Making good decisions with simple heuristics

1
Making good decisions with simple heuristics
Peter M. Todd IU Informatics, Cognitive Science,
and Psychology Center for Adaptive Behavior and
Cognition, MPI, Berlin
Max Planck Institute for Human Development Berlin
2

• Which US city has more inhabitants,
• San Diego or San Antonio?

3

• Which US city has more inhabitants,
• San Diego or San Antonio?

Americans 62 correct
Germans ? correct
Germans 100 correct
...but not any more!
4
The Recognition Heuristic
Definition When deciding which of two objects
is greater on some criterion, if one object is
recognized and the other is not, then infer that
the recognized object has the higher value.
But when, and why, could it possibly work?
Ecological rationality....
5
How to be rational, part 1

• To make a rational decision, e.g. between which
  of two alternatives is greater in some way
• In a perfectly predictable world
• find perfect predictor (cue) and use it
• In our world, surrounded by uncertainty, we could
  (?)
• gather all available (uncertain) cues
• use laws of probability to combine them
• e.g. choose alternative with highest value
• S (cue-present?) (cue-importance)

6
How to be rational, part 2

• We can still be rational without considering all
  possible information and alternatives, if
• we measure costs (inc. opportunity costs) of
  looking up info and alternatives, and benefits of
  considering more
• we add these costs and benefits into our expected
  gain from making the decision
• we optimize our expected gain under these
  constraints
• But optimization under constraints can require
  even more mental computation and effort

7
How to be rational, part 3

• Given our bounded rationality (Simon, Selten)
• Human memory, cognitive capacity, and time are
  limited, i.e. bounded
• Moreover, most real-world problems cannot be
  solved by optimization
• The only plausible/possible approach is
• to make decisions within our bounds by using
  approximate methods heuristics, shortcuts, rules
  of thumb, e.g. based on recognition, satisficing
• We cant use perfect info, and we cant use all
  info, so we have to search for info to use

8
Visions of Cognition
Superhuman
BoundedRationality
Unbounded Rationality
Optimization under Constraints
9
The adaptive toolbox
A collection of decision mechanisms, including
simple heuristics and rules... Some pre-wired,
some learned... Applicable to specific decision
tasks and in particular domains... That rely on
and exploit the structure of information in the
environment... To yield ecological rationality.
10
What bounds our bounded rationality?

• Bounds are not primarily
• limited memory
• limited computational power
• counterexamples (e.g. savants, sensory systems)
  show these bounds could be gotten around if
  benefits outweighed costs

11
Bounds from nature of the world

• Because the world is Decisions must
  be
• uncertain
  robust ? simple
• competitive
  fast ? frugal
• External environment not only bounds decision
  making, it also enables it
• Simplicity and frugality are achieved by
  exploiting the structure of environments, letting
  the world do some of the work to
  produce ecological rationality

12
Bounded/Ecological Rationality
Ignorance-based decision making
Satisficing
One-reason decision making
Elimination heuristics
Gigerenzer, Todd the ABC Research Group (1999).
Simple Heuristics That Make Us Smart. Oxford
University Press. Gigerenzer Selten (Eds.)
(2001). Bounded Rationality The Adaptive
Toolbox. MIT Press.
13
Bounded/Ecological Rationality
Ignorance-based decision making
Satisficing
One-reason decision making
Elimination heuristics

• Recognition Heuristic

• Categorization by Elimination
• Elimination by Aspects
• QuickEst

• Fixed Aspiration Levels
• Dynamic Aspiration Levels

• Take The Best

• Take The Last
• Minimalist

• Innovation by Ignorance

• Fast and Frugal Classification

Gigerenzer, Todd the ABC Research Group (1999).
Simple Heuristics That Make Us Smart. Oxford
University Press. Gigerenzer Selten (Eds.)
(2001). Bounded Rationality The Adaptive
Toolbox. MIT Press.
14
Where to find/study heuristics

• Important decision problems
• selection, estimation, categorization...
• Important adaptive domains (can be social and
  also include cultural influences)
• survival reproduction
• resources health habitat mate choice
  raising offspring
• foraging exchange

15
How to make simple heuristics

• Combinations of building blocks for
• guiding search (for information or options)
• stopping search
• making a decision based on results of search
• Combinations of other heuristics
• all chosen to take advantage of available
  structure of information in the environment

