Coevolution of Cooperation, Causal Cognition and Mindreading

Coevolution of cooperation, causal cognition and mindreading

H. Clark Barrett,1,* Leda Cosmides2 and John Tooby3 1Center for Behavior, Evolution and Culture; Department of Anthropology; University of California; Los Angeles, CA USA; 2Center for Evolutionary Psychology; Department of Psychology and 3Anthropology; University of California; Santa Barbara, CA USA

he evolution of cooperation between On the other hand, models of the evo- Tunrelated individuals has long been lution of cooperation via reciprocity (the a puzzle in evolutionary biology. Formal trading of cooperative favors between models show that reciprocal altruism unrelated individuals) show that it is is approximately as stable as kin-based about as easy to evolve as cooperation via altruism when cooperators can assort. kin selection, if cooperative individuals are Why, then, is reciprocal altruism so rare? able to assort with like-mindedly coopera- We suggest that the key lies in the dif- tive individuals, excluding cheaters from ficulty of assortment based on underly- illicitly reaping the benefits of cooperation ing intentions: if individuals are able to and destabilizing the system (reviewed in reliably detect others’ cooperative intent ref. 1–3). And yet, cooperation between then cooperation is stable, but detect- unrelated individuals is thought to be ing intentions is notoriously difficult, far rarer in the biological world than especially when there are incentives to cooperation between kin.4 This suggests deceive. For this reason, we suggest, that the “assortment problem” is, in fact, there is likely to be a coevolutionary rela- what makes the evolution of cooperation tionship between human cooperativeness difficult. and our skills of social causal cognition; From a mathematical perspective, not it is not a coincidence that we are both only are the assortment problems in kin extraordinarily social, cooperating with selection and reciprocal altruism identi- non-kin to a degree not seen in other spe- cal, they also pose similar informational cies and extraordinarily good at inferring problems in that both require the infer- others’ beliefs, intentions and motiva- ence or detection of a hidden property by tions, a skill sometimes known as min- prospective cooperative partners: shared dreading. We discuss results of a recent genes in the case of kin selection, and Key words: evolution, cooperation, cheater study that provides evidence for this shared cooperative intent (or design) in detection, free-riding, mindreading coevolutionary view of cooperation and the case of reciprocal altruism. It is here, social cognition. however, where the resemblance might Submitted: 06/05/10 end. One reason that kin-based coopera- Accepted: 06/06/10 Why are humans so astonishingly success- tion is thought to be so common is that, Previously published online: ful at sociality? The evolution of coopera- as Hamilton5 pointed out, descent from a www.landesbioscience.com/journals/cib/ tion has been a long-standing puzzle in the common ancestor is a highly statistically article/12604 study of human evolution and in evolution- reliable cue to probability of shared genes, ary biology more generally, because coop- and biologists have discovered several DOI: 10.4161/cib.3.6.12604 eration suffers from a serious evolutionary reliable means of inferring shared ances- *Correspondence to: Clark Barrett; stability problem: the fitness incentives to try, such as direct observation of birth by Email: [email protected] reap the benefits of others’ cooperative- mothers, experience of parental care by ness while shirking oneself—also known offspring, and coresidence during child- Addendum to: Cosmides L, Barrett HC, Tooby J. 6 Adaptive specializations, social exchange, and as “cheating” or “free-riding”—are high. hood by siblings. In contrast, the internal the evolution of human intelligence. Proc Natl This means that in the absence of mecha- states that must be detected for coopera- Acad Sci USA 2010; 107:9007–14; PMID: 20445099; nisms for preventing cheating, cooperation by assortment to be stabilized—dis- DOI: 10.1073/pnas.0914623107. tion quickly unravels.1 positions, motivation structures and

