sign in sign up
<<
Home , Evolutionary psychology, Gerd Gigerenzer

Better than Rational: Evolutionary and the Invisible Hand @ ;

The American Economic Review, Vol. 84, No. 2, Papers and Proceedings of the Hundred and Sixth Annual Meeting of the American Economic Association (May, 1994), 327-332.

Stable URL: http://links.jstor.org/sici?sici=0002-8282%28199405%2984%3A2%3C327%3ABTREPA%3E2.O.CO%3B2-9

The American Economic Review is currently published by American Economic Association.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.j stor.org/journals/aea.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is an independent not-for-profit organization dedicated to creating and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact [email protected].

http://www.jstor.org/ Fri Sep 10 17:38:04 2004 Better than Rational: and the Invisible Hand

Several years ago, we attended an inter- Gigerenzer, 1991). And for the many behav- disciplinary seminar on what were pur- ioral domains where standards of ported to be "biases" in negotiation be- are unclear or undefined, economics is havior. The economists, , and presently mute. biologists present were mulling over the data From a broader scientific perspective, this when, suddenly, a prominent economist lit formulation is decidedly odd. Rational be- up. "Ah, I see,'" he said, "behavior is either havior is not, in any , the state of rational or it's psychological." nature. Not behaving at all is the state of This formulation stuck in our , be- nature in a universe that includes lifeless cause it seemed to succinctly give voice to a planets, prebiotic soup, mountains, trees, tacit assumption held by many economists and tables. All departures from this state of -one that we think works to the detriment inaction require explanation. Moreover, the of economics, by isolating it from the rele- behavioral repertoires of various animals vant parts of , psychology, and the differ profoundly from one another, and rest of the natural sciences. This assump- this must be explained as well: bats cannot tion is that rational behavior is the state of speak, and we cannot navigate through nature, requiring no explanation. Explana- echolocation. Humans and other animals tions that invoke the cognitive processes reason, decide, and behave by virtue of that actually generate human choices are computational devices embodied in neural required only when behavior deviates from tissue. Therefore, a complete causal expla- this state of nature. In this view, economics nation of any behavior-rational or other- is grounded in assumptions of rational be- wise-necessarily invokes theories about the havior, is theoretically constructed out of architecture of these computational devices. what logically follows from assuming ratio- The rationality of a behavior is irrelevant to nal behavior, and gains specificity by plug- its cause or explanation. ging in a variety of variables that are kept Every economic model entails theories exogenous to economics, such as prefer- about these computational devices, but they ences. Merchants of the ad hoc and exoge- are usually left implicit, buried in the as- nous, psychologists are called in only to sumptions of the model. At the moment, provide second-order corrections to eco- most economists rely on the implicit (and nomic theory, usually by furnishing a cata- somewhat vague) theory that these compu- log of oddities and quirks in human reason- tational devices somehow embody "rational" ing (e.g., "biases" and "fallacies"-many of decision rules. But developing a more accu- which are turning out to be experimental rate, useful, and well-defined substitute for artifacts or misinterpretations; see G. this black box is now a realistic goal. Results from the newly emerging field of - ary psychology suggest that (i) explicit, well-specified models of the human *Department of Psychology, University of Califor- can significantly enhance the scope and nia, Santa Barbara, CA 93106, and Department of specificity of economic theory, and (ii) ex- , University of California, Santa Barbara, plicit theories of the structure of the human CA 93106, respectively. For enlightening discussions, mind can be made endogenous to economic we warmly thank , Robert Nozick, models in a way that preserves and expands and Paul Romer. For financial support, we are grateful to the McDonell Foundation and NSF Grant No. their elegance, parsimony, and explanatory BNS9157-449 to Tooby. power. 32 7 328 AEA PAPERS AND PROCEEDINGS MAY 1994