16
Aspects of environment structure

• Distributions of cues and object values
• Costs of information and time pressure
• Location of information (internal/memory,
  external)
• Cues available simultaneously?
• Objects available simultaneously?
• Representation of information
• e.g. probabilities or frequencies
• Representation of problems
• e.g. list of cues or decision tree

17
Searching for info to make decisions

• Traditional rational approach
• gather all available cues, weight and combine
• Simplest alternative
• random choiceno info search needed
• default choice (e.g. organ donation)no other
  info needed
• Simple, smarter alternative
• use recognition heuristic (if you can)
  search only for recognition info about options

18
Ignorance-based decision making the Recognition
Heuristic
19
The Recognition Heuristic
Definition When deciding which of two objects
is greater on some criterion, if one object is
recognized and the other is not, then infer that
the recognized object has the higher value.
Ecological rationality The recognition heuristic
is successful when the environment is structured
such that recognition (and ignorance) is
systematic rather than random, i.e., when
recognition (or lack of) correlates with the
criterion.
20
Recognitions building blocks
Search search for recognition information about
objects to choose between Stop stop search as
soon as recognition is assessedno further
information sought Decision select recognized
object, or guess All built on much more complex
underlying machinery, distilled into recognition
21
A test of the recognition heuristic

• 22 American students tested on pairs of 25 or 30
  largest German cities
• which city is larger, A or B?
• Results plotted number of cases where each person
  could have used recognition heuristic, and number
  of cases where they did choose in accordance with
  heuristic

22
Do people use the recognition heuristic?
250
200
Number of testsin which therecognition
heuristic couldbe applied
150
Number of Tests
100
Number of testsconsistent with the
recognitionheuristic
50
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Individual Participants
Mean (median) use of recognition heuristic 90
(93)
Goldstein Gigerenzer (2000)
23
Less-is-more effects in individuals (and in
groups)
correct
Middle Triplet
Eldest Triplet
Middle Sister
Eldest Sister
Youngest Sister / Triplet
Number of cities recognized
24
The Less-is-More Effect
The expected proportion of correct inferences c is
where
is the number of recognized objects is the total
number of objects is the recognition validity is
the knowledge validity
n
N
a
ß
½ is the proportion correct from guessing
A less-is-more effect occurs when
a gt ß
25
Where does recognition come from?

• Recognition comes from exposure to mediators that
  mention objects (e.g. newspapers,
  other people)

Accessible Environment
Mediator
Surrogate correlation
Ecological correlation
Unknown Environment
Mind
INFERENCE
Criterion
Recognition
Recognition validity
26
Recognition heuristic exploits environment
structure
27
Ignorance can be informative.
Why?
When ignorance is systematic, not random, it
actually carries information
28
(No Transcript)
29
Paying for the
name.
30
(No Transcript)
31
Recognition use in group decisions

• How do groups reach consensus when some members
  lack knowledge?
• Experiment
• 84 Berlin students, all asked to recognize and
  compare pairs of 40 U.S. cities (overall
  recognition validity.72, knowledge.65)
• Then divided into groups of 3 and asked to reach
  consensus on further city pairs
• Deliberations videotaped answers modeled on 2940
  decisions made

32
Models of group decisions used

• Simple majority rule
• Recognition-based majority rule
• Knowledge-based majority rule
• Recognition-then-knowledge lexicographic rule
  (recognition-first)
• Knowledge-then-recognition lexicographic rule
  (knowledge-first)

33
Do groups use recognition?
34
Who do groups listen to?

• Look at decisions in those groups where
  recognition-based choice differs from
  knowledge-based choice
• 1 rec, 2 know 59 choices go with rec
• 2 rec, 1 know 76 choices go with rec
• 1 rec, 1 know 61 choices go with rec
• Groups sometimes give up on majority and go with
  most accurate individual (rec.-user)

35
Less-is-more effects in groups?
36
Using the Recognition Heuristic in Politics

• Julian Marewski
• Wolfgang Gaissmaier
• Anja Dieckmann
• Lael Schooler
• Gerd Gigerenzer
• Max Planck Institute for Human Development,
  Berlin

37
Idea of the Study
Can a simple strategy like
the recognition heuristic be used
to forecast the outcomes
of political elections?
38
Introducing the Election Setting
Setting 2004 Elections to the Parliament of the
German Federal State of Brandenburg

