Transcript Chap10b
Cognitive Processes
PSY 334
Chapter 11 – Judgment and
Decision-Making
Inductive Reasoning
Processes for coming to conclusions that
are probable rather than certain.
As with deductive reasoning, people’s
judgments do not agree with prescriptive
norms.
Baye’s theorem – describes how people
should reason inductively.
Does not describe how they actually
reason.
Baye’s Theorem
Prior probability – probability a
hypothesis is true before considering the
evidence.
Conditional probability – probability the
evidence is true if the hypothesis is true.
Posterior probability – the probability a
hypothesis is true after considering the
evidence.
Baye’s theorem calculates posterior
probability.
Burglar Example
Numerator – likelihood the evidence
(door ajar) indicates a robbery.
Denominator – likelihood evidence
indicates a robbery plus likelihood it
does not indicate a robbery.
Result – likelihood a robbery has
occurred.
Baye’s Theorem
H
~H
E|H
likelihood
likelihood
likelihood
robbery
E|~H likelihood
of being robbed
of no robbery
of door being left ajar
during a
of door ajar without robbery
P( E | H ) P( H )
P( H | E )
P( E | H ) P( H ) P( E |~ H ) P(~ H )
Baye’s Theorem
P(H) = .001
P(~H) = .999
P(E|H) = .8
P(E|~H) = .01
from police statistics
this is 1.0 - .001
Base rate
P( E | H ) P( H )
P( H | E )
P( E | H ) P( H ) P( E |~ H ) P(~ H )
(.8)(.001)
P( H | E )
.074
(.8)(.001) (.01)(.999)
Base Rate Neglect
People tend to ignore prior probabilities.
Kahneman & Tversky:
70 engineers, 30 lawyers vs 30 engineers,
70 lawyers
No change in .90 estimate for “Jack”.
Effect occurs regardless of the content of
the evidence:
Estimate of .5 regardless of mix for “Dick”
Cancer Test Example
A particular cancer will produce a
positive test result 95% of time.
If a person does not have cancer this gives
a 5% false positive rate.
Is the chance of having cancer 95%?
People fail to consider the base rate for
having that cancer: 1 in 10,000.
Cancer Example
Base rate
P(H) = .0001
P(~H) = .9999
P(E|H) = .95
P(E|~H) = .05
likelihood of having cancer
likelihood of not having it
testing positive with cancer
testing positive without cancer
P( E | H ) P( H )
P( H | E )
P( E | H ) P( H ) P( E |~ H ) P(~ H )
(.95)(.0001)
P( H | E )
.0019
(.95)(.0001) (.05)(.9999)
Conservatism
People also underestimate probabilities
when there is accumulating evidence.
Two bags of chips:
70 blue, 30 red
30 blue, 70 red
Subject must identify the bag based on the
chips drawn.
People underestimate likelihood of it
being bag 2 with each red chip drawn.
Probability Matching
People show implicit understanding of
Baye’s theorem in their behavior, if not in
their conscious estimates.
Gluck & Bower – disease diagnoses:
Actual assignment matched underlying
probabilities.
People overestimated frequency of the
rare disease when making conscious
estimates.
Frequencies vs Probabilities
People reason better if events are
described in terms of frequencies
instead of probabilities.
Gigerenzer & Hoffrage – breast cancer
description:
50% gave correct answer when stated as
frequencies, <20% when stated as
probabilities.
People improve with experience.
Judgments of Probability
People can be biased in their estimates
when they depend upon memory.
Tversky & Kahneman – differential
availability of examples.
Proportion of words beginning with k vs
words with k in 3rd position (3 x as many).
Sequences of coin tosses – HTHTTH just
as likely as HHHHHH.
Gambler’s Fallacy
The idea that over a period of time things
will even out.
Fallacy -- If something has not occurred
in a while, then it is more likely due to
the “law of averages.”
People lose more because they expect
their luck to turn after a string of losses.
Dice do not know or care what happened
before.
Chance, Luck & Superstition
We tend to see more structure than may
exist:
Avoidance of chance as an explanation
Conspiracy theories
Illusory correlation – distinctive pairings are
more accessible to memory.
Results of studies are expressed as
probabilities.
The “person who” is frequently more
convincing than a statistical result.
Decision Making
Choices made based on estimates of
probability.
Described as “gambles.”
Which would you choose?
$400 with a 100% certainty
$1000 with a 50% certainty
Utility Theory
Prescriptive norm – people should
choose the gamble with the highest
expected value.
Expected value = value x probability.
Which would you choose?
A -- $8 with a 1/3 probability
B -- $3 with a 5/6 probability
Most subjects choose B
Subjective Utility
The utility function is not linear but
curved.
It takes more than a doubling of a bet to
double its utility ($8 not $6 is double $3).
The function is steeper in the loss region
than in gains:
A – Gain or lose $10 with .5 probability
B -- Lose nothing with certainty
People pick B
Framing Effects
Behavior depends on where you are on
the subjective utility curve.
A $5 discount means more when it is a
higher percentage of the price.
$15 vs $10 is worth more than $125 vs
$120.
People prefer bets that describe saving
vs losing, even when the probabilities
are the same.