Lecture 18: Judgment, Decisions, and Reasoning (Ch. 13)

Judgments vs Judgment Heuristics

Judgments: the estimates that people make of important real-world quantities (height, weight, intensity, probability, etc.)

Judgment Heuristics: easy, natural strategies for making judgments

• They provide answers that are often reasonable but systematically biased 

Availability

One important quantity that people must estimate is the probability of events

• People often estimate probability on the basis of examples of the event that are readily available in memory

  • But when we remember is not always more probable

Accuracy of Human Risk Assessment (Lichenstein, Slovic, Fishhoff, Layman, & Combs, 1978)

Lichenstein et al. asked people to judge the rates of many causes of death (tornado, botulism, cancer, stroke, flood, car accident, etc.)

• People overestimate the probability of rare events but underestimate the probability of common events

Judged vs Actual Mortality Rates (Hertwig, Pachur, & Kurzenhauser, 2005)

• Dramatic events are overestimated (ex: flood, even though there hadn’t been a flood in Germany for years where the data had been collected)

• Non-dramatic events are underestimated 

Availability and Risk Assessment

People overestimate accidents and natural disasters

• They are dramatic and receive a lot of attention (media coverage) and hence are available in memory (plane crashes, shark attacks, etc.)

This can lead people to make real-life decisions that are life-threatening

Availability and Risky Decisions (Gigerenzer, 2004)

In 2004, Gigerenzer analyzed patterns of travel after 9/11

• In the following months there was a decrease in air travel

• At the same time, there was a 2.9% increase in freeway driving that led to an increase in driving deaths

Availability ≠ Probability

We’ve seen that there are many factors that determine what we can remember

• Semantic elaboration

• Interference

• Imagery (especially when vivid)

But factors that make something more memorable does not necessarily make it more probable

Representativeness

People often make judgments of probability on the basis of similarity or representativeness

1. Conjunction fallacy

2. Neglect of the law of large numbers

3. Base rate neglect

1. Conjunction Fallacy (Tversky & Kahneman, 1982)

People sometimes judge the conjunct as more probable

• According to probability theory, the conjunction of two events cannot be greater than either of the events themselves

The conjunct is less probable but more representative of (or similar to) Linda:

• Linda is a bank teller - more probable but less representative

• Linda is a bank teller and active in the feminist movement - more representative but less probable

2. Neglect of the Law of Large Numbers (Tversky & Kahneman, 1974)

Larger samples provide stronger evidence than smaller ones

• people often ignore sample sizes

3. Base Rate Neglect

When estimating the probability of an event given evidence, people often ignore the event’s base rate (the probability of the event in the absence of any evidence)

Base Rate Neglect (Tversky & Kahneman, 1973)

In 1973, Tversky and Kahneman asked subjects to identify an individual in a group of 100 as an engineer to a lawyer

• The population’s distribution was specified

  • 30/70 condition: 30 engineers, 70 lawyers

  • 70/30 condition: 70 engineers, 30 lawyers 

• A description of an individual is provided (engineer, neutral, lawyer, none)

• Subjects are asked if that individual is an engineer

Experimental Design and Results (Tversky & Kahneman, 1973)

People completely ignore the condition and make their judgments purely on the description they read

Another Example (Tversky & Kahneman, 1982)

In order to make this judgment, you can’t ignore the distribution of blue and green cabs. 

• When you have info about the real world (base rates), you can’t just ignore that

Misdiagnosing Real Diseases (Eddy, 1982)

Chance of breast cancer:

• Age 20-24: 1 in 110,000

• Age 40-44: 1 in 790

• Age 70-74: 1 in 215

Actual false positive rate: 10-33%

• Harvard study in 1998 estimated 33%

There are documented cases where doctors have badly misjudged the probability of breast cancer by ignoring base rates

What’s Wrong?

Base rate neglect is one of the most common mistakes people make while reasoning probabilistically

Much research has investigated why people ignore base rates

1. Causal relevance of base rates in unclear

2. People are not good at reasoning with probabilities vs frequencies

1. Causal Relevance of Base Rates

Tversky and Kahneman (1982) also presented a version of the “cab” problem in which the causal relevance of the base rates was emphasized

2. Base Rates as Frequencies (Gigerenzer, Hell, & Blank, 1988)

Gigerenzer et al. tested whether base rates are used when expressed as frequencies

Experimental Design and Results (Gigerenzer, Hell, & Blank, 1988)

Probabilities vs Frequencies (Gigerenzer, Hell, & Blank, 1988)

Why are frequencies preferable to probabilities? According to Gigerenzer:

• Evolution has prepared people to deal with the frequency of observed events

• In contrast, the notion of probabilities are a relatively recent invention

Summary of Base Rate Neglect

Base rate neglect is one of the most common mistakes people make while reasoning probabilistically

But use of base rates more likely when:

• Individuating info is not present

• Base rates are causally related to judgment

• Base rates expressed as frequencies rather than probabilities

Interim Summary

These phenomena are sometimes referred to as cognitive illusions

• Like perceptual illusions, they’re consistent and persistent discrepancies between a true state of affairs and its mental representation 

Confirmation Bias

An unconscious tendency to process info in a way that supports one’s prior beliefs or values

• There are numerous examples of people clinging to beliefs that should be revised or discarded

Types of Confirmation Bias

1. Biased information search

2. Biased evidence evaluation

1. Biased Information Search

Ideally, the evidence that people seek out to test their beliefs should lead to correct beliefs being adopted

In fact, people tend to seek out only evidence that is likely consistent with their existing beliefs

Real World Biased Search (Shafir, 1993)

In 1993, Eldar Shafir asked subjects to choose which parent should receive custody of a child

• One parent’s had strong positive and negative features, the other’s features were neutral

• The wording of the question was varied

  • Which parent would you award custody?

  • Which parent would you deny custody?

• How would the wording affect subjects’ choice?

Results (Shafir, 1990)

2. Biased Evidence Evaluation

The evaluation of evidence should be independent of whether it agrees or disagrees 

Myside bias: people tend to discount evidence they disagree with

Myside Bias (Lord, Ross, & Lepper, 1979)

In 1979, Lord et al. presented subjects with (functional) studies about controversial topics (ex: capital punishment)

• Subjects had strong prior attitudes

  • Proponents: in favor of capital punishment

  • Opponents: against capital punishment

• Each subject read two studies, one supporting capital punishment, one not

The Two Studies (Lord, Ross, & Lepper, 1979)

Experimental Design (Lord, Ross, & Lepper, 1979)

Results (Lord, Ross, & Lepper, 1979)

Reasons for Confirmation Bias

• Maintenance of self-esteem

• Avoidance of cognitive dissonance

• Simplified cognitive processing

• Backfire effect: disconfirming evidence sometimes leads to a mistaken belief becoming stronger

Summary of Confirmation Bias

People are often biased in how they:

• Test hypotheses: they tend to look for positive confirming evidence

• Evaluate evidence: they’re more critical of evidence they don’t agree with

Everyone should remain open to the possibility that we’re sometimes wrong