Decision Making 3.pdf

Decision-Making Lecture 3 FAST-AND-FRUGAL HEURISTICS Acknowledgement of Country Tarnanthi exhibition space, Art Gallery of South Australia A recap of last week  Timeline t u re 1 L ec Prospect Theory...

Decision-Making Lecture 3 FAST-AND-FRUGAL HEURISTICS Acknowledgement of Country Tarnanthi exhibition space, Art Gallery of South Australia A recap of last week  Timeline t u re 1 L ec Prospect Theory (1979) e3 tu re 4 tu r Lec Lec re 2 Fast and Lec tu Heuristics Bayesian models frugal and biases of cognition heuristics (1974) (1994) (1991) The University of Adelaide Slide 4  An assumption of rationality The University of Adelaide Slide 5  Rationality Classical view Bounded view People are perfect with respect People have limited capacity to knowledge storage, and for information processing, but integration of that knowledge how they use this capacity is with personal preferences. optimised to the environment in which they evolved. The University of Adelaide  Prospect Theory and rationality As discussed by Levy (1997, p. 101), Kahneman and Tversky stated the following about the extent to which Prospect Theory considers human behavior boundedly rational: The curves proposed by Prospect Theory follow from Expected Utility Theory, which claims to demonstrate that human decisions are rational in a classical sense Loss aversion reflects the fact that pain is more urgent to attend to than pleasure. Risk aversion reflects the fact that organisms settle into habits. The University of Adelaide Slide 7  Heuristics-and-biases and rationality According to the heuristics-and-biases view, decisions usually have to be made quickly and thus often rely on simple rules of thumb, or heuristics. Heuristic processing is expected to produce accurate judgements in most situations while generating systematically inaccurate – or “biased” – judgements in some cases. The associated research program focuses on identifying those supposedly rare cases where people make errors, but little is said about what makes those cases different from others. The main criticism of the heuristics-and-biases approach is that it engages little with the question of how people achieve accuracy in most everyday decisions. The University of Adelaide Slide 8 Plan for today  Lecture outline Focus questions Lecture sections What are fast-and-frugal The ‘fast-and-frugal’ approach: heuristics, and how are they Key concepts different from earlier conceptions of heuristics? In particular, what is: o the "less can be more" principle, o the "Take-the-Best" The ‘Take-the-best’ heuristic heuristic? The University of Adelaide Slide 10  Readings for this lecture The University of Adelaide Slide 11 The ‘fast-and-frugal’ approach: Key concepts  Timeline Prospect Theory (1979) Heuristics and Fast and frugal Bayesian models biases heuristics of cognition (1974) (1991) (1994)  Decision-making as a boundedly rational process where ‘less can be more’ Less can be more: Decision-making is a Less-Can-Be-More process, in that there is always a point in the decision-making process when doing more computation or considering more information becomes detrimental to making a good (i.e., more accurate) decision. Less-Can-Be-More is also referred to as ‘ecological rationality’. Definition of a heuristic Heuristics are fast and frugal in that they have simple building blocks: searching, which is the process of looking up cues in Gerd Gigerenzer order of importance for the specific environment stopping, which occurs as soon as one cue allows forming a judgement based on that critical cue The University of Adelaide  Fast and frugal heuristics Under this definition, heuristics are fast, in that they involve few computations, and frugal in that they search only a portion of relevant information—only a portion of relevant cues. Heuristics are accurate, in that simpler alternatives to considering all the available information are available for many decisions. This is the case because cognition is adapted to the environment. Heuristics can, thus, be thought of as the contents of an ‘adaptive cognitive toolbox’. The University of Adelaide Slide 15  Fast and frugal heuristics According to Gigerenzer, people are highly attuned to their environments, and can, thus, select cues and heuristics that are most appropriate in various environments. Heuristics identified by Gigerenzer and his colleagues are listed on this slide. The ability to select appropriate cues The Take-the-Best heuristic and heuristics is likely to be shaped The Recognition heuristic by a number of processes: evolutionary hard-wiring The Fluency heuristic individual learning (within that Tallying environment and similar ones) social processes – imitation and explicit teaching The University of Adelaide Slide 16  Evidence for the ‘fast-and-frugal’ model Gigerenzer provided evidence for his claim that people are highly attuned to their environments. He did this in a series of studies that showed 'reversals' of the biases demonstrated by Kahneman and Tversky when questions were asked in a way that was sensitive to the differences people might see across different environments. The University of Adelaide Slide 17  Evidence for the ‘fast-and-frugal’ model More specifically, Gigerenzer pointed out that most of the survey questions used by Kahneman and Tversky to demonstrate biases required participants to estimate probability or likelihood. Recall one of the questions used by Kahneman and Tversky to demonstrate the phenomenon of base rate neglect—a bias they attributed to the application of the representativeness heuristic: The University of Adelaide Slide 18  Evidence for the ‘fast-and-frugal’ model The answer is 2%, and we can calculate this