Evolutionary Game Theory

Total Page:16

File Type:pdf, Size:1020Kb

Evolutionary Game Theory Magazine R503 ecology. If prey is abundant, predators cost of helping an unrelated Primer will increase for a while. But this individual? Suppose that the benefit increase reduces the abundance of (to the recipient) is b and the cost (to prey, and eventually leads to a the donor) is c, with c < b. If an Evolutionary game decrease of the predators. individual is equally likely to be theory potential recipient or donor in a Hawks and doves given encounter, then a population Karl Sigmund* and Intraspecific fights provide a first of cooperators would earn, on † example of changes in a population average, b – c per interaction, and be Martin A. Nowak that are dependent on the frequency better off than a population of of a trait. Assume that there are two defectors earning 0. But no matter Ever since Darwin read Malthus, the behaviourally distinct morphs: how the population is composed, an theory of evolution has benefited ‘hawks’ escalate a fight until the individual would always increase its from the interaction of ecology with injury of one contestant settles the economics. Evolutionary game theory outcome, ‘doves’ stick to some belongs to this tradition: it merges conventional display (a pushing The hawk–dove game population ecology with game theory. match, for instance, where injuries Game theory originally addressed are practically excluded) and take In the hawk–dove game, G is the gain in problems confronted by decision flight if the adversary escalates. If fitness resulting from winning the contest, and C is the cost in fitness due makers with diverging interests (for most contestants are doves, the hawk to an injury (see Figure 1). instance, firms competing for a morph will spread; but if most market). The ‘players’ have to choose contestants are hawks, escalating a Figure 1 between strategies whose payoff conflict will lead with probability ½ depends on their rivals’ strategies. to injury (see green box and Figure 1). Type of adversary This interdependence leads to a Even this oversimplified example Hawk Dove mutual ‘outguessing’, as with chess shows that heavily armed species (she thinks that I think that she (where the risk of injury is large) are (G-C)/2 G thinks…). There usually is no particularly prone to conventional Hawk solution that is unconditionally displays (see Figure 2). wins payoff 0 G/2 optimal, no matter what the A wide variety of behavioural Dove Contestant who co-players are doing. traits — and purely morphological or Current Biology In the context of evolutionary physiological characters, like the biology, the two basic notions of height of trees or the length of The payoff matrix for the hawk–dove game. game theory, namely strategy and antlers — are submitted to payoff, have to be re-interpreted. A frequency-dependent selection and strategy is not a deliberate course of are amenable to game analysis. Such No morph is unconditionally better than the other. If the object of the fight is action, but an inheritable trait; payoff traits may influence conflicts of not worth the injury, then the dove morph is Darwinian fitness (average interest between two individuals, for will spread. Hawks can spread only if reproductive success). The ‘players’ instance, territorial disputes (between their frequency is below G/C. If G is less are members of a population, all neighbours), the length of the than C, a mixed population of hawks and doves will evolve. It is conceivable that competing for a larger share of weaning period (between parents and some phenotype plays a mixed strategy, descendants. offspring), or the division of parental escalating only with a certain frequency. If several variants of a trait occur investment (between male and Such a mixed strategy escalating with in a population, then natural selection female). But frequency-dependent probability G/C is evolutionarily stable; no mutant with a different propensity to leads to an increase in the frequency selection also occurs without escalate can invade. If there exists an of those variants with higher fitness. antagonistic encounters. The sex asymmetry between the two contestants If the success of a trait does not ratio is an example of this (if it is (larger versus smaller, for instance, or depend on its frequency, this will biased towards males, it pays to owner versus intruder), this will alter the game. In such cases a conditional eventually lead to the fixation of the produce daughters, and vice versa). strategy will emerge which uses the optimal variant. But if the success of a asymmetry as a cue, for instance the trait is frequency-dependent, then its The prisoner’s dilemma so-called ‘bourgeois strategy’: if owner, increase may lead to a composition of The evolution of cooperation escalate; if intruder, display. Some the