Simulpast may v2
- 1. Social Mechanisms, Agent-Based Modelling and the Evolution of Cooperation Mauricio Salgado PhD Analytical Sociology and Institutional Design Group (GSADI) – UniversitatAutònoma de Barcelona Mauricio.Salgado@uab.cat Barcelona – May 2013 1
- 3. Why cooperation? 3 Definitions Cooperation: when two or more individuals engage in costly joint actions that result in mutual benefit Altruism (Unconditional Cooperation): Acts that benefit others at a personal cost Donor (c) Recipient (b)
- 5. Why cooperation? 5 Ingrid Loyau-Kennett (48) ―Being a cub leader I have my first aid so when I saw this guy on the floor I thought it was an accident then I saw the guy was dead‖ ―And then when I went up there was this black guy with a revolver and a kitchen knife (…) and he said 'move off the body‘.‖ (…) ―I started to talk to him and I started to notice more weapons and the guy behind him with more weapons as well. By then, people had started to gather around. So I thought OK, I should keep him talking to me before he noticed everything around him.‖
- 6. Why cooperation? 6 Cooperation is what makes us humans
- 7. Why cooperation? 7 ―If you try to do something cooperative with a chimp—point out something, show them where some food is—their attention wanders all over the place, but if you compete with them over food, they are zeroed in like a laser. All their cognitive skills are on.‖ Michel Tomasello But we still have to explain how did we go from ―chimp-like-selfishness‖ to ―human super-cooperation‖
- 8. Why cooperation? The Higgs boson of evolutionary anthropology: How did we get that way? 8
- 10. Social Mechanisms 10 Two main advantages: 1. Knowledge about mechanisms increases the possibility of causal analysis in the absence of nomological laws, and 2. it helps to open the ‗black box‘ of social dynamics in order to provide the microfoundations of the observed phenomena.
- 11. Social Mechanisms 11 Definition A mechanism consists of entities and activities, organized such that they are productive of regular changes from start or set-up to finish or termination conditions. Entities (and their properties) are the things that engage in activities, and activities are the producers of change. The dualistic nature of this definition makes it adequate for social sciences that deal with individual agents (‗entities‘) and their actions (‗activities‘).
- 12. ―Mechanisms consist of entities (with their properties) and the activities that these entities engage in, either by themselves or in concert with other entities. These activities bring about change, and the type of change brought about depends on the properties of the entities and how the entities are organized spatially and temporally‖ Hedström, P. Dissecting the Social (2005) Opening the black box Social Mechanisms 12
- 13. Social Mechanisms 13 1. A set of entities, e1 to en, vary with respect to a property pi 2. In a certain environment the benefit, b1, bestowed on entity i is a function of its property p1 3. The value of bi, influences the relative frequency of entities with different properties at subsequent time periods. (Properties that perform better in the environment will become more frequent) Selection Mechanisms
- 14. Social Mechanisms 14 Evolutionary game dynamics describe: • Frequency-dependent selection. • The outcome of the game is related to reproductive success. • Payoff determines fitness. • Reproduction can be genetic or cultural. Genetic Reproduction It means that individuals leave genetic offspring that inherit their strategy. Cultural Reproduction It means that individuals are imitated by others; thereby strategies reproduce by imitation or learning.
- 15. 15 Ontology Is the mechanisms made of ‘real’ entities and activities or is it just an as-if story? Regularity In which instances can the mechanisms be generalized from one case to another? Transparency Is the presence of the mechanism clearly observable? Intelligibility Dos the presence of the mechanism improve the intelligibility of the explanation? Reduction Base What are the mechanism ‘microfoundations’, the local rules that bring about the ‘macro-properties’? Generative Sufficiency Can the model’s microfundations generate the observed macroproperties? Empirical Adequacy Are the model’s microfundationsplausible in relation to some theory or empirical data?