At present, economics hovers, scientifi- of relationships that link inputs to outputs, cally unsupported and isolated in mid-air, can deduce the structure of a computer theoretically levitating on the assumption of program. Hence cognitive can ask: rationality. But this is unnecessary and lim- do our brains embody procedures that carry iting. There are tight causal and analytic out the rules of logical inference? or calcu- connections between economics, psychol- late Bayesian probabilities? or cause us to ogy, and (Tooby and avoid incest? or allow us to echolocate? Cosmides, 1992a): Of course, the information-processing structure of the mind is fantastically com- 1. -an invisible-hand plex, so conducting experiments blindly, process-is the only component of the without knowing what to look for, has not evolutionary process that produces com- been an efficient research strategy. This is plex functional machinery in organisms, where the integration of evolutionary biol- such as the vertebrate eye (Richard ogy and has proved so Dawkins, 1986). useful. The applicability of evolutionary bi- 2. Natural selection built the decision-mak- ology is based on a simple but powerful ing machinery in human minds. idea. Form follows function: the properties 3. This of cognitive devices generates all of an evolved mechanism reflect the struc- economic behavior. ture of the task it evolved to solve. 4. Therefore, theories of economic behav- This approach has teeth because there is ior necessarily include theories about the only one class of problems that evolution structure of the cognitive mechanisms produces mechanisms for solving: adaptive that generate that behavior. Moreover, problems. These are problems that recurred the design features of these devices de- across many generations during a species' fine and constitute the human universal evolutionary history, and whose solution principles that guide economic decision- statistically promoted in an- making. cestral environments. By identifying and modeling the adaptive problems humans In other words, natural selection's invisible faced during their evolution, researchers can hand created the structure of the human make educated guesses about the designs of mind, and the interaction of these minds is the computational devices the hu- what generates the invisible hand of eco- brain embodies, and about many of the nomics (e.g., Tooby and Cosmides, 1992a): specific design features they required to be one invisible hand created the other. able to solve these problems. Armed with these models, researchers can then design I. Evolutionary Psychology experiments that can detect and map the features of these complex devices-features The brain is a complex computational that no one would otherwise have thought device, a system that takes sensory informa- to test for. tion as input, transforms it in various ways, stores it, analyzes it, integrates it, applies 11. A New View of the Mind decision rules to it, and then translates the output of those rules into the muscular con- The application of these methods is lead- tractions that we call "behavior." For the ing to a fundamentally new view of the most part, humans have no more conscious architecture of the human mind. Previously, access to the structure of these programs the mind was thought to resemble a gen- and the decision rules they embody than to eral-purpose computer: initially free of any the processes through which the kidneys content that had not originated in the select what to excrete. Nevertheless, the and the social world (the as- structure of these information-processing sumption), and equipped only with a small programs can be mapped in much the same number of content-independent rules of in- way that a programmer, by studying the web ference (e.g., rules drawn from logic, mathe- VOL. 84 NO. 2 INVISIBLE-H AND THEORIES 329 matics, and probability theory, or associa- ral problems is the primary reason why tive rules). problem-solving specializations were fa- Converging lines of evidence from an ar- vored by natural selection over general-pur- ray of disciplines are replacing this view pose problem-solvers. Despite widespread with a model in which the human cognitive claims to the contrary, the human mind is architecture resembles a large and hetero- not worse than rational (e.g., because of geneous network of functionally specialized processing constraints)-but may often be computational devices. Because biological better than rational. On evolutionarily re- evolution is a slow process, and the modern current computational tasks, such as object world has emerged within an evolutionary recognition, grammar acquisition, or speech eye-blink, these devices are inherited from comprehension, the human mind greatly the past and remain functionally specialized outperforms the best artificial problem-solv- to solve the particular distribution of prob- ing systems that decades of research have lems that were characteristic of humans' produced, and it solves large classes of hunter-gatherer past, rather than those of problems that even now no human- the modern world (e.g., habitat selection; engineered system can solve at all. foraging; social