population where other variants through reciprocation is a particularly intraspecific conflicts use a long assessment phase to detect an do better; this can be analysed by extensive chapter of evolutionary asymmetry in, for instance, size, by means means of game theory. This is similar game theory. Why should a selfish of a pushing match or a parallel walk. to what happens in population gene, or ‘fitness maximiser’, bear the R504 Current Biology, Vol 9 No 14 fitness by refusing help, and hence Figure 2 further round is sufficiently high, we would not see cooperation. then the presence of even a small Game theorists have number of so-called ‘retaliators’ is encapsulated this tug-of-war enough to favour cooperation. The between common good and selfish best known example of such a interest in the so-called prisoner’s retaliatory strategy is ‘tit-for-tat’. A dilemma game. In this game, a player tit-for-tat player cooperates in the has two possible strategies C (to first round and from then on always cooperate) and D (to defect). Two C repeats whatever the co-player did in players will get a reward R which is the previous round. higher than the punishment P After promoting the emergence A clash between male red deer. Only rarely obtained by two D players. But a D does the fight escalate beyond a pushing of cooperators, retaliatory strategies player exploiting a C player obtains a match. (Photograph provided by Tim are often superseded by more payoff T (temptation to defect) Clutton-Brock, Department of Zoology, tolerant strategies, for instance which is higher than R, and this University of Cambridge, UK.) ‘generous tit-for-tat’ or ‘win–stay, leaves the C player with the sucker’s lose–shift’. A generous tit-for-tat payoff S, which is lower than P. So, reproduces inside a bacterium. In player cooperates after an opponent’s because T > R > P > S, a rational this instance, one phage variant defection with a certain probability, player will always play D, which is produces less of the intracellular whereas a win–stay, lose–shift player the better move no matter what the products needed for replication than repeats the previous move after co-player is doing. the other, and thus may be said to receiving a high payoff in a round, Many species engage in defect. The payoff values can be but otherwise switches to the other interactions which seem to be of the measured with some precision: they move. Both generous tit-for-tat and prisoner’s dilemma type. Vampire satisfy the rank ordering required for win–stay, lose–shift players return to bats feed each other, monkeys the prisoner’s dilemma. cooperation after an erroneous engage in allogrooming, birds and There are several ways in which defection, whereas tit-for-tat players vervet monkeys utter alarm calls, the prisoner’s dilemma can be do not. A great many theoretical guppies and stickleback cooperate in overcome. With the phages, selection results and computer simulations predator inspection, hermaphroditic for a particular trait-group operates show that under very general sea bass alternate as egg-spenders whenever the virus population is so conditions, populations of defectors and lions engage in cooperative small that most bacterial hosts are can be invaded by small groups of hunting or joint territorial defense. invaded by only one virus. A phage stern retaliators, who pave the way But often, attempts to specify the will then most likely interact only for more tolerant populations, which, payoff values of these behavioural with members of its own clone, and in turn, can eventually ‘soften up’ to interactions lead to doubts as to this is when cooperators are better such a degree that defectors may whether one is seeing a bona fide off than defectors. More generally, take over again. prisoner’s dilemma. It is difficult to any form of associative interaction The best examples of reciprocal measure the fitness of fish darting in favours cooperation. Such association strategies may be found in human and out of shoals or of monkeys can be due to kinship, to partner societies. But here, reciprocation is hiding in the bush. It may be that choice, to ostracism of defectors or often indirect. An act of assistance is some of these are simply instances of simply to spatial structure (or returned, not by the recipient, but by by-product mutualism, in which both ‘population viscosity’). a third party. Such indirect players are best served by cooperating reciprocation can be based on and none is tempted to defect, or an Repeated interactions score-keeping by ‘discriminate instance of the so-called snowdrift Among higher organisms, interactions altruists’ (who help only those game, in which the best reply to the of the prisoner’s dilemma type are individuals that have not refused co-player’s C is a D, but the best reply probably repeated between the same help too often).