- 16. 16 The Evolution of Cooperation
- 17. Evolution of Cooperation 17 Alter Ego Cooperate Defect CooperateDefect + + - + + - - - Prisoner‘s Dilemma Alter Ego Cooperate (Quiet) Defect (Fink) Cooperate (Quiet) Defect (Fink) 3 , 3 0 , 5 5 , 0 1 , 1
- 18. Evolution of Cooperation 18 The Stag Hunt Alter Ego Cooperate Defect CooperateDefect + + - + + - - - Alter Ego Cooperate (Stag) Defect (Hare) Cooperate (Stag) Defect (Hare) 2 , 2 0 , 1 1 , 0 1 , 1 Jean-Jacques Rousseau No temptation to defect
- 19. Evolution of Cooperation 19 C C C C C C C C C C C DMutation C C D C D DSelection D D D D D DSelection Declining Average Fitness Without any mechanism for the evolution of cooperation, natural selection favours the evolution of defection
- 20. Evolution of Cooperation 20 Kin Selection Direct Reciprocity Indirect Reciprocity Group Selection Parochial Altruism Five Mechanisms
- 21. 21 Kin Selection
- 22. Kin Selection 22 Agents cooperate only with those related genetically ―I will jump into the river to save two brothers or eight cousins‖ C DD First Generation C CC N Generation Selection
- 23. Kin Selection 23 Definition The beneficiaries share some genes with the altruist; the survival and reproduction of the beneficiaries contribute to the propagation of the altruist’s genes. Ontology Realist Regularity It works in small familial groups Transparency Yes Intelligibility Yes Reduction Base Individuals’ genes Generative Sufficiency Yes Empirical Adequacy Although it might work in small and familial groups with little migration, it does not explain cooperation to unrelated individuals (commitment falls off precipitously as genetic distance increases between individuals)
- 24. Kin Selection 24 In meerkats, helpers vary widely in the number of food items they gave to pups. The level of contributions that helpers make to rearing pups is not significantly correlated with variation in kinship to the litters they are rearing. (Clutton-Brock et al., 2001) In this picture, a male golden lion tamarin at Colchester Zoo (The United Kingdom) took the unusual step of carrying other primates‘ offspring that he shares his enclosure with. This tamarin male has taken on the role of carrying for two silvery marmoset youngsters. Limitations: It cannot explain cooperative breeding
- 26. Direct Reciprocity 26 Agents play the Iterated Prisoner's Dilemma (IPD) ―You scratch my back and I'll scratch yours‖ Tit for Tat Interactions Alter Ego T1 T2
- 27. Direct Reciprocity 27 Definition Both the beneficiaries and the cooperators face repeated encounters; the cooperator’s decision to cooperate is based on what the beneficiary has done to her in previous encounters Ontology Realist Regularity It works in contexts of repeated encounters within small groups Transparency Yes Intelligibility Yes Reduction Base Individuals in repeated interactions Generative Sufficiency Yes Empirical Adequacy Although it might work in small groups, it is a weak mechanism in larger groups (ability to directly monitor trustworthiness in reciprocation decreases rapidly as the number of transactions multiply)
- 28. Direct Reciprocity 28 Limitations: • It cannot explain short-term or one-shot cooperative interactions in games such as ―The Ultimatum Game‖ or ―The Dictator Game‖ • It cannot explain the production of ―public goods‖ • In natural populations, errors occur: occasional mistakes between two TFT players cause long runs of mutual backbiting. C C C C C C D D D D D TFT TFT‘ Error! C …
- 30. Indirect Reciprocity 30 Reputation ―I help you and somebody will help me‖ D R W Ct1 D R WCt1+n D R W Dt1 Dt1+n Donor‘s reputation increases Donor‘s reputation decreases
- 31. Indirect Reciprocity 31 Definition Both the beneficiaries and the altruists interact with each other only occasionally; the altruist’s decision to cooperate is based on information about what the beneficiary has done to others in previous encounters Ontology Realist Regularity It works when information about the beneficiaries’ reputation is available Transparency Yes Intelligibility Yes Reduction Base Individuals in occasional interactions Generative Sufficiency Yes Empirical Adequacy Although it might work in larger groups, it requires dense connections or cultural artifacts that register individuals’ reputation
- 32. Indirect Reciprocity 32 ―Indirect reciprocity is a plausible mechanism to sustain cooperation within dense, bounded social networks that are stable through time‖ Limitations: Heroism in warfare?
