DANCEST 805 Lecture Notes - Lecture 5: Representativeness Heuristic, Conjunction Fallacy, Daniel Kahneman
Task 5 – Decision Making
Introduction to Decision – Making
Judgement
= Individuals use various cues to draw inferences about situations and events
• Evaluated in terms of accuracy
• Component of larger decision-making process→ concerned with assessing, estimating, inferring what events will
occur and what decision maker’s evaluative reactions to those outcomes will be
Decision-Making
= Individuals choose amongst various options.
• Evaluated in terms of consequences of decisions
• Involves problem-solving – individuals try to make best possible choice from options
• Is not the same as problem solving (PS)
1) PS involves generating own options, DM involves options that are already present
2) 2) DM concerned with preferences, PS concerned with solutions
Judgement
• We often change our opinion of the likelihood of something based on new information
→ Strength of our beliefs often increased or decreased by new information
• Bayes’ theorem – probability of event, based on prior knowledge of conditions that might be related to the event
→ focused on situations in which there are two possible beliefs or hypotheses (e.g., X is lying vs. X is not lying), and showed how new data
or information change the probabilities of each hypothesis being correct
→ need to assess the relative probabilities of the two hypotheses before the data arc obtained (prior odds). We also need to calculate the
relative probabilities of obtaining the observed data under each hypothesis (likelihood ratio).
→ Bayesian methods evaluate the probability of observing the data, D, if hypothesis A is correct, written p (D/HA), and if hypothesis B is
correct, written p (D/HB)
→ Left side: relative probabilities of hypotheses A and B in the light of the new
data (probabilities we want to work out)
→ Right side: prior odds of each hypothesis being correct before the data were
collected multiplied by the likelihood ratio based on the probability of the data
given each hypothesis
→ Taxi-cab problem (Kahneman & Tversky) – taxicab is involved in a hit-and-run accident one night; 85% taxicabs belong to green
company & 15% to blue company; due to prevalence of cabs from separate companies, the probability taxi was blue is 41% and
probability it was green is 59% (when combined with probability derived from ability of eye-witness to identify cabs under appropriate
visibility conditions)
Neglecting base rates
• Base-rate information – relative frequency with which an event occurs, or an attribute is present in the population
→ People take less account of this than they should – e.g. in taxicab example, base-rate information about relative numbers of Green
& Blue cabs was ignored, focused mainly on evidence of witness: 80% of the time she was right in identifying cab colour, so they
said probability cab is blue is 80%
→ We fail to utilize base-rate information because we use heuristics (cognitively un-demanding)
• Tversky and Kahnemann: Most people given judgement tasks make considerable use of rules of thumb or heuristics (reduce effort)
• Intelligence and cognitive ability – unrelated to performance on most judgement tasks
• Fast-and-frugal heuristics – involve rapid processing of relatively little information
• base-rate information is sometimes both relevant and generally used
→ we possess valuable causal knowledge that allows us to make accurate judgements using base-rate information in everyday life. ln
the laboratory → however, the judgement problems we confront often fail to provide such knowledge
→ breast cancer experiment: if another cause was mentioned, participants were far more likely to take full account of the base-rate
information → corresponds to real life
• many people use base-rate information when they understand the underlying causal factors
• we also use base-rate information when strongly motivated to do so → when it is advantageous
Representativeness
heuristic
= assumption that an object or individual belongs to a specified category because it is representative (typical) of
that category
• Assumption that representative or typical members of a category are encountered most frequently e.g. when
given description of individual and asked to estimate probability they have a certain occupation; you would
do it in terms of the similarity between the individual’s description & your stereotype of the occupation
• Conjunction fallacy – mistaken belief that combination of two events (A and B) is more likely than one of the
events on its own
→ still found even when almost everything possible is done to ensure participants interpret the
problem correctly
Availability heuristic
= Assumption that frequencies of events can be estimated accurately based on how easy or difficult it is to retrieve
relevant information from long-term memory e.g. causes of death attracting publicity (murder) are judged to be
more likely than those than do not (suicide), even when opposite is the case
• sometimes overridden by deliberate thought
→ availability heuristic when under cognitive load
• Availability-by-recall mechanism – based specifically on number of people an individual recalls having died
from given risk
• Affect-heuristic – using one’s own emotional responses to influence rapid judgements or decisions
• Fluency mechanism – involves judging number of deaths from given risk by deciding how easy it would be to
bring relevant instances to mind but without retrieving them
• Support theory – any given event will appear more or less likely depending on how it is described (more
explicit description of event has greater subjective probability than same event described in less explicit
terms); influences even expert’s judgements
• Kahneman and Tversky showed several general heuristics or rules of thumb (e.g., representativeness heuristic, availability heuristic)
underlie judgements in many different contexts
→ Instrumental in establishing the field of judgement in research
→ Ideas hugely influential within psychology but also other fields
→ Plenty of evidence that most people prefer to minimize the cognitive demands on them by using heuristics
• Reasons to use heuristics:
→ Advantage of speed – we produce approximately correct judgements very rapidly
→ Robust because they can be used regardless of limited usefulness when information is sparse
→ We don’t like thinking hard if we can avoid it
→ Does not make sense to devote so much effort to making precise judgements in rapidly changing world
• Limitations of heuristics-and-biases approach:
→ heuristics identified by Kahneman and Tversky are vaguely defined
→ theorizing based on the heuristics-and-biases approach has been limited
→ Heuristics is used in many different ways by different researchers and is in danger of losing its meaning
→ Difficult to explain how effort is reduced while using heuristics
→ Sometimes unfair to conclude people’s judgements are biased and error-prone
→ Errors occur because participants misunderstand problem
→ Focuses on biased processing, but problem is often with whether people have quality information available
