The Effect of Emotions on Economic Decision-Making

Report
The Effect of Emotions
on Economic Decision-Making
MAS 630: Affective Computing
Javier Hernandez Rivera
[email protected]
Contents
•
•
•
•
•
•
Motivation & Project Goals
Background
Experimental Setting
Data Synchronization & Visualization
Preliminary Data Analysis
Conclusions
2
Motivation
&
Project Goals
3
Affect in Decision Making
Emotions have been long neglected in decision making (DM)
in favor of a deliberative and reason-based decision making
(Shafir, Simonson, & Tversky, 1993)
Why? Affect can lead us to irrational decision making
(ignoring the odds or negative consequences)
Playing
the lottery
Smoking
Flying
by plane
Happy
Makes
People
Feel
Relaxed
Fearful
4
Project Goals
What?
• Validate current economic DM theories (e.g.,Somatic
Marker Hypothesis) in different settings
• Understand how negative emotions (fear and anger)
affect the DM process
How?
• Emotion elicitation
• Two-armed Bandit task
• Electrodermal activity (EDA)
Why?
• Understand the role of emotions in DM
• Explore the benefits and limitations of most common
emotional responses to catastrophes
5
Background
6
Roles of Emotions in Decision Making
1) Minimize negative emotions
vs
2) Emotions as common currency
3) Encode and recall information
Positive
Negative
4) Motivator of information
processing and behavior
(Peters E., Vastfjall D., Garling T. & Slovic P, 2006)
7
Factors that Influence Decision Making
Perceived
value
Time1
Uncertainty2,3
Risk3,4
time
Visceral States
Sexual
Ownership
Arousal5
Relaxed7
Hunger6
1
(Lowenstein, 1992)
4(MacGregor
7(Pham,
Disgusted8
Sad8
2(Bar-Anan.,
Wilson & Gilbert , 2009)
et al., 2005) 5(Ariely & Loewenstein, 2006)
Hung, Gorn, 2011)
8(Lerner,
3(Lerner,
6(Read
Small & Loewenstein, 2004)
& Tiedens, 2006)
& Leeuwen, 1998)
8
Decision Making and Physiology
Somatic Marker Hypothesis (SMH)
Theory: Physiological responses (a.k.a. somatic markers), learned in daily life
activity, consciously or unconsciously influence the decision-making process.
(Bechara A., Damasio H., & Tranel D. 1991, 1997)
Experiment: Iowa Gambling task
A
B
C
D
Disadvantageous decks
Lead to overall loss
Advantageous decks
Lead to overall gain
Risky option (high variance)
Safe option (low variance)
x 100 Trials
Observation: Higher EDA responses before choosing risky and disadvantageous
options, even before people could consciously identify the risky decks.
9
Anger and Fear
Most common emotional reactions after catastrophic events such as
the terrorist attacks of 9/11 or the economical crisis
Anger
Fear
Appraisal to
negative events1
Certainty
Control
Uncertainty
Uncontrolled
Influence on
Decision Making1
Risk-seeking
Optimistic assessments
Risk-averse
Pessimistic assessments
Low
High
Physiologycal
Responses2
1(Lerner
and Keltner, 2000,2001)
2(Lerner,
Dahl, Hariri & Taylor, 2006)
10
Experimental Setting
Designed & conducted by Hyungil Ahn
(Ahn, 2010)
11
Experimental Setting
Neutral
Fear
x 25 Trials
x 25 Trials
Bet Money
Option 1
Option 2
Bet Money
Option 1
Option 2
Gain
Loss
-
+
Safe option
(low variance)
is better
Anger
Emotions
Domain 1
Ownership
-
+
Risky option
(high variance)
is better
Domain 2
Risk + Uncertainty
Experimental Setting: 1 Trial
1
1
2 3
4
5
6
EDA
3
2
Time
4
5
6
13
Data Synchronization
&
Visualization
14
Data Synchronization
Surveys
Loss-pass
filter
(0.16 Hz cutoff
frequency)
15 participants were excluded
because of corrupted signals
(artifacts, low response)
20 participants were excluded
because of missing information
Number of Participants
Frames
EDA
(20 Hz)
Normalization
Scale each
subject
between
0 and 1
Best Option
Filtering
Neutral
Anger
Fear
Gain
3
3
5
Loss
4
5
5
Neutral
Anger
Fear
Safe
7
8
10
Risky
7
8
10
25 x
participants
2
= 1250
sessions
trials
Task Activity
15
Data Visualization (Neutral)
Video
(B B)
.