• First Vote
• Candidates

• Second Vote
• Political Parties

Parliament of the German Federal State of
Brandenburg
39
Questionnaire Design
Candidates Recognition Test

?
?
Candidates Prediction Task (Paired Comparisons)
40
1. Can the Recognition Heuristic Be Useful to
Predict Election Outcomes?
Candidates
Newspaper namings
.45 - .61
.51 - .77
Ecological Correlation
Surrogate Correlation
Recognition Correlation
Name Recognition
Election Outcome
.36 - .40

Measure of strength of Association Gamma
The media may be responsible for the association
between recognition and election outcomes.
41
2. Is it Possible to Predict Election Outcomes
Using Citizens Collective Name Recognition?
Accuracy 70 (Candidates)
Accuracy 80 (Parties)
Citizens Recognition
Citizens collective recognition of party and
candidate names predicts the election outcomes
well.
42
3. Are Citizens Who Recognize Fewer Candidates
Better at Predicting Elections?
90
Dots Means
Bars 95 Confidence Intervals
80
n31
n10
n32
n33
n33
70
n6





Total Proportion of Correct Election Forecasts

Citizens with less recognition knowledge derived
as accurate predictions as citizens with more
recognition knowledge.
60

n27
50
40
0
33
66
100
Percentage of Recognized Candidates
43
When to go against recognition?

• Use anti-recognition (take the unrecognized
  option) when
• you want to explore (and the environment is safe,
  so you can use trial-and-error learning)e.g.
  food/restaurant choice, trying Stinking Bishop
  cheese
• there could be profit from taking unknown things
  (e.g., contrarian stock investment)
• But this still just requires simply searching for
  recognition info about options

44
Searching for more information?

• When should you rely on just (anti/)recognition,
  and when should you search for more info?
• Answer stick with recognition when it has high
  validity compared to other possible
  infootherwise, search for more (better) info
• Validity how often a cue tells you the right
  answer (or your preferred choice) by itself
• validity ( right answers) / (right
  wrong)

45
Searching for info to make decisions

• Traditional rational approach
• gather all available cues, weight and combine
• Simplest alternative
• random choice no info search
• default choiceno other info needed
• Simple, smarter alternative
• use recognition heuristic (if you can)
  search for recognition info only
• If you cant use recognition use (and search
  for) as few cues as you can get away with

46
Which city is bigger?

• Dresden
  Leipzig
• Cues
• Recognized? yes
  yes
• National capital no no
• Insect restaurant yes no
• Soccer team no yes
• Intercity train yes yes
• State capital no yes
• University yes yes

47
Insect restaurant cue
48
Insect restaurant cue
Dishes at the Espitas restaurant in Dresden
include maggot ice cream, maggot salads and
maggot cocktails. The restaurant is importing
the "nutritious and extremely tasty" maggots from
Mexico.
49
Thats why its called a heuristic

• (sometimes even excellent cues can lead you
  astray Leipzig has 506,000 inhabitants, and
  Dresden, despite its insect restaurant, has only
  504,000....)

50
Where is information found?

• In the external world (including on paper, in an
  experiment, on the Web)
• In memory, including recognition
• In others, e.g. social networks, groups
• In culture/institutions
• In combinations of these
• rats (use recognition memory, but from others)
• restaurant-goers (use info from others and from
  own culture)

51
Flowchart of one-reason DM
52
(No Transcript)
53
Building blocks for one-reason decision making

• Search many possibilities, e.g. validity
• Stop on first cue that discriminates
• Decision select object indicated by cue that
  stopped search

54
One-reason decision making Take The Best
55
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
56
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
57
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
stop search
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
58
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
59
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
60
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
61
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
62
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
63
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
64
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
65
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
stop search
66
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
67
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
68
Which city has a higher homelessness rate?
LosAngeles
NewYork
NewOrleans
Chicago
Rent control
1
1
0
0
stop search
Vacancy rate
1
1
0
0
Temperature
1
1
1
0
Unemployment
1
1
1
1
Poverty
1
1
1
1
Public housing
1
0
0
1
69
How accurate is Take The Best?

