Judgement and Decision Making 1

Memory distortion (hindsight bias)

"I knew it all along!"

Inevitability (hindsight bias)

"It was bound to happen", "It couldn't have turned out any other way!"

Forseeability (hindsight bias)

"They should have known better, anybody could have seen this coming"

Hindsight bias

Given knowledge of an outcome, we tend to exaggerate the degree to which the outcome could have been predicted beforehand. 1. Memory distortion 2. Inevitability 3. Forseeability

Availability Heuristic: Drawing examples

1. Examples leap readily to mind when the event is a common occurrence 2. It is hard to come up with examples when the event is uncommon/rare 3. Availability works fine when memory is unbiased. -Events that really are more frequent really are easier to recall. -But memory can be biased... --the way we encode and retrieve information. --Examples you draw from are usually not equally accessible.

Availability in Real Life

1. Plane crash vs. car accident. -Why are people more afraid of the plane crash? ---Media coverage. ---Car crash is much more likely, but rarely make the front page. ---Plane crashes receive increased and extensive coverage, no matter where the crash occurred. -Unusual events recalled more easily. 2. Playing the lottery. -Never hear about people not winning the lottery, even though it happens all the time. -Only hear about people winning the lottery-availability leads people to overestimate their odds of winning.

Gambler's fallacy

An erroneous belief that a random process will automatically keep track of the outcomes in order to make the overall rate of an outcome in the short run equal to the overall rate of that outcome in the long run.

Tversky & Kahneman, 1973, 1974: Letter Experiment

Asked: Are there more English words that start with K or that have K as the third letter? -Is easier to think of words that start with k. --kind, kangaroo, kick, key. -Ease of search leads participants to estimate that letters that start with k are more frequent. -Hard to think of words with k as the third letter. --Like, make, acknowledge... -There are twice as many words that have k as a third letter than have k to start.

Fischhoff, Slovic & Lichtenstein (1977):

Fischhoff, Slovic & Lichtenstein (1977). Participants answered general knowledge questions: "Is absinthe a liqueur or a precious stone?" -For questions about which participants were 100% confident, they were only right 75% of the time.

Decision Making

Generating, evaluating, and selecting among a set of relevant choices, where the choices involve some uncertainty or risk. Based on a judgement, make a choice from among several alternatives. The outcome is uncertain.

Base rate

How often an item occurs in a population

Heuristic

Informal strategy that provides a quick solution but does not guarantee a precisely correct one.

Summary

Judgments based on heuristics (availability, representativeness) can be accurate. But biases exist: --Memory biases can lead to biased judgments based on availability. --Representativeness can lead people to ignore principles of probability. ----Gambler's fallacy. ----Conjuction fallacy. ----Base-rate fallacy. Anchoring effects lead people to discount new information and stick with initial estimates. We aren't all that good at judging what we know and when we knew it. --Hindsight bias. --Overconfidence.

The Focus on Errors

Much of the research on judgement and decision-making relies on errors people make. Heuristics usually allow us to make the best decisions we can given processing limitations and limited information. -Knowing what heuristics we use can actually help you make sure you're using them properly and not being led astray. The patterns of errors that we see tell us something about how these processes typically operate. Errors are the price we pay for fast, efficient processing.

Tversky and Gilovich (1989): Cold Facts about the "Hot Hand" in Basketball

Observers see evidence of the "hot hand" where none exists. -Records of individual players (from the 76ers, Nets and Knicks) show no evidence of the success of a shot depending on the one that came before. -Free throws (Celtics): shooting percentage for the second shot not different following hits and misses.

Conjunction fallacy

Occurs when people estimate that the odds of two uncertain events happening together are greater than the odds of either event happening alone.

Overconfidence

Our confidence in our judgements is often much too high. Confidence ratings are higher than actual accuracy Arrogance in decision making

anchoring and adjustment: population of Chicago

Participants asked: "Is the population of Chicago more or less than 200,000 (5 million)?" Are then asked: "What is the actual population of Chicago?" -People are unduly influenced by the suggestion in the initial question. -Estimates are significantly higher in the 5 million condition than the 200,000 condition. (Tversky and Kahneman, 1974)

Tversky and Kahneman (1973): Fame Experiment

Participants shown list of names and were asked to estimate how many women's name they saw and how many men's names they saw. There were 10 women's name and 10 men's names. -Women's names were all famous women, the men's names were not. From what we know of memory - the participants probably did not memorize the list, so how did they generate their answers? Female names were more easily brought to mind (names of famous people, men's names were not). -When they are easier to recall, you have the impression that there are more of them. -81% of participants judged that there were more women's name than men's names. Evidence of the availability heuristic.