Recommended publications
  • Evolutionary Game Theory: ESS, Convergence Stability, and NIS
    Evolutionary Ecology Research, 2009, 11: 489–515 Evolutionary game theory: ESS, convergence stability, and NIS Joseph Apaloo1, Joel S. Brown2 and Thomas L. Vincent3 1Department of Mathematics, Statistics and Computer Science, St. Francis Xavier University, Antigonish, Nova Scotia, Canada, 2Department of Biological Sciences, University of Illinois, Chicago, Illinois, USA and 3Department of Aerospace and Mechanical Engineering, University of Arizona, Tucson, Arizona, USA ABSTRACT Question: How are the three main stability concepts from evolutionary game theory – evolutionarily stable strategy (ESS), convergence stability, and neighbourhood invader strategy (NIS) – related to each other? Do they form a basis for the many other definitions proposed in the literature? Mathematical methods: Ecological and evolutionary dynamics of population sizes and heritable strategies respectively, and adaptive and NIS landscapes. Results: Only six of the eight combinations of ESS, convergence stability, and NIS are possible. An ESS that is NIS must also be convergence stable; and a non-ESS, non-NIS cannot be convergence stable. A simple example shows how a single model can easily generate solutions with all six combinations of stability properties and explains in part the proliferation of jargon, terminology, and apparent complexity that has appeared in the literature. A tabulation of most of the evolutionary stability acronyms, definitions, and terminologies is provided for comparison. Key conclusions: The tabulated list of definitions related to evolutionary stability are variants or combinations of the three main stability concepts. Keywords: adaptive landscape, convergence stability, Darwinian dynamics, evolutionary game stabilities, evolutionarily stable strategy, neighbourhood invader strategy, strategy dynamics. INTRODUCTION Evolutionary game theory has and continues to make great strides.
    [Show full text]
  • Econstor Wirtschaft Leibniz Information Centre Make Your Publications Visible
    A Service of Leibniz-Informationszentrum econstor Wirtschaft Leibniz Information Centre Make Your Publications Visible. zbw for Economics Banerjee, Abhijit; Weibull, Jörgen W. Working Paper Evolution and Rationality: Some Recent Game- Theoretic Results IUI Working Paper, No. 345 Provided in Cooperation with: Research Institute of Industrial Economics (IFN), Stockholm Suggested Citation: Banerjee, Abhijit; Weibull, Jörgen W. (1992) : Evolution and Rationality: Some Recent Game-Theoretic Results, IUI Working Paper, No. 345, The Research Institute of Industrial Economics (IUI), Stockholm This Version is available at: http://hdl.handle.net/10419/95198 Standard-Nutzungsbedingungen: Terms of use: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Documents in EconStor may be saved and copied for your Zwecken und zum Privatgebrauch gespeichert und kopiert werden. personal and scholarly purposes. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle You are not to copy documents for public or commercial Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich purposes, to exhibit the documents publicly, to make them machen, vertreiben oder anderweitig nutzen. publicly available on the internet, or to distribute or otherwise use the documents in public. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, If the documents have been made available under an Open gelten abweichend von diesen Nutzungsbedingungen die in der dort Content Licence (especially Creative Commons Licences), you genannten Lizenz gewährten Nutzungsrechte. may exercise further usage rights as specified in the indicated licence. www.econstor.eu Ind ustriens Utred n i ngsi nstitut THE INDUSTRIAL INSTITUTE FOR ECONOMIC AND SOCIAL RESEARCH A list of Working Papers on the last pages No.