• Performance tested across 20 environments
• Homelessness rates (50 cities in the United
  States)
• Attractiveness judgments of famous men and women
• Average motor fuel consumption per person (all
  states in U.S.)
• Rent per acre paid (58 counties in Minnesota)
• House price (Erie, Pennsylvania)
• Professors' salaries (at a midwestern college)
• Car accident rate per million vehicle miles
  (Minnesota highways)
• High school drop-out rates (all high schools in
  Chicago)

Czerlinski, Gigerenzer, Goldstein (1999)
70
Take The Bests performance across 20 real-world
data sets
  Czerlinski, Gigerenzer, Goldstein
(1999) e.g., professor salaries, homelessness
rates, school drop-out rates
71
Generalization performance on 20 problems
75
NO TRADE OFF
Take The Best
Linear Rule(unit weight)
70
MultipleRegression
Predictive Accuracy
65
TRADE OFF
60
6
7
7.7
0
1
2
3
4
5
Frugality
72
Martignon Hoffrage (1999)
73
When do people decide with 1 reason?

• First answer Not always!
•    not when info is presented all at once
•    not when mistakes are costly
•    not when decisions must be justified
• But do stop and use one reason when
•     information must be searched for
•    time pressure, info takes time to find or
  is costly
•     mistakes dont matter much
•     no need for accountabilityor if its
  easier to
• explain one important cue than many
  weights
•     moral decisions (protected values)

74
Designing institutions for one-reason decision
makers

• If institutional rules are lexicographic, then
  people can reason with them quickly and
  transparently
• Examples
• traffic right-of-way decisions
• soccer championship advancement decisions

75
Institutions to obscure decisions
Old-fashioned slot machine, where chances of
getting a triple can be inferred from the wheels,
vs. new computerized machine, where chances are
independent from whats shown
76
What are the governance implications?

• Is use of heuristics by individuals reassuring or
  worrying? What about by governments?
• Do we need paternalistic policies to overcome (or
  work with) peoples heuristics?
• How can environments be structured to help people
  make better choices for better self-governance?
  Through transparency (lexicographic traffic
  rules), or nudges (organ donation defaults),
  or...?

77
Conclusions Cognition the fast and frugal way

• The minds adaptive toolbox
• contains simple heuristics...
• built from building blocks for search, stopping
  and decision...
• that exploit the structure of information in the
  environment...
• to produce good decisions and yield ecological
  rationality with little information or
  processing.
• People use such heuristics in a variety of
  domains, and their decision making can be aided
  by taking this into account

78
For more information...

• Gigerenzer, Todd, the ABC Research Group
  (1999). Simple Heuristics That Make Us Smart.
  Oxford University Press.
• Gigerenzer Selten (Eds.) (2001). Bounded
  Rationality The Adaptive Toolbox. MIT Press.
• Gigerenzer (2002). Calculated Risks. Simon
  Schuster.
• Gigerenzer (2007). Gut feelings. Simon
  Schuster.
• Todd, Gigerenzer, the ABC Research Group
  (2009). Ecological rationality Intelligence in
  the world. Oxford.
• Me pmtodd_at_indiana.edu
• The ABC group www.mpib-berlin.mpg.de/abc
• ABC-West at IU www.indiana.edu/abcwest

79
(No Transcript)
80
Studying ecological rationality

• Goal Explore how simple mental mechanisms can
  exploit environment structure to yield adaptive
  behavior (e.g. ABC Research Group, 1999)
• Identify important decision problems
• Analyze structure of decision environment
• Propose specific simple heuristics
• Use simulation to see if they can work
• Use analysis to find when/where they can work
• Test experimentally/empirically for evidence that
  they really are at work

81
Types of environment structure

• Aspects of criterion values of objects
• uniform, normal, J-shaped distributions
• Aspects of cue dimensions
• many vs. few (scarcity compared to of objects)
• distribution of importance, validity, etc.
• positive/negative intercorrelation (friendly vs.
  unfriendly environments)
• Aspects of cue values
• binary/continuous
• uniform, normal, J-shaped distributions, rarity

82
Idea of the Study (2)

• We predict the election outcome using potential
  voters collective recognition of names
• Potential voters predict the election outcomes
• Analyses of the election environment to
  understand why recognition can be used for
  election forecasts

Prediction with recognition
Election Outcomes Candidates Parties

Public Naming Candidates Parties
Election Outcomes Candidates
Parties
Recognition Candidates Parties