522 Communicative & Integrative Biology Volume 3 Issue 6 Autophagic Punctum article addendum

intentions to cooperate—are notoriously empirically. Evolutionary game theory are inferred to have access to information difficult to reliably infer. models, for example, typically assume an that would allow them to manipulate a The informational difficulty of figuring extremely impoverished cognitive inter- cooperative situation in their own inter- out what another individual intends to do face for “mindreading,” involving simple est. We varied this factor independently of in strategic social contexts, including both rules that look only at a few prior instances “benefit,” by creating a scenario in which cooperation and strategic competition, has of an individual’s past cooperative behav- potential cheaters could be thwarted by long been recognized in the study of the ior (as opposed to, e.g., inferring their the use of code numbers to identify chil- evolution of intelligence.7-9 Intentions can incentive structures based on individual dren. Removal of ability to cheat reduced be both concealed and faked, and when characteristics or positions in a social net- vigilance for cheating by ~20 percentage there is an incentive to do so—as in the work or using other cues to intent or moti- points. case of free-riding—there are formidable vation). In psychology, on the other hand, Finally, we investigated a factor directly informational barriers to the evolution mindreading is typically studied as a skill related to theory of mind: intent to cheat. of reliable intention detectors. Unlike in isolation, without asking in what socio- Again, we varied this factor independently kinship detection, it is likely that there ecological contexts it is most likely to have of the other two factors, by explicitly cue- are no simple magic bullets for detecting evolved, nor how it is likely to interact by ing subjects to the possibility of either the internal states of others that matter design with other cognitive mechanisms, intentional or accidental rule-breaking. for cooperation. Instead, the evolution- such as mechanisms that stabilize coop- While some accounts of cooperation might ary pathway to cooperative intelligence is eration (reviewed in refs. 11 and 12). suggest that accidental and intentional probably long, involving many steps, each We recently conducted a study13 that rule-breaking are equally significant from of which incrementally improves organ- explicitly investigated the relationship a fitness point of view because they have isms’ ability to react strategically to the between cognitive mechanisms for sta- equal effects in the world, an “assortment” social world via progressively more fine- bilizing cooperation—in particular, the view of cooperation suggests that what is tuned capacities to detect and represent its cheater detection mechanism14—and most important is identifying the underly- hidden causal structure. mechanisms of social causal cognition. ing motives, in order to sort cheaters from Given these considerations, it is per- Using the Wason selection task, a standard non cheaters. Consistent with this pro- haps not a coincidence that humans are method of experimentally studying cheater posal, we found vigilance for cheating to both hyper-social and that we possess a detection, we investigated three aspects of be substantially higher in the intentional suite of cognitive capacities that appear social causal cognition that we expected to than the accidental conditions, again, by well-suited to inferring and representing interact with vigilance for cheaters in coop- about ~20 percentage points. Performance the hidden causal structure of the social erative contexts, using a standard scenario was highest when all three factors were world and of other people as social actors. in which school volunteers are sorting chil- present. One such set of capacities, sometimes dren into good and bad school districts These results provide support for the called “mindreading” or “theory of mind,” and have the potential to cheat by giving idea that the evolution of cooperation is widely thought to be an evolved adap- children illicit access to better schools than requires, or is facilitated by, the coevolu- tation for inferring the hidden internal their parents have paid for. tion of mechanisms that make assort- states of others, including their knowl- The first factor we investigated, which ment between like-minded individuals edge, motivations and intentions.10 More we called “benefit,” refers to the ability to possible. Moreover, they provide support broadly, we appear to be uniquely good compute others’ incentive structures in for the proposal that assortment prob- at causal cognition and especially, social cooperative contexts: their personal incen- lems are solved not by a single cognitive causal cognition: we devote substantial tive to cheat or free ride (i.e., the degree mechanism, but many. In our study, infer- mental resources to figuring out why other to which they stand to benefit from doing ences of intention, inferences of incentive people do things, representing their social so). We investigated this by activating an structure and inferences of informational relationships, assessing their skills, com- evolved psychology of kinship: we pre- access all had independent effects on the puting their motivational structures and dicted that if subjects knew that another capacity to detect cheating. In fact, we sus- so on. individual could benefit his or her own pect that this is just the tip of the iceberg: We suggest that this is not a coinci- offspring by free-riding, they would be a large number of cognitive capacities dence: human cooperativeness and our more vigilant for cheating, as measured by likely coevolved because of their syner- ability to solve free-rider problems have turning over exactly those cards necessary gistic effects in enabling human coop- likely coevolved with our capacities for to detect cheating in the Wason selection eration and sociality. We mean this not social causal cognition and theory of mind. task. This is what we found, with vigilance just in the narrow sense of tit-for-tat reci- While this potential evolutionary synergy for cheating increasing by ~20 percentage procity, but in all the ways that humans has long been recognized in the concept points when the potential for nepotistic can benefit through cooperative gains in of Machiavellian intelligence, in practice benefits could be inferred. trade, including the evolution of language, the relationships between cooperation, Second, we measured an aspect of with its reliance on relationships between causal cognition and theory of mind have social causal cognition which we called communicative intent and informational been little-studied, both theoretically and “Ability:” the degree to which individuals validity,15 and the evolution of other social