    [Show full text]
  • Introduction to Game Theory I
    Introduction to Game Theory I Presented To: 2014 Summer REU Penn State Department of Mathematics Presented By: Christopher Griffin [email protected] 1/34 Types of Game Theory GAME THEORY Classical Game Combinatorial Dynamic Game Evolutionary Other Topics in Theory Game Theory Theory Game Theory Game Theory No time, doesn't contain differential equations Notion of time, may contain differential equations Games with finite or Games with no Games with time. Modeling population Experimental / infinite strategy space, chance. dynamics under Behavioral Game but no time. Games with motion or competition Theory Generally two player a dynamic component. Games with probability strategic games Modeling the evolution Voting Theory (either induced by the played on boards. Examples: of strategies in a player or the game). Optimal play in a dog population. Examples: Moves change the fight. Chasing your Determining why Games with coalitions structure of a game brother across a room. Examples: altruism is present in or negotiations. board. The evolution of human society. behaviors in a group of Examples: Examples: animals. Determining if there's Poker, Strategic Chess, Checkers, Go, a fair voting system. Military Decision Nim. Equilibrium of human Making, Negotiations. behavior in social networks. 2/34 Games Against the House The games we often see on television fall into this category. TV Game Shows (that do not pit players against each other in knowledge tests) often require a single player (who is, in a sense, playing against The House) to make a decision that will affect only his life. Prize is Behind: 1 2 3 Choose Door: 1 2 3 1 2 3 1 2 3 Switch: Y N Y N Y N Y N Y N Y N Y N Y N Y N Win/Lose: L W W L W L W L L W W L W L W L L W The Monty Hall Problem Other games against the house: • Blackjack • The Price is Right 3/34 Assumptions of Multi-Player Games 1.
    [Show full text]
  • Navigating the Landscape of Multiplayer Games ✉ Shayegan Omidshafiei 1,4 , Karl Tuyls1,4, Wojciech M
    ARTICLE https://doi.org/10.1038/s41467-020-19244-4 OPEN Navigating the landscape of multiplayer games ✉ Shayegan Omidshafiei 1,4 , Karl Tuyls1,4, Wojciech M. Czarnecki2, Francisco C. Santos3, Mark Rowland2, Jerome Connor2, Daniel Hennes 1, Paul Muller1, Julien Pérolat1, Bart De Vylder1, Audrunas Gruslys2 & Rémi Munos1 Multiplayer games have long been used as testbeds in artificial intelligence research, aptly referred to as the Drosophila of artificial intelligence. Traditionally, researchers have focused on using well-known games to build strong agents. This progress, however, can be better 1234567890():,; informed by characterizing games and their topological landscape. Tackling this latter question can facilitate understanding of agents and help determine what game an agent should target next as part of its training. Here, we show how network measures applied to response graphs of large-scale games enable the creation of a landscape of games, quanti- fying relationships between games of varying sizes and characteristics. We illustrate our findings in domains ranging from canonical games to complex empirical games capturing the performance of trained agents pitted against one another. Our results culminate in a demonstration leveraging this information to generate new and interesting games, including mixtures of empirical games synthesized from real world games. 1 DeepMind, Paris, France. 2 DeepMind, London, UK. 3 INESC-ID and Instituto Superior Técnico, Universidade de Lisboa, Lisboa, Portugal. 4These authors ✉ contributed equally: Shayegan Omidshafiei, Karl Tuyls. email: somidshafi[email protected] NATURE COMMUNICATIONS | (2020) 11:5603 | https://doi.org/10.1038/s41467-020-19244-4 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-19244-4 ames have played a prominent role as platforms for the games.
    [Show full text]
  • Evolutionary Game Theory: a Renaissance
    games Review Evolutionary Game Theory: A Renaissance Jonathan Newton ID Institute of Economic Research, Kyoto University, Kyoto 606-8501, Japan; [email protected] Received: 23 April 2018; Accepted: 15 May 2018; Published: 24 May 2018 Abstract: Economic agents are not always rational or farsighted and can make decisions according to simple behavioral rules that vary according to situation and can be studied using the tools of evolutionary game theory. Furthermore, such behavioral rules are themselves subject to evolutionary forces. Paying particular attention to the work of young researchers, this essay surveys the progress made over the last decade towards understanding these phenomena, and discusses open research topics of importance to economics and the broader social sciences. Keywords: evolution; game theory; dynamics; agency; assortativity; culture; distributed systems JEL Classification: C73 Table of contents 1 Introduction p. 2 Shadow of Nash Equilibrium Renaissance Structure of the Survey · · 2 Agency—Who Makes Decisions? p. 3 Methodology Implications Evolution of Collective Agency Links between Individual & · · · Collective Agency 3 Assortativity—With Whom Does Interaction Occur? p. 10 Assortativity & Preferences Evolution of Assortativity Generalized Matching Conditional · · · Dissociation Network Formation · 4 Evolution of Behavior p. 17 Traits Conventions—Culture in Society Culture in Individuals Culture in Individuals · · · & Society 5 Economic Applications p. 29 Macroeconomics, Market Selection & Finance Industrial Organization · 6 The Evolutionary Nash Program p. 30 Recontracting & Nash Demand Games TU Matching NTU Matching Bargaining Solutions & · · · Coordination Games 7 Behavioral Dynamics p. 36 Reinforcement Learning Imitation Best Experienced Payoff Dynamics Best & Better Response · · · Continuous Strategy Sets Completely Uncoupled · · 8 General Methodology p. 44 Perturbed Dynamics Further Stability Results Further Convergence Results Distributed · · · control Software and Simulations · 9 Empirics p.