83
Design of the Study

• I.
• Questionnaire Study
• N 172 citizens of at least 18 years of age
• Living in the voting districts of data collection
• Recruitment 1 day and 2 weeks before election
• II.
• Environmental Analyses
• Counting the number of articles including
  candidate and party names published in 2 local
  newspapers within 3 months before the election

84
Questionnaire Design (1)

• 1. A) Recognition tests for 11 candidates
• B) Prediction tasks for 11 candidates
• (paired comparisons)
• 2. A) Recognition tests for 15 parties
• B) Prediction tasks for 15 parties
• (rank ordering ? computing paired comparisons
  from rank order)

Counterbalancing of tests and tasks,
Randomized item orders with tests and tasks
Counterbalancing of tests and tasks,
Randomized item orders with tests and tasks
85
Design Questionnaire (3)

Parties Prediction Task (Ranking)
86
Environmental Analyses Counting the Number of
Articles Published in two Newspapers
Candidates Number of Mentions in Newspapers
Parties Number of Mentions in Newspapers
87
Research Questions
1. Why can the recognition heuristic be useful to
predict election outcomes? 2. Is it
possible to predict the election outcomes using
citizens collective recognition of candidate and
party names? 3. Can citizens predict the
election outcomes by relying on the recognition
heuristic? 4. Is there evidence that citizens
rely on the recognition heuristic to predict
election outcomes? 5. Are citizens better off
when they do not rely on the recognition
heuristic to predict election outcomes? 6. Are
citizens who recognize fewer candidates better in
predicting the election outcomes than citizens
who recognize more candidates?
Public Naming
The recognition heuristic
can be used to predict election outcomes.
Positive correlations
Election Outcomes
88
1b. Why Can the Recognition Heuristic Be Useful
to Predict Election Outcomes?
Political Parties
Measure of strength of Association Gamma
The media may be responsible for the association
between recognition and election outcomes.
89
Societal Risks, Causes of Death, and Diseases
How (and How Accurately) People Assess Them

• Thorsten Pachur
• Ralph Hertwig, Stephanie Kurzenhäuser
• Max Planck Institute for Human Development, Berlin

90
How Well Can People Assess Health Risk
Frequencies?

• Public has a skewed perception of relative
  mortality rates. (Frost, Frank Maibach, 1997)
• Do people know which risks lead to many deaths
  and which risks lead to few? They do not. In
  fact, they make huge blunders. (Sunstein, 2001)

• Lichtenstein, Slovic, Fischhoff, Layman and Combs
  (1978) In their studies participants judged
  the frequencies of a set of 41 causes of death
  (e.g., car accident, lightning, homicide,
  emphysema)
• People do not have accurate knowledge of the
  risks they face.
• Frequency judgments were given both in a pair
  comparison task and direct frequency estimation
  task.

91
Hypothesized "Process" Underlying the Judgment
Inaccuracies

• Lichtenstein et al. Availability heuristic
  (Tversky Kahneman, 1973)
• But
• Availability heuristic only post-hoc
  explanation, not tested based on predictions
• Availability heuristic rather vague concept
  (Betsch Pohl, 2002 Gigerenzer, 2000)
• So new mechanism Social circle heuristic, with
  precise search, stopping, and decision rules

92
Which risk is more prevalent in population? A or
B
- -
Recognition
-
Choose the alternative to which the cue points

Oneself
-
- -
Guess

Family
-
Social circle
- -

Friends
-
- -

Acquaintances
-
- -

- -/
Other cues
-
93
How much choice is beneficial? (Iyengar Lepper,
2000)
24 alternatives
6 alternatives
40 of customers stopped
60 of customers stopped
30 of them bought
3 of them bought
94
A further choice study

• Iyengar/Lepper (2000)
• People given choice of 1 from 6 or from 30
  chocolates to sample, then rated enjoyment of
  process and satisfaction with choice
• Results
• Extensive-choice (30-option) process was more
  enjoyable and difficult
• Limited-choice (6-option) selection was more
  satisfying and tasty
• Limited-choice setting led to more purchasing
  (taking chocolate rather than money as payment)

95
More alternatives are NOT always better! Why?
Making a decision is more difficultwhen there
are too many alternatives.
96
How much thinking is useful?(Wilson Schooler,
1991)

• Subjects tasted/rated 5 strawberry jams, with or
  without analyzing reasons
• Ratings compared with expert opinions

97
Results of the jam tasting study(Wilson
Schooler, 1991)

• Analysis of reasons lowered liking of jams
• Unanalyzed ratings followed expert opinions
  (r.55) analyzed ratings did not (r.1)
• Preference shifts can be temporary, but still
  lead to regretted choices initially
• (How does this affect Chernevs results?)

