Natl Acad Sci USA 111:10889–10896. 11 von Neumann J, Morgenstern Y (1944) Theory of Games and 25 Cressman R, Tao Y (2014) The replicator equation and other 37 Dobzhansky T (1973) Nothing in biology makes sense except in Economic Behavior (Princeton Univ Press, Princeton). game dynamics. Proc Natl Acad Sci USA 111:10810–10817. the light of evolution. Am Biol Teach 35(3):125–129.

10784 | www.pnas.org/cgi/doi/10.1073/pnas.1411483111 Skyrms et al. Downloaded by guest on September 30, 2021