exchange; competition from How can this be? General-purpose sys- small armed groups; parental care; language tems are constrained to apply the same acquisition; contagion avoidance; sexual problem-solving methods to every problem rivalry) (for discussion, see Tooby and and can make no special assumptions about Cosmides [1992aI). The fact that these de- the problem to be solved. Specialized prob- vices are (a) specialized rather than lem-solvers are not handicapped by these general-purpose and (b) specialized to solve limitations. Many facts and relationships seemingly exotic ancestral problems rather relevant to particular types of adaptive than all problems, or modern problems, problems were stably true of the world dur- leads to markedly different sets of predic- ing (e.g., incestuous mat- tions about and decision- ings produced a high proportion of birth making. In addition, this view implies that defects; human grammars were limited to a cultural differences are vastly overstated, restricted set of patterns). Natural selection because beneath existing surface variability could equip humans' cognitive specializa- all humans share the same of set of prefer- tions with design features and problem-solv- ence-generating and decision-making de- ing strategies that exploited the presence of vices. Finally, the tabula rasa assumption is these problem-specific regularities to solve being discarded: human mental content does particular classes of recurrent problems in not simply originate in the external world. efficient ways appropriate only to that class. Specialized mental mechanisms inject re- Triggered by cues that a particular problem current content into human thought across type has been encountered, a network of cultures (through privileged hypotheses and dedicated computers can selectively deploy conceptual primitives, specialized represen- from its large repertoire those specialized tational formats, privileged preference gen- procedures that are well designed for solv- erators, et~.). ing that particular problem. For the problem One point is particularly important for domains they are designed to operate on, economists to appreciate: it can be demon- specialized problem-solving methods perform strated that "rational" decision-making in a manner that is better than rational; that methods (i.e., the usual methods drawn from is, they can arrive at successful outcomes logic, mathematics, and probability theory) that canonical general-purpose rational are computationally very weak: incapable of methods can at best not arrive at as effi- solving the natural adaptive problems our ciently, and more commonly cannot arrive ancestors had to solve reliably in order to at all. Such evolutionary considerations sug- reproduce (e.g., Cosmides and Tooby, 1987; gest that traditional normative and descrip- Tooby and Cosmides, 1992a; , tive approaches to rationality need to be 1994). This poor performance on most natu- reexamined (Tooby and Cosmides, 1992a,b; 330 AEA PAPERS AND PROCEEDINGS MAY 1994

Robert Nozick, 1994). (Unfortunately, be- of exchange. Needless to say, a specialized cause decision theorists and philosophers social exchange logic and associated circuits have almost equated rational methods with allows gains in trade to be identified by general-purpose analytic tools, the analysis individuals, trades to be arranged, cheaters and development of domain-specific meth- to be excluded, and hence markets to ods has been relatively neglected.) emerge spontaneously (Cosmides and Tooby, 1992). Equally important, experi- 111. Reasoning Instincts mental evidence suggests that humans have specialized circuits for understanding From this perspective, the human mind is threats, as well as recognizing bluffs and powerful and intelligent not because it con- double-crosses. The ability of humans to tains general-purpose rational methods (al- understand and properly interpret each though it may include some), but primarily other's threats also allows and structures because it comes equipped with a large ar- the emergence of social predators, coercive ray of what one might call "reasoning in- coalitions, governments, and other extortive stincts." Although- instincts are often social arrangements. thought of as the polar opposite of reason- Moreover, models of ing, a growing body of evidence indicates and do not seem to capture accu- that humans have many reasoning, learn- rately the distinctive organization of human ing, and preference circuits that (i) are reasoning. Such models were based on plau- complexly specialized for solving the spe- sible notions of limited processing capacity cific adaptive problems our hominid ances- (e.g., Herbert Simon, 1956; D. Kahneman tors regularly encountered; (ii) reliably de- et al., 1982) and seemed to make sense of, velop in all normal human beings; (iii) devel- for example, the experimentally arrived at op without any conscious effort: (iv) develop consensus that humans lacked the ability to without any formal instruction; (v) are ap- do Bayesian reasoning. Recent results from plied without any awareness of their under- , however, show that lying logic; and (vi) are distinct from more bumblebees engage in probabilistic induction general abilities to process information or that the is considered "too behave intelliaentlv.