    [Show full text]
  • Game Theory and Evolutionary Biology 1
    Chapter 28 GAME THEORY AND EVOLUTIONARY BIOLOGY PETER HAMMERSTEIN Max-Planck-Institut für Verhaltensphysiologie REINHARD SELTEN University of Bonn Contents 1. Introduction 931 2. Conceptual background 932 2.1. Evolutionary stability 932 2.2. The Darwinian view of natural selection 933 2.3. Payoffs 934 2.4. Game theory and population genetics 935 2.5. Players 936 2.6. Symmetry 936 3. Symmetric two-person games 937 3.1. Definitions and notation 937 3.2. The Hawk-Dove garne 937 3.3. Evolutionary stability 938 3.4. Properties ofevolutionarily stable strategies 940 4. Playing the field 942 5. Dynamic foundations 948 5.1. Replicator dynamics 948 5.2. Disequilibrium results 951 5.3. A look at population genetics 952 6. Asymmetric conflicts 962 * We are grateful to Olof Leimar, Sido Mylius, Rolf Weinzierl, Franjo Weissing, and an anonymous referee who all helped us with their critical comments. We also thank the Institute for Advanced Study Berlin for supporting the final revision of this paper. Handbook of Garne Theory, Volume 2, Edited by R.J. Aumann and S. Hart © Elsevier Science B.V., 1994. All rights reserved 930 P. Hammerstein and R. Selten 7. Extensive two-person games 965 7.1. Extensive garnes 965 7.2. Symmetric extensive garnes 966 7.3. Evolutionary stability 968 7.4. Image confrontation and detachment 969 7.5. Decomposition 970 8. Biological applications 971 8.1. Basic qnestions about animal contest behavior 972 8.2. Asymmetric animal contests 974 8.3. War of attrition, assessment, and signalling 978 8.4. The evolution of cooperation 980 8.5.
    [Show full text]
  • Game Theory: Penn State Math 486 Lecture Notes
    Game Theory: Penn State Math 486 Lecture Notes Version 1.1.1 Christopher Griffin « 2010-2011 Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 United States License With Major Contributions By: James Fam George Kesidis Contents List of Figuresv Chapter 1. Preface and an Introduction to Game Theory xi 1. Using These Notes xi 2. An Overview of Game Theory xi Chapter 2. Probability Theory and Games Against the House1 1. Probability1 2. Random Variables and Expected Values6 3. Conditional Probability7 4. Bayes Rule 12 Chapter 3. Utility Theory 15 1. Decision Making Under Certainty 15 2. Advanced Decision Making under Uncertainty 23 Chapter 4. Game Trees, Extensive Form, Normal Form and Strategic Form 25 1. Graphs and Trees 25 2. Game Trees with Complete Information and No Chance 28 3. Game Trees with Incomplete Information 32 4. Games of Chance 35 5. Pay-off Functions and Equilibria 37 Chapter 5. Normal and Strategic Form Games and Matrices 47 1. Normal and Strategic Form 47 2. Strategic Form Games 48 3. Review of Basic Matrix Properties 50 4. Special Matrices and Vectors 52 5. Strategy Vectors and Matrix Games 53 Chapter 6. Saddle Points, Mixed Strategies and the Minimax Theorem 57 1. Saddle Points 57 2. Zero-Sum Games without Saddle Points 60 3. Mixed Strategies 63 4. Mixed Strategies in Matrix Games 66 5. Dominated Strategies and Nash Equilibria 66 6. The Minimax Theorem 71 7. Finding Nash Equilibria in Simple Games 76 8. A Note on Nash Equilibria in General 79 iii Chapter 7. An Introduction to Optimization and the Karush-Kuhn-Tucker Conditions 81 1.
