Confirmation Bias - University of Waterloo

Download Report

Transcript Confirmation Bias - University of Waterloo

Confirmation Bias
• We tend to ask questions that will confirm our
hypotheses
• Snyder & Swan - told to find out whether someone is an
introvert people tend to ask questions like “what
factors make it hard for you to really open up to
people?” Told to find out whether someone is an
extrovert people tend to ask questions like “What kind
of situations do you seek out if you want to meet new
people?”
• When they let people answer these questions and other
people listened to only the answers they thought the
respondents were like the questions asked
• Base rates (Study 3) and monetary incentives to be
accurate (Study 4) do not eliminate this effect
Snyder & Swan - Confirmation Bias
Studies 1 & 2 combined
8
7
6
5
4
3
2
1
0
Extrovert Hypothesis
Introvert Hypothesis
Number of Extrovert Questions
Number of Neutral Questions
Number of Introvert Questions
The Trope & Bassok Response to
the Snyder & Swan Study
• People would prefer diagnostic questions to the
types of questions Snyder & Swan allowed them
to choose.
• What questions do people prefer in testing
whether someone is an extravert, “Do you shy
away from social interactions?” or “Do you talk
loudly?”
• People preferred, “Do you shy away from social
interactions?”
• Further testing showed that people do prefer
hypothesis confirming questions, but
diagnosticity of the questions is a stronger
Snyder, Tanke, & Berscheid (1977) Behavioral Confirmation
• Men believe women are attractive or unattractive
and then they have a real conversation. The two
sides of the conversation were taped separately.
• Confirmatory Hypothesis Testing - men ask
women questions that elicit physical
attractiveness stereotype (men were more
pleasant to women perceived to be attractive)
• Behavior Confirmation - the men’s behavior
creates confirmation of the stereotype (men’s
attitude toward the women created it’s own
reality)
Self-Fulfilling Prophecies
• Word, Zanna, & Cooper (1974)
– Study 1 - White participants interview Black or White
Confederate
– Study 2 - White confederates trained to interview like
interviewers in Study 1 interview White participants
• Rosenthal & Jacobson - teacher expectancy effects
– In a number of classrooms students are picked out at
random and designated to be late bloomers who are
going to exhibit a large spurt in their learning ability.
– These students did do better and it seems to be
mediated by teacher’s interactions with these students.
Word, Zanna & Cooper - Study 1
14
12
10
8
6
4
2
0
Distance
Interview Length
Speech Errors
Ethnicity of Job Candidate
African-Americans European-Americans
Word, Zanna, & Cooper - Study 2
8
7
6
5
4
3
2
1
0
Distance
Interviewer
Friendliness
Speech Errors
Job Candidate was treated like
African-American European-Americans
Perceiving Causality
• Nisbett & Wilson Studies
– Rasing the Tide
– Power saw experiment
– Which pantyhose do you like?
• Wilson Relationship Studies
– List all the positive and negative qualities of your
partner
– Tell me how much you feel about your partner
– Which predicts better?
Raising the Tide
Percent
Saying Target
Word
25
20
15
10
5
Cued
Not Cued
Perceiving Relationships Between Variables
• 2 X 2 tables and how we don’t use them
• Illusory Correlation
– Chapman & Chapman studies
– Hamilton studies - Stereotype formation through
illusory correlation?
• Failure to Detect Actual Correlation
– Chapman & Chapman studies
– Jennings, Amabile, & Ross - correlation between
walking sticks and height
Hypothesis Testing and 2 X 2 Tables
Occurred
Yes
No
Yes
A
X
No
Y
Z
Predicted
Illusory Correlation
Number of Behaviors
20
15
10
5
0
Group A
Group B
Actual Behaviors
Group A
Group B
Perceived Behaviors
Desirable Behaviors Undesirable Behaviors
When are people accurate?
• People can perceive correlations pretty well if
they are high
• When their theories are right about how the world
works people are fairly accurate
• When they are familiar with the domain - such as
whether they agree with others evaluations of
people they know well
Ratings of ability vs. personality (Kunda &
Nisbett, 1986)
• People are much more accurate in estimating
consistency across situations for abilities than
for personality
• They correctly perceived that people’s abilities
vary across situations (someone who is smart in
math may not be in English)
• They incorrectly overestimated the consistency
of people’s personality (someone who is shy in
class will be shy at parties)
How Do We Make More Complex
Judgements with Many Variables?
• We understand things better when they are
presented in a storyline - Story model of jury
decision making (Pennington & Hastie studies)
– People spontaneously construct stories when
viewing a trial
– How easy or difficult it is to construct a story
influences the verdict that juries make.
• We like stories best if they have
– Explanatory breadth
– Simplicity
– Extent to which they can be explained by other
information
Counterfactual Thinking What Caused This to Go Wrong?
• Imagine you get to the airport and you missed
your plane
– By an hour or by five minutes
– What would make you more upset?
• Anticipatory Regret
• Notion of Normality and Mental Models
– Closeness of the Counterfactual to the Actual
Event- Airplane Example, Olympic Medals
– Exception vs. Routine - exception causes more
counterfactual thinking
– Controllability
Other Factors that Affect Our
Counterfactual Thoughts
• Action vs. Inaction - do you feel worse if you did
something or didn’t do something?
– Black Jack (Miller & Taylor, 1995)
– The importance of time (Gilovich & Medvec, 1995) what we regret more depends on when it occurred
• Ease of Mental Replication (Miller, Turnbull, &
McFarland, 1989) - Stealing cookies from the cookie
jar
• Functions of Counterfactuals
– May prevent future problems
– Let us feel better by imagining how things could have