→ People make correct and approximately accurate judgements
→ Research is artificial and detached from realities of everyday life – emotional and motivational factors have not been assessed until
recently
• Limitations of fast-and-frugal heuristics
→ Major heuristics are used much less often than predicted theoretically
→ Some heuristics are not as simple as researchers have claimed – e.g. difficult to organize cues hierarchically in terms of validity in take-
the-best heuristic
→ De-emphasizes importance of decision in question when applied to decision-making – e.g. women want to consider all relevant
evidence before choosing one of two men to marry rather than stopping after a single discriminatory cue
→ Unable to specify conditions under which certain heuristics are selected over others
Judgement theories
Support theory
• Tversky and Kochler (1994) put forward their support theory based partly on the availability heuristic
• Key assumption: an event appears more or less likely depending on how it is described
→ Must distinguish between events themselves and the description of those events
• More explicit event descriptions have greater subjective probability for two main reasons;
1) An explicit description often draws attention to aspects of the event less obvious in the non-explicit description
2) Memory limitations may prevent people remembering all the relevant information if it is not supplied
• Let participants judge the risk of a terrorist attack over the following six months vs. judging the probability of an attack plotted by al-Qaeda
or not plotted by al-Qaeda
→ overall estimated probability of a terrorist attack was greater (0.30 + 0.18) when the two major possibilities were made explicit
than when they were not (0.30)
• doctors also showed the effect → subjective probabilities were higher for explicit descriptions
• explicit description can reduce subjective probability if it leads us to focus on low-probability causes
• providing an explicit description can reduce subjective probability by making it more effortful to comprehend an event
• received empirical support
• shows how the availability heuristic can lead to errors in judgement
• limitations:
→ precise reasons why providing an explicit description generally increases an event's subjective probability are not clear
→ explicit descriptions can reduce subjective probability if they lead individuals to focus on low-probability causes
→ explicit descriptions can also reduce subjective probability if they are hard to understand
→ theory is oversimplified
it assumes the perceived support for a given hypothesis provided by relevant evidence is independent of the rival
hypothesis or hypotheses → however, people often compare hypotheses and so this independence assumption is
incorrect
Fast-and-frugal heuristics
• Rapid processing of little information
• take-the-best heuristic is a key fast-and-frugal heuristic
→ Take the best, ignore the rest e.g. deciding whether Cologne or Herne has larger population: assuming most valid cue is whose city
names you recognize; if both are recognized, you use another cue such as whether city has cathedral
• Take-the-best strategy has 3 components:
1) Search rule – search cues (name recognition, cathedral) in order of validity
2) Stopping rule – stop after finding a discriminatory cue (cue applies to only one of possible answers)
3) Decision rule – choose outcome
• Recognition heuristic – using knowledge that only one out of two objects is recognized to make judgement: recognized object has higher
value with respect to criterion (least cognitively demanding of all heuristics)
→ important
• Effective in spite of simplicity
• Which heuristic to use on judgement?
→ a two-step process
→ First, the nature of the task and individual’s memory limit the number of available heuristics
→ Second, people select one of them based on the likely outcome of using it and its processing demands
• Individuals with little knowledge can sometimes outperform those with more knowledge
• These heuristics surprisingly effective in spite of their simplicity
• Individuals with little knowledge can sometimes outperform those with greater knowledge
• Use of the recognition heuristic more complex than assumed
→ people generally also consider why they recognize an object and only then decide whether to use the recognition heuristic
• using the take-the-best heuristic is also more complex than suggested by Gigerenzer
→ heuristic requires us to organize the various cues hierarchically based on their validity
→ often do not have sufficient knowledge of cue validities
• when the approach is applied to decision making it de-emphasizes the importance of the decision
→ decision making may stop after finding a single discriminatory cue when deciding
→ most consider more
Findings
• Selecting the recognized object does not necessarily mean the recognition heuristic was used – it could have been chosen for other reasons
• People are much more likely to use the recognition heuristic when it is valid than when it is not
→ Correlation of +0.64 between validity and use of this heuristic
• Simple heuristics can be moderately effective
→ Outperform judgements based on much more complex calculations
• Hiatus heuristic – the rule of thumb that only customers who have purchased goods fairly recently remain active customers
→ hiatus heuristic correctly categorized 83% of customers, whereas the comparable figure for the complex model was only 75%
→ less-is-more effect – the approach based on less information was more successful
• recognition heuristic often not used when information was inconsistent (airport city experiment)
• take-the-best strategy used less often than predicted theoretically
→ only 33% used all 3 components of the strategy
• less likely when cost of information are low and cue validities unknown
→ more detailed processes then
• more intelligent participants more likely to use the take-the-best strategy when it was the best one
• factors determining which processing strategy used on judgement task
→ information redundancy – redundancy high when different cues provide similar information but low when cues provide different
information
simple heuristics better in situations with high redundancy
→ simple strategies work best when environmental information is simple
→ take-the-best strategy was more likely to be used than a more complex strategy with high information redundancy but the
opposite with low information redundancy
Natural frequency hypothesis
• Evolutionary account of strengths and weaknesses of human judgements
• Natural sampling – process of encountering instances in a population sequentially
• Gigerenzer and Hoffrage (1999) claimed our evolutionary history makes it easy for us to work out the frequencies of different kinds of
events
→ Ill-equipped to deal with factions and percentages → poor at base rates
• Judgement improves when problems use frequencies – as a result of our evolution, we find it easy to work out frequencies of different kind
of events, but we find it difficult to deal with fractions and percentages