98
More thinking is NOT always better! Why?
Considering the reasons for a decision may lead
to including reasons that are not actually useful.
99
3. Can Citizens Use the Recognition Heuristic to
predict Election Outcomes?
Accuracy (a) citizens could obtain by relying on
the recognition heuristic
Candidates M 71 Mdn 73 SD 22 n
164
Parties M 84 Mdn 85 SD 11 n
163
Citizens can derive accurate predictions on the
election outcomes by relying on the recognition
heuristic.
100
4a. Do Citizens Rely on the Recognition Heuristic
to Predict Election Outcomes? (Candidates)
M 80 Mdn 86 SD 21
100
1.00
80
60
Proportion of Predictions Consistent with the
Recognition Heuristic
40
The recognition heuristic can account for the
majority of participants election forecasts.
20
00
0.00
5
9
1
164 Participants
4
45
49
21
25
33
37
53
57
61
65
69
73
77
81
85
89
93
97
13
17
101
105
109
113
117
121
125
133
7
101
4b. Do Citizens Rely on the Recognition Heuristic
to Predict Election Outcomes? (Parties)
M 77 Md 82 SD 22
100
80
60
Proportion of Predictions Consistent with the
Recognition Heuristic
The recognition heuristic can account for the
majority of participants election forecasts.
40
20
00
163 Participants
102
5. Are Citizens Better off If Their Predictions
do not Always Follow the Recognition
Heuristic?(Parties)
Accuracy (a) citizens could obtain by relying on
recognition
Accuracy citizens did in fact obtain
M 84 SD 11 n 163
M 75 SD 20 n 163
Mean difference 8, 95 CI on this difference
5 to 12 (p .001)
For predictions on parties citizens could have
derived more accurate election forecasts on
average if their predictions had always followed
the recognition heuristic.
103
Can the Recognition Heuristic be Used to Predict
Election Outcomes? Yes!
Wrap up

• The media may be responsible for the association
  between recognition and election outcomes.
• Citizens collective recognition of party and
  candidate names predicts the election outcomes
  well.
• Citizens can derive accurate predictions on the
  election outcomes by relying on the recognition
  heuristic.
• The recognition heuristic can account for the
  majority of participants election forecasts.

104
What shapes our cognitive mechanisms

• Herbert Simon rational behavior is shaped by the
  structure of task environment and computational
  capabilities of agent, fitting together like two
  blades of scissors
• Egon Brunswik mind and environment co-adapt to
  each other like husband and wife texture of
  environment captured in weights given to cues to
  form proximal impression of distal stimulus

105
What shapes our cognitive mechanisms

• Roger Shepard structures in mind mirror
  structures in environment, e.g. in color
  perception, object motion
• James Gibson mind is designed to directly pick
  up affordance structures in environment
• Amos Tversky/Daniel Kahneman heuristics use
  ecologically valid clues (but can lead to
  systematic biases/errors)

106
How do people select a heuristic?

• Can people tell what heuristic to apply in
  different situations? If so, how?
• Rieskamp (2003)
• have people make paired comparisons in a
  situation with feedback supporting either
• Weighted combination of cues
• Non-compensatory use of first discriminating cue
• What happens over many trials?

107
Can people learn when to use Take The Best?
100
NoncompensatoryFeedback
90
80
70
60
Choices predicted by Take The Best ()
50
40
CompensatoryFeedback
30
20
10
0
0-24
25-48
49-72
73-96
97-120
121-144
145-168
Feedback Trials
Rieskamp Otto (2001)
108
Other relevant research directions

• How can people make choices (preference, not
  inference) in the face of overwhelming
  information and number of options?
• How can people make sequential decisions when
  only one option is encountered at a time, but
  better ones may come later? (e.g., parking
  places)
• How do people learn about environment structure,
  e.g. what cues are more important?
• How do people divide up resources that they have
  worked together to produce?
• How do people decide on portion sizes to consume?
• How can we structure environments to help people
  to make better decisions? (e.g. frequency formats)