Evolutionary Cognitive Science: Understanding Brain, Memory, and Reasoning

1. Chapter 9 Evolution, Thought and Cognition

2. Some Points to Remember • Costs and benefits • Evolution doesn’t optimize systems; design to the level of “good enough” • Inclusive fitness

3. Costs of our Large Brain • Energetically expensive (20% energy budget) • Risk of CNS damage • Birthing complications • From evolutionary perspective, what’s the benefit that justifies the costs?

4. What’s the Brain Do? • Biological computer • Computational mechanisms to deal with environmental challenges • Computational theory of mind • Develop computational models of brain function • Test • Substrate neutrality - hardware (mostly) doesn’t matter

5. Levels of Explanation • Computational Theory • What problems was brain evolved to solve • Representation and Algorithm • What abstract mental computations is the brain evolved to execute to meet its goals • Hardware Implementation • How does the physical brain actually work to carry out computations

6. Evolution Applied to Cognitive Science • Visual perception • Memory • Categorization and reasoning

7. Visual Perception • What is the visual system for? • Gives a representation of the external world • Question is one of representational accuracy • Many cases where visual system does not represent the external world “as is” • Is this a design flaw, or an adaptation?

8. Optical Illusions • Show that the internal representation is not the same as the external features

9. Hermann Grid

10. Hering Illusion

11. Julian Beever’s Pavement Art

12. Intentional (Mis)representation • Visual system doesn’t represent the world as it actually is • Marr (1982) argues that this is not an error, but an adaptation • Brain processes visual input and turns it into something useful

13. Brain evolved to function in the real world • Visual illusions play with this • Visual representation by brain interprets the input into a something that is more beneficial to viewer • Fills in missing pieces, maintains colour consistency, adds scale and perspective • Value of visual processing lies in keeping the individual alive long enough to reproduce (and maybe longer)

14. Memory • Value: use past experience to predict future events. • Preparedness • Episodic and Semantic • Specific experiences vs. general facts • Inceptive and derived • All information stored at time of experience vs. processed “summaries” of experience

15. Cost:Benefit in Memory • Recovery of complete encoded information • Speed and ease of recall • Depending on situation, different a balance is required

16. Categorization • A technique to parse information space • Prototypes (“stereotypes”) • Succinct, but non-inclusive • “Majority rule” • Increases retrieval speed and ease, but inaccuracies may occur as a byproduct

17. Faulty Memory • Why isn’t memory perfect? • Schacter’s seven sins of memory • Transience, absent-mindedness, blocking, misattribution, suggestibility, bias, persistence

18. Reasoning and Problem-Solving • Variability exists in environment • Heuristics • “Short-cuts” for problem solving • Not always correct • Algorithms • Computationally “expensive” • Guarantee a correct answer

19. Representational Fallacies • Conjunction fallacy • For event 1 and event 2 to be true, event 1 has to occur first, and is therefore more likely • E.g. Linda is 31 years old, single, outspoken, and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-nuclear demonstrations. Which of the following statements about Linda is more probable? 1. She is a bank teller. 2. She is a bank teller who is active in the feminist movement. • What is more representative of the real world? • Brain mechanisms evolved to solve real world problems…

20. Gambler’s fallacy • A run of bad luck must eventually be replaced with good luck • E.g. Coin toss. Which is more likely: HHHTTT or HTTHHT? • An algorithm interpretation would say neither is more likely • A representational heuristic, though, results in the second option, because it appears more “random”, i.e., more like the real world

21. The probability of something occurring often depends on something else happening first, for which there is also some ambiguity • Bayes Theorem is a statistical principle that calculates the probability of an event being true given the probability of earlier events occurring • People generally don’t problem solve according to Bayes Theorem • Demonstrates Base-rate Neglect (failure to take prior probabilities into account) • But, restructure problem into one of frequencies rather than probabilities, and people do much better

22. Frequency vs. Single-Case Probabilities • Representational problems may be like visual illusions: not actually flaws in the evolved system, but adaptations to operating in a particular (real) environment • Cosimides & Toobey (1996) argue that the human brain is good at dealing with frequencies (i.e., repeatedly occurring events), but not single-case probabilities (one-off events)

23. Frequency Based Decisions • Optimal foraging theory • How should animals partition limited time to maximize gain of required resources? • Basically, an issue of choice • Choice behaviour learned by making repeated choices and preferentially shifting towards those that give more benefits • In essence, based upon frequency of “reward”

24. Difficulty with Single-Case Probabilities • Require particular reference classes to be useful • Non-generalized • E.g., “Odds of winning lottery less than the odds of being struck by lightening.” • But…is this for someone who works outdoors? Lives on a high hill in the open prairie? Has metal golf clubs?

25. Conditional and Logical Reasoning • Not really that good at using rules of logic • E.g., In science, a theory can only be disproven, never proven • Much better at conditional reasoning

26. E K 4 3 Johnson-Laird & Wason (1970) • If p, then q logical rule • Card with vowel has even number on back. • Which card(s) do you turn over to test the rule?

27. Coke Beer 16 19 Griggs & Cox (1982) • If a person is drinking alcohol, they must be over 19 years of age • Imagine you are police checking for underage drinkers

28. Cheat Detection Theory • Cosimides (1989) • Important for social exchange, reciprocity • Due to social nature of humans, evolved modules for detecting freeloading are expected

29. Domain Specific Algorithm • Difficult to do abstract logic task • Underage drinking task triggers mental modules for cheat detection • “Social contract infringement” • Omit police cover story and performance much closer to abstract logic task (Pollard & Evans, 1987)

30. Information Gain Theory • Oaksford & Chater (1994) • Two tasks dealing with entirely different domains • Abstract task: determine truth or falsehood of a rule (an indicative task) • Underage drinking task: not concerned with truth, but with obligations (deontic task)

31. Testing for Rules • Indicative tasks • Reject rule based on finding contradictory evidence • E.g., “all swans are white”; now test • Deontic tasks • Can’t prove rules true or false • E.g., “Under 19 not allowed to drink.” But finding someone breaking the rule doesn’t make it false

32. Presented with Indicative Task • Act to reduce level of uncertainty about world • Rarity assumption: in most cases, finding out something that is true is more informative than finding out something not true • So, in WST, more likely to choose q card than not-q card • Usually, positive information more useful than negative information

33. Presented with Deontic Task • Task requires you to take some perspective towards the rule, such as enforcing it • Rarity assumption does not apply here • High value placed on catching violators • Rational choice is to select p and not-q

34. Which Theory? • Information gain theory explains wider range of logical reasoning tasks than cheat detection theory • Humans as “informavores” • Humans consume information in an analogous way to other animals consume food