- - In other words. these limited" in capacity to perform (see e.g., reasoning, learning, and preference circuits Leslie Real, 1991). This suggested that the have all the hallmarks of what people usu- artificial nature of the experiments usually ally think of as "instincts" (Pinker, 1994). conducted might not be triggering latent They make certain kinds of inferences just competences. As it turns out, although it is as easy, effortless, and "natural" to humans true that people are bad at calculating the as spinning a web is to a spider or building a probability of a single event, when probabil- dam is to a beaver. ities are expressed as frequencies, the stan- It is important to identify and map these dard "fallacies" and "biases" seem to dis- specialized computational devices, because appear, and statistically naive subjects they powerfully shape economic processes behave like good Bayesians (Gigerenzer, (e.g., by making some social interactions 1991; Cosmides and Tooby, 1994). From an easy for the participants to understand and evolutionary perspective, this is not surpris- engage in, and others more difficult to un- ing. Our hunter-gatherer ancestors were derstand or to successfully navigate). For awash in statistical information in the form example, humans do not seem to have avail- of the encountered frequencies of real able on-line circuits that perform many logic events; in contrast, the probability of a sin- operations (e.g., modus tollens). However, gle event was inherently unobservable to experimental evidence indicates that hu- them (a single event either happens or it mans do have evolved circuits dedicated to does not). Natural selection can only be a more specialized task of equal (or greater) expected to have built mechanisms to complexity: detecting cheaters in situations exploit information in the form that was VOL. 84 NO. 2 INVISIBLE-HAND THEORIES 331 regularly available to our ancestors. Such everywhere to find alternative sets of rules have implications about how to be reasonable, depending on how closely humans process the various forms of infor- their particular economic environment mim- mation available to them about their eco- ics various ecological condi- nomic environments and supports the view tions. that, instead of bounded rationality, hu- For example, humans evolved in small mans have certain nonconscious natural bands who lived by hunting animals and competences that are far better than is gathering plant foods. Hunting (in many presently appreciated. environments) is a high-variance activity, in which luck plays a major role and individu- IV. Universal Preferences als run a significant chance of coming back and "Rules of the Game" empty-handed for several days running. Op- timality analyses from The most straightforward application of indicate that under circumstances of high evolutionary psychology to economics in- variance for individual foragers, band-wide volves the likelihood that the two communi- food-sharing is individually beneficial, a ties collaboratively might be able to create a form of -pooling that smoothes out what science of preferences (e.g., the session on would otherwise be a feast-or-famine cycle "Preferences" at the 1994 Allied Social Sci- for individuals and families. Because forag- ence Association meeting featuring papers ing and sharing decisions are complex adap- by Gary Becker and Casey Mulligan, Robert tive problems that humans faced for mil- Frank, and Paul Romer). A psychological lions of years, humans should have evolved architecture that simply acquired an arbi- cognitive programs specialized for solving trary set of preferences, provided they were them; and recent evidence from the study of present in the social environment, could not modern hunter-gatherers suggests that they have been plausibly produced by the evolu- have. These mechanisms monitor local in- tionary process (Tooby and Cosmides, formation about factors such as resource 1992a). Instead, evolutionary psychology variance, using it as a "switch" to turn vari- should be able to supply a list of human ous alternative sharing programs on or off. universal preferences, and of the proce- These mechanisms should make sharing dures by which additional preferences are rules appealing in conditions of high vari- acquired or reordered. Models of such ance, and unappealing when resource ac- mechanisms