by applying a mathematical tool called Bayes’ Rule (Gigerenzer, 1991). As you might recall from the lecture on heuristics and biases, participants’ answers to this question tend to be '95%'—rather different to the answer suggested by Bayes’ Rule. Gigerenzer argued that, because heuristics are matched to the structures of specific environments—each with their own unique features—questions about probabilities need to be asked in a way that references the populations from which people or objects with a certain characteristic (say, a disease) are being drawn. Effectively, Gigerenzer suggested that any question about probability can be rephrased to convey exactly the same probabilities in a 'natural frequency' format. For the disease question above, the natural frequency version would sound as follows: The University of Adelaide Slide 19  Evidence for the ‘fast-and-frugal’ model The answer is 1 out of 51, which equals 2%. You know you have tested 1,000 people and so—given the statement 'out of every 1,000 people, 50 perfectly healthy people test positive for the disease'—you know that there will be 50 positive identifications that are, in fact, false positives. You know also, that the test basically comes out positive every time it is given to a person who has the disease. And you know that, in a group of 1,000 people randomly drawn from the American population, there will be one sick person. So, 51 tests out of 1,000 will come back positive, and only one of these will be a correct identification (i.e., a hit). Gigerenzer (1991) reviewed two studies by Cosmides and Tooby (1990) which showed that presenting the disease question to participants in the natural frequency format resulted in correct answers being provided by 76% of participants in one study, and 92% in another. A number of studies—reviewed by Barbey and Sloman (2007)—have observed similar trends. The University of Adelaide Slide 21  Evidence for the ‘fast-and-frugal’ model Recall the ‘Linda problem’ that elicits the conjunction fallacy – another supposed consequence of the representativeness heuristic. Do you find this version of the problem easier? 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 is most likely? 1. Linda is a bank teller 2. Linda is a feminist bank teller There are 100 women like Linda. How many of them are: 1. bank tellers 2. bank tellers and active in the feminist movement  Evidence for the ‘fast-and-frugal’ model As Gigerenzer (1997) reports in the first reading, when shown the ‘natural frequency’ version of the problem, only 20% of people select the more specific answer (2). With the original version, 90% of people select the more specific answer. The University of Adelaide Slide 23 The Take-the-Best heuristic  Definition of Take-the-Best One structural element of environments that heuristics exploit is the degree of correlation (and hence, redundancy) among cues. In this case, the term redundancy refers to one cue being as good as another. For example, when you’re looking at cues about which house to inspect in advertisements, out of the two cues —'three or more bedrooms' and 'double garage'—you only need to know the value of one cue to be quite certain about the value of the other. Houses with three or more bedrooms tend to also have a double garage, and houses with a double garage tend to also have a larger number of bedrooms. Take-the-Best has been suggested to be a heuristic that exploits redundant cues in various environments to produce judgements that are at least as accurate (and in some cases, more accurate) than strategies involving consideration of all cues. Gigerenzer, proposed that Take-the-best is often used when deciding between two recognisable alternatives – say, when deciding between two rental properties to inspect, or two known routes for driving to work. The University of Adelaide Slide 25 Take-the-best heuristic Which city has a larger population – Frankfurt or Leipzig? Gigernzer & Gaissmaier, 2011 Gigerenzer & Goldstein, 1996 Gigerenzer & Goldstein, 1996 3 5 9 8 6 4 2 1 7 Gigerenzer & Goldstein, 1996 3 5 9 8 6 4 2 1 7 Gigerenzer & Goldstein, 1996 Take-the-Best heuristic 1. Search rule: Search through cues in order of their validity. 2. Stopping rule: Stop the search on finding the first cue that discriminates between the alternatives. 3. Judgement rule: Choose the alternative with the “yes” value of the stopping cue. Gigernzer & Gaissmaier, 2011 Take-the-Best heuristic 1. Search rule: Search through cues in order of their validity. 2. Stopping rule: Stop the search on finding the first cue that discriminates between the alternatives. 3. Judgement rule: Choose the alternative with the “yes” value of the stopping cue. Gigernzer & Gaissmaier, 2011 Take-the-Best heuristic 1. Search rule: Search through cues in order of their validity. 2. Stopping rule: Stop the search on finding the first cue that discriminates between the alternatives. 3. Judgement rule: Choose the alternative with the “yes” value of the stopping cue. Gigernzer & Gaissmaier, 2011 Take-the-Best heuristic 1. Search rule: Search through cues in order of their validity. 2. Stopping rule: Stop the search on finding the first cue that discriminates between the alternatives. 