Are Humans Rational? SymSys 100 Lecture April 7, 2009 AndersonвЂ™s Rational Approach to Cognition вЂў Why do we see the regularities that we see in various Cognitive tasks? вЂў One answer: вЂ“ Because of characteristics of the mechanisms of human thought вЂў Another answer: вЂ“ Because the regularities represent the optimal response to the characteristics of the environment Example: Forgetting вЂў Why do we forget? вЂў Is it because: вЂ“ our memory is imperfect вЂў Or is it because: вЂ“ it costs us something to keep information around andвЂ¦ вЂ“ Information from a long time ago is less likely to be useful than information weвЂ™ve been exposed to recently. Power Law of Forgetting (Anderson and Schooler, 1991) пЂb or: P пЂЅ AT log( P ) пЂЅ log( A ) пЂ b log( T ) Which topic is most likely to be in tomorrowвЂ™s paper? вЂў Two headlines from 100 days ago: вЂ“ Focus on weatherization is shift in energy costs вЂ“ Blagojevich prosecutors seek ruling on tapes вЂў Two headlines from yesterday: вЂ“ North Korean missile launch was a failure вЂ“ Santana is sharp in MetsвЂ™ opening victory Power Law of Recurrence Rational or Irrational? The card selection task вЂњIf a card has an even number on one side, then it has a vowel on the otherвЂќ Circle the cards to turn over: 3 8 A N вЂњIf you are drinking alcohol, then you must be over 18вЂќ Circle the cards to turn over: coke beer 22 17 Explanations of Abstract Card Selection Task Results вЂў Oaksford & Chater (вЂњA rational analysis of the selection task as optimal data selectionвЂќ): вЂў Their example: вЂ“ вЂ�If you eat tripe, then you will get sickвЂ™ вЂ“ Participants turn over cards that they expect will give them the most information вЂў Premise and consequence are thought to pick out rare situations and thus to be highly informative. вЂ“ Therefore participants turn over the even number with the highest probability) then the vowel, with the next highest. вЂњThe purpose of a rational analysis is to show that behavior is optimally adapted to the environment.вЂќ Is the Human Mind вЂ�The Best of All Possible Minds?вЂ™ A parable from VoltaireвЂ™s Candide вЂў вЂў вЂў вЂў вЂў вЂў Candide inquired into the cause that had reduced Pangloss to so miserable a condition. "Alas," replied the preceptor, "it was love; love, the comfort of the human species; love! tender love!вЂњ "Alas,вЂќ cried Candide, вЂњBut how could this beautiful cause produce in you so hideous an effect?" Pangloss answered thus: "O my dear Candide, you remember Pacquette, that pretty wench, who waited on our noble Baroness; in her arms I tasted the pleasures of Paradise, which produced these Hell torments. She was infected with an ailment, and perhaps has since died of it; she received this present of a learned Franciscan; he was indebted for it to an old countess, who had it of a captain of horse вЂ¦ who had it in a direct line from one of the fellow adventurers of Christopher Columbus." "O Pangloss," cried Candide, "what a strange genealogy is this! Is not the devil the root of it?" "Not at all," replied the great man, "it was a thing unavoidable, a necessary ingredient in the best of worlds; for if Columbus had not caught in an island in America this disease, which contaminates the source of generation, and is evidently opposed to the great end of nature, we should have had neither chocolate nor cochineal. вЂ¦ In the meantime, we may safely affirm, that, when an army of thirty thousand men engages another equal in size, there are about twenty thousand infected with syphilis on each side." Other Explanations вЂў Ambiguity: People interpret вЂ�ifвЂ™вЂ¦ вЂ�thenвЂ™ to mean different things in different contexts (вЂњif and only ifвЂќ, maybe). вЂў Availability: Statement of the rule does not mention the odd number, thus participants donвЂ™t think about it much. вЂў Why do people do better with the drinking example? вЂ“ Laws and social rules bring to mind a concern about the importance of compliance, leads to search for possible violations. вЂў Did evolution endow us with innate вЂ�Cheat detectorsвЂ™, as proposed by Tooby and Cosmides? вЂ“ Only some people think so! Rational or Irrational? The Conjunction Fallacy Tversky & Kahneman, 1983 Feminist Bank Tellers Bank Tellers Feminists T&KвЂ™s Explanation of the Conjunction Fallacy вЂў вЂњWe propose that a judgment of probability or frequency is commonly biased toward the natural assessment that the problem evokes. Thus, the request to estimate the frequency of a class elicits a search for exemplars, the task of predicting vocational choice from a personality sketch evokes a comparison of features, and a question about the co-occurrence of events induces an assessment of their causal connection. These assessments are not constrained by the extension rule. Although an arbitrary reduction in the extension of an event typically reduces its availability, representativeness, or causal coherence, there are numerous occasions in which these assessments are higher for the restricted than for the inclusive event.вЂќ Framing Effects 80 70 60 50 40 30 Or after the same cover storyвЂ¦ 20 10 0 Lives Saved Certain Lives Lost Uncertain Why does this happen? вЂў We evaluate outcomes relative to a baseline or neutral point. вЂў Losses loom larger than gains. вЂў Both gains and losses show a diminishing returns effect. Are T&K Overplaying the Laws of Probability? Chase, Hertwig & Gigerenzer, 1998 вЂў вЂњProponents of [competing views] agree on one critical point: rationality requires reasoning in accordance with the rules of probability theory. вЂ¦ [But] no single conception of probability is shared by all statisticians and philosophersвЂ¦ In our view, wherever a normвЂ™s applicability depends on our interpretation of probability in this way, we are not justified in treating it as an unequivocal norm of sound reasoning.вЂќ Visions of Rationality вЂў вЂ�Bounded RationalityвЂ™ (Simon, 1957) вЂ“ People donвЂ™t have the resources it would take to be rational; they are always working under constraints. вЂў вЂњExpecting peopleвЂ™s inferences to conform to classical rational norms in such complex environments requires believing that the human mind is a вЂ�Laplacean demonвЂ™: a supercalculator with unlimited time,knowledge, and computational power.вЂќ (Chase et al., TiCS, 1998) вЂ“ Instead of optimizing, says Simon, they вЂ�satisficeвЂ™. вЂў A satisficing strategy may often be (near) optimal if the costs of the decision-making process itself, such as the cost of obtaining complete information, are considered in the outcome calculus. Fast and Frugal Heuristics вЂў Which has the largest population: San Diego or San Antonio? San Diego: San Antonio: 1,336,865 2,031,445 вЂў The вЂ�recognition heuristicвЂ™: вЂ“ if you recognize one object and not the other, then infer that the recognized object has the higher value on the target variable; if you do not recognize either object, then guess. вЂњTake the bestвЂќ вЂў What happens if you recognize both cities? вЂ“ Use вЂ�Take the BestвЂ™ вЂў Search for facts about each city, starting with those most likely to be good predictors of the answer. вЂ“ How many professional sports teams? вЂ“ How many universities? вЂ“ How many people from there have you heard of? вЂў When you find a variable where you know the answer for both cities, you stop, and choose on that basis. вЂў Often this is more accurate than considering all available sources of information. Are Humans Rational? вЂў Yes вЂ“ Anderson, Chater вЂў No вЂ“ Tversky, Kahneman, (Voltaire) вЂў They do pretty well with limited resources вЂ“ Simon, Gigerenzer вЂў Your opinion?