what are cognitive biases?

International Society for Intelligence Research
Annual Conference (Graz, 2014)
Individual Differences in
Cognitive Biases
Evidence Against One-Factor Theory of Rationality
Predrag Teovanović, Institute of Psychology, Faculty of Philosophy, University of Belgrade
Goran Knežević, Department for Psychology, Faculty of Philosophy, University of Belgrade
Lazar Stankov, Institute for Positive Psychology and Education, Australian Catholic University
Cognitive Biases
 WHAT ARE COGNITIVE BIASES?
Systematic deviations of cognitive processes’ outcomes from what is normatively
considered to be rational behavior.
 HOW MANY COGNITIVE BIASES THERE ARE?
14
Tversky and Kahneman (1974)
53
Baron (2008)
100+
Wikipedia (2014)
 HOW TO CLASSIFY COGNITIVE BIASES?
Tversky and Kahneman (1974): Representativeness, Availability and Anchoring
Haselton, Nettle and Andrews (2005): Error Managment, Heuristics, Artificial
Stanovich (2009): Focal Bias, Override Failure, Mindware Gap, Egocentric Processing
Arnott (2006): Adjustment, Confidence, Memory, Statistical, Situation (...)
Kahneman & Frederick (2005): Coherence Rationality and Reasoning Rationality
NORMATIVE
DESCRIPTIVE
Individual Differences
 HOW DO COGNITIVE BIASES RELATE TO EACH OTHER?
Stanovich and West (1998, 2000) – Positive manifold
De Bruin, Parker and Fischhoff (2005, 2007) – Decision Making Competence
 HOW DO COGNITIVE BIASES RELATE TO INTELLIGENCE?
Stanovich & West (2008) – Some biases correlate with measures of intelligence
De Bruin, Parker and Fischhoff (2007) – DMC higly correlate with intelligence
NORMATIVE
DESCRIPTIVE
Research Questions
1. Can individual differences in cognitive biases be reliably measured?
2. How do cognitive biases relate to each other?
Is there single latent factor of rational behavior?
3. Can cognitive biases be reduced to gf, gc, NFC, CRT and Openness?
NORMATIVE
DESCRIPTIVE
Cognitive Biases
Anchoring Effect
Belief Bias
Normative
Model
Coherence
De Bruin et
Pohl
al. (2007)
(2004)
Consistency Judgment
Carter et al.
(2007)
Ref. Point
Availability
Confirmatory
Logic
Accuracy
Overconfidence Bias
Calibration
Accuracy
Hindsight Bias
Coherence
Consistency
Memory
Imperf. Corr.
Output Eval.
Probability T.
Accuracy
Thinking
Att.Focus
Base Rate
Utility T.
Accuracy
Base Rate Neglect
Sunk Cost Effect
Outcome Bias
Coherence
OTHER CLASSIFICATIONS
Thinking
Baron
(2008)
Availability
Judgment Motivational Control Illusion
Judgment Imperf. Corr.
Commitment
Consistency Judgment Imperf. Corr.
Confirmatory
Tversky and Kahneman (1974): Representativeness, Availability and Anchoring
Haselton, Nettle and Andrews (2005): Error Managment, Heuristics, Artificial
Stanovich (2009): Focal Bias, Override Failure, Mindware Gap, Egocentric Processing
Arnott (2006): Adjustment, Confidence, Memory, Statistical, Situation (...)
Kahneman & Frederick (2005): Coherence Rationality and Reasoning Rationality
Instruments
Cognitive Biases
Other Measures
Anchoring Effect
Raven’s Progressive Matrices (Raven, Court & Raven, 1979)
Belief Bias
Swaps Test (Stankov, 2000)
Overconfidence Bias
Three-Dimensinal Space Test (Wolf et al., 1992)
Hindsight Bias
Vocabulary Test (Knežević & Opačić, 2011)
Base Rate Neglect
Test of Analogies (Wolf et al., 1992)
Sunk Cost Effect
Test of Synonyms-Antonyms (Wolf et al., 1992)
Outcome Bias
Cognitive Reflection Test (CRT; Frederick, 2005)
gf
gc
Openness (McCrae & Costa, 2004)
Participants
Need For Cognition Scale (NFC; Cacciopo et al., 1984)
All measures were computer administrered in two sessions, one week apart.
Participants were 243 undergraduate students (22 males; mean age 19.83, SD=1.31).
They were tested in grops of approximately 20.
Overall testing time was 60 minutes per session.