Tversky and Kahnman (1971): Law of Small Numbers

People expect small samples to resemble in every respect the populations from which they are drawn. In actuality, small samples are much more likely to deviate from the population and therefore are a less reliable bias on which to build a conclusion than are larger samples. Examples: -The hot hand. -Hitting streaks (superstititons).

Why does the conjunction fallacy occur?

People ignore the rules of probability when making judgements. They instead base their judgements on their assessments of similarities to a particular category.

Tversky and Gilovich (1989): The "hot hand".

Performed controlled experiments with players from Cornell's varsity men's and women's basketball teams... 1. Found that the probability of a hit following a hit (.47) was the same as the probability of a hit following a miss (.48) 2. Asked the players and an observer to bet on the outcome of each upcoming shot. --Bet high - win 5¢ for a hit, lose 4¢ for a miss --Bet low - win 2¢ for a hit, lose 1¢ for a miss Players were unsuccessful at predicting their performance (average correlation = .02) and the observers weren't very good either (.04). -However, the bets made by both players and observers were correlated with the outcome of the previous shot (r=.40 and r=.42). -Both players and observers relied on the outcome of the previous shot when making their predictions. ---But this betting strategy was no better than chance! This is what the hot-hand hypothesis predicts.

The Gambler's Fallacy and Coin Tosses

Someone tosses a fair coin 6 times, and it comes up heads every time. -What happens on the 7th toss? -Is it "due" to come up tails? Every flip of the coin is an independent event. -Coin does not keep track of its past performance/behavior. -The chance of getting heads on the 7th toss is exactly the same as the chance of getting tails In the long run the number of times heads comes up should equal the number of times tails comes up, but in the short run proportions do not and will not usually be even

Algorithm

Specific rule or procedure, often detailed or complex, that guarantees a correct answer.

Availability Heuristic

Tendency to form a judgment on the basis of information is readily brought to mind. -We assess the frequency or probability (likelihood) of an event based on how easy it is to come up with relevant examples.

Judgement

The human ability to infer, estimate and predict the character of unknown events. Drawing a conclusion based on limited information.

Conjunction rule

The probability of the conjunction of two events cannot be larger than the probability of its constituent events.

Base rate fallacy

The tendency to ignore information about general principles in favor of very specific but vivid information. Using prototypical or stereotypical factors while ignoring actual numerical information.

Descriptive approach

The way we actually do think in a given situation.

Normative approach

The way we ought to think in a given situation.

anchoring and adjustment: math problem

Two groups of high school students are given 5 seconds to estimate a complex (identical) expression: -Group 1: estimates 8x7x6x5x4x3x2x1 to be 2,250. -Group 2: estimates 1x2x3x4x5x6x7x8 to be 512. Both estimates were too small (actual answer = 40,320). People tend to perform the first few steps of multiplication and then extrapolate. -Extrapolation tends to be too little rather than too much. Group 1 estimate was much lower because their anchor (initial approximation) was lower than that of Group 2.

Representativeness Heuristic

Used when calculating probability. Ignore statistical information and focus on similarity of an instance to its category. The belief that outcomes will always reflect characteristics of the process that generated them

Representative Heuristic and Fallacies

Using similarity, ignoring other relevant info (e.g. sample size, base rates). -Gambler's Fallacy: Belief in the law of small numbers. -Base Rate Fallacy: Ignoring overall frequency of events. -Conjunction Fallacy: Believing that a conjunctive statement is more probable that either of its component statements.

hot hand vs gamblers fallacy

When even short sequences look non-random, the representativeness heuristic can lead us to two different predictions? 1. In coin tosses - we bet on tails after a run of heads (the sequence will right itself, even out). 2. In sports - we bet on a hit after a run of hits (the streak will continue). Perhaps when we can imagine non-random causes for the streaks and slumps we see, we imagine those effects will continue. -When the sequence is generated by a system where no cause is apparent, we imagine those streaks will even out.

Kahneman & Tversky, 1973: Lawyers and Engineers

You randomly draw Jack from a sample of 100, and are asked to determine if it is more likely Jack in an engineer or a lawyer. -Are told that in the sample, there are 30 engineers and 70 lawyers. -Are also provided a description of Jack (45 yrs old, generally conservative, careful, ambitious, shows no interest in political or social issues, and spends most of his free time on his many hobbies which include mathematical puzzles.) Given both sources of information, participants made their judgments based solely on the description. -In fact, they reported the same results when the proportions were reversed (30 lawyers, 70 engineers). -Participants ignore base rates in favor of representativeness.

Anchoring

a decision-making heuristic in which final estimates are heavily influenced by initial value estimates. Judgements are too heavily influenced by initial approximations.