should be able to address crual is a matter of effort rather than of long-standing problems in economics by luck (Cosmides and Tooby, 1992). In other widening the scope of preferences beyond words, different "rules of the game" can be the usual notion of goods and services (e.g., triggered in a lawful way by specific kinds of preferences for participation in coalitions ecological variables. [Tooby and Cosmides, 19881, for as a Knowing that such computational devices function of sex and age, and even for cer- exist could, for example, serve as the basis tain "rules of the game" [Cosmides and for deeper and more scientifically satisfying Tooby, 19921). theories of the conditions under which cer- Indeed, it is frequently assumed that the tain ideas and arise and are rules of economic interaction are the out- maintained (Cosmides and Tooby, 1992). come of historical processes, created by For example, in the modern world, the op- "society" or by the action of "visible hands" eration of Pleistocene-forged variance-sen- -explicit, conscious, human intentions. sitive sharing programs would probably However, humans' evolved computational make cost-sharing for medical care more devices may also tacitly supply much of the psychologically appealing than for many structure of these "games." Indeed, our other goods, because illness is seen to have evolved psychology may have alternative a large random component. Less obviously, modes of operation that prompt humans knowing the structure of such cognitive 332 AEA PAPERS AND PROCEEDINGS MY 1994 devices could help economists understand (forthcoming). why certain economies grow so slowly. The Dawkins, Richard. The blind watchmaker. New savings and investment necessary for eco- York: Norton, 1986. nomic growth may be difficult to achieve- Gigerenzer, G. "How to Make Cognitive Illu- even when average per capita income is sions Disappear: Beyond Heuristics and increasing-if there is a substantial amount Biases," in Wolfgang Stroebe and Miles of effort-independent variance in economic Hewstone, eds., European review of social welfare (caused, e.g., by political unrest or psychology, Vol. 2. Chichester, U.K.: frequent natural disasters). In such condi- Wiley, 1991, pp. 83-115. tions, an ethic of widespread sharing could Kahneman, D.; Slovic, P. and Tversky, A. Judg- be triggered, and resources that might ment under uncertainty: Heuristics and bi- otherwise be saved and invested might in- ases. Cambridge: Cambridge University stead be consumed. Press, 1982. Lastly, cognitive specializations deter- Nozick, Robert. "Invisible-Hand Explana- mine what inferences will be triggered by a tions." American Economic Review, May situation and what kinds of information will 1994 (Papers and Proceedings), 84(2), pp. be attention-grabbing, memorable, learn- 314-18. able, communicable. Their structure there- Pinker, Steven. . New fore determines which ideas can be easily York: Morrow, 1994. replicated from mind to mind, and which Real, Leslie. "Animal Choice Behavior and cannot (, 1990). Explicit models the Evolution of Cognitive Architecture." of these mechanisms will eventually allow Science, 30 August 1991, 253, pp. 980-86. economists to develop well-grounded theo- Simon, Herbert. "Rational Choice and the ries about the Hayekian distribution of Structure of the Environment." Psycho- knowledge. logical Review, March 1956, 63(2), pp. In sum, economic theory can increasingly 129-38. be grounded in the theoretically and empiri- Sperber, Dan. "The Epidemiology of Beliefs," cally derived models of human decision- in Colin Fraser and George Gaskell, making machinery that are presently being eds., The social psychological study of constructed within evolutionary psychology. widespread beliefs. Oxford: Clarendon, 1990, pp. 25-44. REFERENCES Tooby, J. and Cosmides, L. "The Evolution of War and Its Cognitive Foundations." Cosmides, L. and Tooby, J. "From Evolution Technical report, Institute for Evolution- to Behavior: Evolutionary Psychology as ary Studies, Palo Alto, CA, 1988. the Missing Link," in J. Dupre, ed., The . "The Psychological Foundations of latest on the best: Essays on evolution and Culture," in J. Barkow, L. Cosmides, and optimality. Cambridge, MA: MIT Press, J. Tooby, eds., : Evolu- 1987, pp. 277-306. tionary psychology and the generation of . "Cognitive for Social culture. New York: Oxford University Change," in J. Barkow, L. Cosmides, and Press, 1992a, pp. 19-136. J. Tooby, eds., The adapted mind: Evolu- . " and the tionary psychology and the generation of Multimodular Mind: Grounding Norma- culture. New York: Oxford University tive Theories in Adaptive Problems." Press, 1992, pp. 163-228. Technical Report 92-1, Center for Evolu- . "Are Humans Good Intuitive tionary Psychology, University of Califor- Statisticians After All?" , 1994 nia-Santa Barbara, 1992b.