    [Show full text]
  • Evolutionary Game Theory
    Game Theory and Evolution 02-251 Spring 2019 Much of this material is derived from the book Evolutionary Dynamics by Martin A. Nowak. 1. Evolutionary Games When looking at genetic algorithms, our fitness was a function only of the individual. It didn't take into account the other individuals in the environment. This is unrealistic. Evolution can be a competitive or cooperative process. An bird evolves a beak that can reach particular insects, insects evolve a poison to avoid being eaten by the bird, etc. Cells in a cancer tumor evolve to avoid particular treatments. A virus evolves to avoid a vaccine. A bacteria evolves to not be susceptible to antibiotics. Organisms are evolving various strategies in games of competition and cooperation. This leads to the use of game theory to try to model observed evolutionary phenomena. As with our model of evolution based on NAND gates, these game-theoretic models are idealized and omit many aspects of real evolution. By doing this, we can hope to gain some understanding about fundamental evolutionary principles. \Game theory" was introduced by John von Neumann, a mathematician (and computer scientist before that was a thing). He was one of the architects of the first computers. He also coined the terms \transcription" and \translation". 1.1 Two-player games • A two-player is given by a payoff grid P where the rows and columns represent actions that the players can take. • In each cell of the grid, there are two numbers rij; cij. • There is a \row" player and a \column" player (hence the name \two player") • The row player chooses a row i, the column player chooses a column j, and the row player wins rij and the column player wins cij.
    [Show full text]
  • Evolutionary Game Theory Notes
    Evolutionary Game Theory Notes James Massey These notes are intended to be a largely self contained guide to everything you need to know for the evolutionary game theory part of the EC341 module. During lectures we will work through these notes, starting with a quick skim through Chapter 2 (which should be a revision of already known material for most of you) before moving on to Chapters 3 and 4 where we will spend most of our time. One notable thing missing from these notes are diagrams. You will find these very useful for making sense of these notes and should make sure to copy these down together with anything else covered in lectures that may help your understanding. These notes also contain many exercises, which I recommend you attempt. Separately I will also give you some further exercises and a sample exam question to give an idea of the kind of thing I might ask in an exam. Good luck! Chapter 1 Overview While non-cooperative game theory assumes hyper-rational agents predicting and best responding to each others’ play to lead to equilibria, evolutionary game theory takes a somewhat different approach. Here we dispose with such assumptions in favour of a milder assumption of evolutionary pressure towards the increasing representation of better strategies and study the re- sults.1. Take for example the question of which side of the road to drive a car (absent any laws which solve the problem). If two drivers are heading in opposite directions towards each other, we can identify two mutually ad- vantageous outcomes: where they pass each other with both driving on the left; and where they pass each other with both driving on the right.
    [Show full text]
  • Temporal and Spatial Effects Beyond Replicator Dynamics
    Evolutionary game theory: Temporal and spatial effects beyond replicator dynamics Carlos P. Roca a,∗, José A. Cuesta a, Angel Sánchez a,b,c a Grupo Interdisciplinar de Sistemas Complejos (GISC), Departamento de Matemáticas, Universidad Carlos III de Madrid, Avenida de la Universidad 30, 28911 Leganés, Madrid, Spain b Instituto de Biocomputación y Física de Sistemas Complejos (BIFI), Universidad de Zaragoza, Corona de Aragón 42, 50009 Zaragoza, Spain c Instituto de Ciencias Matemáticas CSIC-UAM-UC3M-UCM, 28006 Madrid, Spain Abstract Evolutionary game dynamics is one of the most fruitful frameworks for studying evolution in different disciplines, from Biology to Economics. Within this context, the approach of choice for many researchers is the so-called replicator equation, that describes mathematically the idea that those individuals performing better have more offspring and thus their frequency in the population grows. While very many interesting results have been obtained with this equation in the three decades elapsed since it was first proposed, it is important to realize the limits of its applicability. One particularly relevant issue in this respect is that of non-mean- field effects, that may arise from temporal fluctuations or from spatial correlations, both neglected in the replicator equation. This review discusses these temporal and spatial effects focusing on the non-trivial modifications they induce when compared to the outcome of replicator dynamics. Alongside this question, the hypothesis of linearity and its relation to the choice of the rule for strategy update is also analyzed. The discussion is presented in terms of the emergence of cooperation, as one of the current key problems in Biology and in other disciplines.