Cognitive Biases
Range
M
SD
Cohen’s d
Anchoring Effect
0–1
0.44
0.18
2.05
Cronbach’s
Alpha
.77
Belief Bias
0–8
4.05
2.34
2.05
.76
0 – 100
32.56
18.99
2.31
.94
Hindsight Bias
0–1
0.33
0.20
0.24
.66
Base Rate Neglect
0–1
0.78
0.22
3.58
.71
Sunk Cost Effect
0–5
2.16
1.06
2.04
.76
Outcome Bias
0–5
1.55
0.91
2.43
.83
Overconfidence Bias
Note: Higher scores indicate more pronuounced biass
Better than α=. 11 (Pohl, 1999)
Intercorrelations of Cognitive Biases
Cognitive Bias
1. Anchoring Effect
2. Belief Bias
3. Overconfidence Bias
4. Hindsight Bias
5. Base Rate Neglect
6. Sunk Cost Effect
7. Outcome Bias
1
2
3
4
5
6
7
-.09
-
-.09
.03
-
.18
.04
-.18
-
.12
.09
-.01
.12
-
.16
.18
.10
.15
.17
-
-.01
.20
.04
.07
.06
.06
-
Factor Pattern Matrix
for Seven Cognitive Bias Measures
Cognitive Bias
Anchoring Effect
Belief Bias
Overconfidence Bias
Hindsight Bias
Base Rate Neglect
Sunk Cost Effect
Outcome Bias
PC Eigenvalue
Variance Explained
Factor 1
.34
Factor 2
.31
-.42
.50
.54
1.539
9.23%
Number of Factors – PC Eignevalues
Extraction Method - Maximum Likelihood
Rotation Method - Direct Oblimin
1.255
8.33%
Cognitive Biases and Other Measures
Raven’s Matrices
Swap Test
3D Space Test
Vocabulary
Synonyms-Antonyms
Analogies
Cognitive Reflection
Need For Cognition
Openness/Intellect
ANC
.03
-.08
.02
-.02
-.11
-.03
-.02
.10
.21
BLF
-.15
-.22
-.19
-.10
-.12
-.19
-.17
.01
.01
OVC
-.21
-.15
-.14
-.08
.00
.02
-.07
-.10
.05
HSB
.05
.06
-.01
-.05
-.08
-.10
-.06
-.05
-.02
BRN
-.24
-.20
-.21
-.16
-.18
-.13
-.22
-.10
-.10
R2=.10
SCE
-.12
-.09
-.14
-.06
-.09
-.04
-.22
-.07
-.11
OUT
-.17
-.06
-.14
.00
-.01
-.21
-.12
-.04
-.01
Intercorrelations of Other Measures
8. Raven’s Matrices
9. Swap Test
10. 3D Space Test
11. Vocabulary
12. Synonyms-Antonyms
13. Analogies
14. Cognitive Reflection
15. Need For Cognition
16. Openness/Intellect
8
.48
.52
.26
.19
.26
.25
.19
.22
9
10
11
12
13
14
15
16
.40
.11
.21
.28
.24
.12
.13
.21
.11
.16
.35
.18
.21
.39
.42
.10
.11
.19
.57
.07
.11
.09
.14
.14
.15
.16
.17
.58
-
Factor Pattern Matrix
for All Sixteen Measures
Cognitive Bias
Anchoring Effect
Belief Bias
Overconfidence Bias
Hindsight Bias
Base Rate Neglect
Sunk Cost Effect
Outcome Bias
Raven’s Matrices
Swap Test
3D Space Test
Vocabulary
Synonyms-Antonyms
Analogies
Cognitive Reflection
Need For Cognition
Openness/Intellect
Gc
TD
Gf
-.38
.64
.49
.54
.49
.72
.85
.62
.95
Bias1
.31
.54
Bias2
.44
.37
.36
.40
1. Experimental reliability of cognitive biases was confirmed
2. Most of cognitive biases showed satisfactory reliability estimates (αs>.70)
3. Intercorrelations among cognitive biases are only of small magnitude (rs<.20)
Positive manifold was not observed
Measurement error did not play role
Two cognitive bias factors could be extracted
These factors were not robust
One-factor theory of rational behavior is not plausible
4. Biases are relatively independent form gc, gf and thinking dispositions
Individual cognitive biases correlated lowly with these measures (rs<.20)
Cognitive bias measures loaded on separate factors
Bias factors were relatively independent form the gc, gf and TD factors
Thank you.