3. Judgement rule: Choose the alternative with the “yes” value of the stopping cue. Gigernzer & Gaissmaier, 2011 Effectively, applying Take-the-Best involves following a decision tree. Evidence of accuracy despite frugality Gigerenzer & Goldstein, 1996 Evidence of accuracy despite frugality Simulation study Gigerenzer & Goldstein, 1996 Evidence of accuracy despite frugality Simulation study: predictions and cue usage patterns across a number of decision strategies Gigerenzer & Goldstein, 1996 Evidence of accuracy despite frugality Simulation study: predictions and cue usage patterns across a number of decision strategies -Take-the-best Gigerenzer & Goldstein, 1996 Evidence of accuracy despite frugality Simulation study: predictions and cue usage patterns across a number of decision strategies -Take-the-best -Minimalist Take-the-best Gigerenzer & Goldstein, 1996 Evidence of accuracy despite frugality Simulation study: predictions and cue usage patterns across a number of decision strategies -Take-the-best -Minimalist Take-the-best -Take-the-last Gigerenzer & Goldstein, 1996 Evidence of accuracy despite frugality Simulation study: predictions and cue usage patterns across a number of decision strategies -Take-the-best -Minimalist Take-the-best Variants of Take-the-best -Take-the-last Gigerenzer & Goldstein, 1996 Evidence of accuracy despite frugality Simulation study: predictions and cue usage patterns across a number of decision strategies -Take-the-best -Minimalist Take-the-best Variants of Take-the-best -Take-the-last -Rere -Weighted tallying (summing) Rational strategies -Multiple linear regression Gigerenzer & Goldstein, 1996 Evidence of accuracy despite frugality Frugality (mean Percentage of correct number of cue judgements values looked up) Take the best 5.9 69 Minimalist 5.7 65 Take the last 5.29 65 Weighted tally 14 to 20 66 Multiple regression 14 to 20 69 Gigerenzer & Goldstein, 1996 Further evidence of accuracy despite frugality Weighted tallying strategy Lee, Blanco & Bo, 2017 Evidence of use Stimulus Feature Poison number pattern level Bergert & Nosofsky, 2007 Stimulus Feature Poison number pattern level Stimulus 1 Bergert & Nosofsky, 2007 Stimulus Feature Poison Standard insect: number pattern level 000000 Stimulus 1 Bergert & Nosofsky, 2007 Stimulus Feature Poison Standard insect: number pattern level 000000 Stimulus 4 Bergert & Nosofsky, 2007 Stimulus Feature Poison Standard insect: number pattern level 000000 Stimulus 4 Bergert & Nosofsky, 2007 Stimulus Feature Poison Standard insect: number pattern level 000000 Stimulus 4 Bergert & Nosofsky, 2007 Feature order for individual participant: Stimulus Feature Poison Standard insect: number pattern level 000000 Textured body: 100000 Bergert & Nosofsky, 2007 One of five unseen pairs: Which is more poisonous? OR Only the single All but the most valid single most feature for valid feature predicting for predicting poison level is poison level present are present Bergert & Nosofsky, 2007  Answer One of five unseen pairs: according to rational Which is more poisonous? strategies OR Only the single All but the most valid single most feature for valid feature predicting for predicting poison level is poison level present are present Bergert & Nosofsky, 2007  Answer One of five unseen pairs: according to Answer rational Which is more poisonous? strategies according to Take-the-best OR Only the single All but the most valid single most feature for valid feature predicting for predicting poison level is poison level present are present Bergert & Nosofsky, 2007 Response patterns across five unseen pairs for which Take-the-best and rational strategies make different predictions: Take-the-best Rational Mixed Bergert & Nosofsky, 2007 Summary  Overall, simulation research on Take-the-best illustrates the sense in which heuristics can be adapted to an environment, allowing decision- making that is fast and frugal, yet accurate. However, experimental research on whether Prospect Take-the-best is used in novel environments has Theory produced mixed results. (1979) Fast and Heuristics Bayesian models frugal and biases of cognition heuristics (1974) (1994) (1991) Take-the-best Take-the-best is just one heuristic proposed under the fast-and-frugal framework. Key general concepts proposed by the framework include: a definition of heuristics as containing three components the less-can-be-more principle (also known as ‘ecological rationality’) – an implication of which is that speed and accuracy of decision-making do not need to be traded off against each other the notion of an adaptive toolbox, from which people draw out appropriate heuristics due to evolutionary hard-wiring, and, within specific environments: individual learning, and social processes (imitation and instruction) evidence of reversal of some of the biases identified by Kahneman and Tversky when the question is presented in a ‘natural frequency’ format that supports judgements about specific environments The University of Adelaide Slide 61 62  References Barbey, A. K., & Sloman, S. A. (2007). Base-rate respect: From statistical formats to cognitive structures. The Behavioral and Brain Sciences, 30, 287–292. Bergert, F. B., & Nosofsky, R. M. (2007). A response-time approach to comparing generalized rational and take- the-best models of decision making. Journal of Experimental Psychology. Learning, Memory, and Cognition, 33, 107– 129. Cosmides, L., & Tooby, J. (1990). Is the mind a frequentist? Paper presented at the Second Annual Meeting of the Human Behavior and Evolution Society, Los Angeles, CA. Gigerenzer, G. (1991). How to make cognitive illusions disappear: Beyond “heuristics and biases”. European Review of Social Psychology, 2, 83–115. Gigerenzer, G., & Goldstein, D. G. (1996). Reasoning the fast and frugal way: Models of bounded rationality. Psychological Review, 103, 650–669. Goldstein, E. B. (2018). Cognitive Psychology: Connecting Mind, Research, and Everyday Experience (5th ed.). Cengage. Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux. Lee, M. D., Blanco, G., & Bo, N. (2017). Testing take-the-best in new and changing environments. Behavior Research Methods, 49(4), 1420–1431. The University of Adelaide Slide 63