www.landesbioscience.com Communicative & Integrative Biology 523 References 11. Ermer E, Guerin SA, Cosmides L, Tooby J, Miller learning and cultural transmission mech- MB. Theory of mind broad and narrow: reasoning anisms, especially ones designed for sys- 1. Axelrod R, Hamilton WD. The evolution of coopera- about social exchange engages ToM areas, precau- tion. Science 1981; 211:1390-6. 16-18 tionary reasoning does not. Soc Neurosci 2006; tematic information sharing. 2. Boyd R, Richerson PJ. The evolution of reciprocity in 1:196-219. We regard this synergistic, coevo- sizeable groups. J Theor Biol 1988; 132:337-56. 12. McCabe KA, Smith VL, LePore M. Intentionality lutionary approach to the evolution of 3. Joshi NV. Evolution of cooperation by reciprocation detection and “mindreading”: Why does game form within structured demes. J Genetics 1987; 66:69-84. matter? Proc Natl Acad Sci USA 2000; 97:4404-9. mental specializations as quite different 4. Hammerstein P. Why is reciprocity so rare in social 13. Cosmides L, Barrett HC, Tooby J. Adaptive spe- than approaches that attempt to identify animals? A Protestant appeal. In: Hammerstein P, cializations, social exchange and the evolution of a single factor or capacity as the explana- Eds. Genetic and Cultural Evolution of Cooperation. human intelligence. Proc Natl Acad Sci USA 2010; Cambridge, MA: MIT 2003; 83-94. 107:9007-14. tion for human uniqueness, as well as dif- 5. Hamilton WD. The genetical evolution of social 14. Cosmides L. The logic of social exchange. Cognition ferent from approaches to “modularity” or behaviour, I. J Theor Biol 1964; 7:1-16. 1989; 31:187-276. cognitive specialization that assume isola- 6. Lieberman D, Tooby J, Cosmides L. The architecture 15. Sperber D, Wilson D. Relevance: Communication of human kin detection. Nature 2007; 445:727-31. and Cognition. 2nd Ed. New York, NY: Blackwell tion of mechanisms, rather than interac- 7. Humphrey N. The social function of intellect. In: 1995. tion between them, to be the hallmark of Bateson PPG, Hinde RA, Eds. Growing Points in 16. Boyd R, Richerson PJ. Culture and the Evolutionary specialization (reviewed in ref. 19). We Ethology. Cambridge UK: Cambridge 1976; 303-21. Process. Chicago, IL: Chicago 1985. 8. Krebs JR, Dawkins R. In: Krebs JR, Davies NB, Eds. 17. Csibra G, Gergely G. Natural pedagogy. Trends Cog suspect that the investigation of evolution- Behavioural Ecology: An Evolutionary Approach. 2nd Sci 2009; 13:148-53. ary synergies between cognitive specializa- Ed. New York, NY: Sinauer 1984; 380-402. 18. Tomasello M, Carpenter M, Call J, Behne T, Moll H. 9. Byrne RW, Whiten A. Machiavellian Intelligence: Understanding and sharing intentions: The origins of tions is likely to shed light on aspects of Social Expertise and the Evolution of Intellect in cultural cognition. Behav Brain Sci 2005; 28:675-91. human cognitive evolution that could not Monkeys, Apes and Humans. New York: Oxford 1988. 19. Fodor J. The Modularity of Mind. Cambridge, MA: be understood via the study of individual 10. Baron-Cohen S. Mindblindness. Cambridge, MA: MIT 1983. MIT 1995. capacities alone.

524 Communicative & Integrative Biology Volume 3 Issue 6