- 34. Group Selection 34 Multilevel Model ―I help you (… and my group will be better)‖ Declining Average Fitness (Extinction)
- 35. Group Selection 35 Definition The altruists’ genes can become fixed within certain groups because of the benefits they bestow on those groups as wholes, even when the effect of these alleles on individuals’ fitness is negative. These groups reproduce faster Ontology Realist Regularity It works when the group benefit of the altruists’ actions is higher than the altruists’ individual costs Transparency Yes Intelligibility Yes Reduction Base Individuals’ genes within groups that compete with each other Generative Sufficiency Yes Empirical Adequacy Although it might work in small or medium-size groups, it requires within- group homogeneity, between group heterogeneity and low migration rates
- 36. Group Selection 36 ―Most human populations do not meet these conditions. Thus, it does not seem that group selection can provide a genetic basis for altruism (…) Group selection mechanism depends on the relative amount of variation within and between groups. If group members are closely related, most of the variation will occur between groups. This is easiest to see if groups are composed of clones (as in colonial invertebrates such as corals). Then there is almost no genetic variation within groups; all the variation is between groups, and selection acts to maximise group benefit.‖ Limitations:
- 38. Parochial Altruism 38 Multilevel Model ―I help you because you belong to my group (… and, therefore, my group will be stronger)‖ Declining Average Fitness (Extinction) Intergroup Conflict
- 39. Parochial Altruism 39 Definition The altruists cooperate with group members and are hostile to individuals from other groups. In contexts of intergroup conflict, internally cooperative groups prevail over less cooperative rival groups Ontology Realist Regularity It works when there is severe intergroup conflict Transparency Yes Intelligibility Yes Reduction Base Individuals within groups that compete with each other Generative Sufficiency Yes Empirical Adequacy Although it might work in contexts of violent intergroup conflict, it does not apply in contexts in which groups cooperate among them
- 40. Parochial Altruism 40 ―However, such generalisations are unlikely to correctly describe the conditions in which our Pleistocene ancestors lived, so parochial altruism seems to be implausible. Small bands of hunter-gatherers, numbering 25 or so individuals, under chronic climate fluctuation, widely dispersed over large areas and unable to fall back on staple foods, would have suffered from high mortality rates, particularly child mortality, due to starvation as well as predation and disease, so they hardly would have been able to sustain warfare against competing groups ‖ Limitations:
- 42. Concluding Remarks 42 Explaining the evolution of cooperation requires the identification of some mechanism at work A mechanism consists of entities and activities, organized such that they are productive of regular changes from start or set-up to finish or termination conditions. Entities (and their properties) are the things that engage in activities, and activities are the producers of change. Ontology, Regularity, Transparency, Intelligibility, Reduction Base, Generative Sufficiency, Empirical Adequacy
- 43. Concluding Remarks Five Mechanisms for the Evolution of Cooperation • Kin Selection: Cooperate with genetic relatives • Direct Reciprocity: I help you, you help me • Indirect Reciprocity: I help you, somebody will help me • Group Selection: Groups of cooperators out-perform other groups • Parochial Altruism: Groups of cooperators out-compete other groups 43
- 44. Concluding Remarks 44 Punishment (e.g., ‗Strong Altruism‘) is not a mechanism for the evolution of cooperation Parsimony If you punish someone who has defected with you, then you are using direct reciprocity If you punish someone who has defected with others, then you are using indirect reciprocity If you punish someone who is not a member of your group, then you are using parochial altruism (a special case of group selection) (Winners do not punish)
- 45. Cooperators • José Antonio Noguera • Francisco J. Miguel 45
- 46. Thank you! Mauricio Salgado PhD Analytical Sociology and Institutional Design Group (GSADI) – UniversitatAutònoma de Barcelona Mauricio.Salgado@uab.cat Barcelona – May 2013 46
Editor's Notes
- #5: Tales about the origins of our species always start off like this: A small band of hunter-gatherers roams the savannah, loving, warring, and struggling for survival under the African sun. They do not start like this: Brave mom Ingrid Loyau-Kennett faced Woolwich terrorists.
- #39: Parochial altruism has gained recent popularity among scholars: these models highlight the idea that individuals are altruistic with group members and are hostile to individuals not of one’s own group (Arrow, 2007; Bowles, 2006; Choi and Bowles, 2007; Sa ̈äksvuori et al., 2011). Since internally cooperative groups prevail over less cooperative rival groups, parochial altruism rests on the evolutionary belief that violent intergroup conflict played a key role in the dawn of human cooperation.