    [Show full text]
  • Types of Evolutionary Stability and the Problem of Cooperation
    Proc. Natl. Acad. Sci. USA Vol. 92, pp. 3596-3600, April 1995 Evolution Types of evolutionary stability and the problem of cooperation (evolution of cooperation/evolutionary game theory/evolutionarily stable strategies/iterated prisoner's dilemma/tit for tat) JONATHAN BENDORt AND PIOTR SWISTAKt tGraduate School of Business, Stanford University, Stanford, CA 94305-5015; and *Department of Government and Politics, University of Maryland, College Park, MD 20742-8221 Communicated by Robert Axelrod, University of Michigan, Ann Arbor, MI, November 14, 1994 ABSTRACT The evolutionary stability ofcooperation is a choices, to cooperate or defect; the same payoff matrix de- problem of fundamental importance for the biological and scribes the outcomes of these choices in all periods. The players social sciences. Different claims have been made about this observe each others' choices. A pure strategy in a repeated issue: whereas Axelrod and Hamilton's [Axelrod, R. & Ham- game between two players specifies an action for one player in ilton, W. (1981) Science 211, 1390-1398] widely recognized a current period as a function of the two players' past sequence conclusion is that cooperative rules such as "tit for tat" are of moves. A mixed strategy is a probability distribution on a set evolutionarily stable strategies in the iterated prisoner's of pure strategies. In dynamical analyses, it is assumed that dilemma (IPD), Boyd and Lorberbaum [Boyd, R. & Lorber- strategies that obtain higher payoffs reproduce at a higher rate. baum, J. (1987) Nature (London) 327, 58-59] have claimed (See ref. 7 for a discussion of the standard model and of its that no pure strategy is evolutionarily stable in this game.
    [Show full text]
  • Real World Games Look Like Spinning Tops
    Real World Games Look Like Spinning Tops Wojciech Marian Czarnecki Gauthier Gidel Brendan Tracey Karl Tuyls DeepMind DeepMind DeepMind DeepMind London London London Paris Shayegan Omidshafiei David Balduzzi Max Jaderberg DeepMind DeepMind DeepMind Paris London London Abstract This paper investigates the geometrical properties of real world games (e.g. Tic-Tac- Toe, Go, StarCraft II). We hypothesise that their geometrical structure resembles a spinning top, with the upright axis representing transitive strength, and the radial axis representing the non-transitive dimension, which corresponds to the number of cycles that exist at a particular transitive strength. We prove the existence of this geometry for a wide class of real world games by exposing their temporal nature. Additionally, we show that this unique structure also has consequences for learning – it clarifies why populations of strategies are necessary for training of agents, and how population size relates to the structure of the game. Finally, we empirically validate these claims by using a selection of nine real world two-player zero-sum symmetric games, showing 1) the spinning top structure is revealed and can be easily reconstructed by using a new method of Nash clustering to measure the interaction between transitive and cyclical strategy behaviour, and 2) the effect that population size has on the convergence of learning in these games. 1 Introduction Game theory has been used as a formal framework to describe and analyse many naturally emerging strategic interactions [30, 10, 9, 28, 20, 11, 6]. It is general enough to describe very complex interactions between agents, including classic real world games like Tic-Tac-Toe, Chess, Go, and modern computer-based games like Quake, DOTA and StarCraft II.
    [Show full text]