(N1 N1)
(N2
Risky Option is Better Video
N2)
.
. .
.
1
1 1
2
(N1
(N N1) 2 1 2
2
2 . 1 2 12 1 1
1
.
.
.
.
.
.
.
B)
.
.
2 1
1 2 11 1
.
.
.
. .
.
.
. .
.
.
(N1 N1)
121 2 1
1
2
.
1
2
11
.
.
.
.
.
.
.
.
.
. .
.
. . . .
. .
. .
.
2
.
.
2 2 111
(N3 N3)
2
211 1
1 1
1 11 1
. . .
1111
. .
.
.
.
.
. . .
11 1 1 1 1 1 1 11 1 1 1
1 112 11
. .
. . .
. . .
. . . . . . . . . . . . . . . .
S)
Gain
Frame
(3 participants)
11 11 1
1
. .
..
. .. .
.
..
2
.
.
.
1
.
.
N2)1
(N2
Selected Options
S) (‘1’ is always
the optimal selection)
(N3 N3)
1 2 2
11 1
1 2 1 1 2 1 1 21 11 1 1 1 1 1 1 1 1
. .
. .
Safe Option is Better
2 22 2
. .
.
. . . . .
2 1
.
.
.
2 222
222 2
. ..
.
. .. .
(N2
N2)
(N3 N3)
S)
2 2
2
.
21 2
1
21 1 1
1 211 11 11 111 1
2 1 111
.
..
..
.
. .
.
Data in
.
.
. . . . . .. .. . .
. . . . .
1
1 111
(N1
(N N1)
(N2
N2)
(N3 N3)
.
1
1
1
11111
2
1111 1
1
1
1
1 2 1 1 . . 1 1 1 1 1 1 111
.
1
.
.
211
. . .
.
.. . . .
111 1112111
..
.
. .
1
. . .
.
.
.
. ...
.
.
.
S)
..
.
..
. . .. .. .
. . .
. .
Neutral
2
(B B)
2
(N1
(N N1)
(N2
1
2 1 21 (N3 N3)
.1
N2)
.
. .
2
2
2
.
.
2
.
2
2
2
12 22 2 1 11
1
2
1 1 . .1 2 .1 . . .
2 .
..
.
.
.
. . .
1. .
.
.
.
.
..
.
(N1
(N N1)
.
.
1
.
(N1
(N N1)
1 122 21 1
.
. ..
(N2
2
2
2
.
.
.
(N2
S)
.. .
.
N2)
(N3 N3)
2 2 1 1 11 1 11 1 2 2 2 2 1 1
2
21
1 21
2
1 21 2 12111
111
. . . . .
. . .. . .
. . .
.
.
N2)
1 2 2 2 22222121221122
12221
. .. . .
. . . . . . . . . . . . . . . . . .. .
. . . ..
7 participants
(350 trials)
.
..
1
2
1
.2
12 1
. 12 21 12
12
1
. .2
1
.
..
.
.
2 1
.
. . . .. . .
.
(N3 N3)
111211121111211 1 1 2 1 1
. . . . . . . . . .. . . . . . . . . .
1 1 11 1 1 2 22
. . . . .. . . . .
. . . .. . . . . .
S)
Loss
Frame
(4 participants)
S)
16
Data Visualization (Anger)
Video
(A1
Risky Option
is Better
Safe Option
Video
is Better
A1)
(A2
2
12 221
.. ..
.
A2)
(A3
. . ....
(A1
. . . ..
.
(A1
12212
.
.
(A1
.
.
.
.
.
2 1 211
1
2 1212
.
.
A3)
S)
Data in
2122112221221222222
...
. .. .
. .. .. . . . . ... .. . .
A2)
(A3
2
1 1
1 2
1 2 1 1 2 2 1 2 2 1 2 2 1 1 1 2212 1
.
.
.
.
.
.
.
.
.
. .
. .
. . .
(A2
. .
A3)
1 2 1 2 112 111211 2 1211 12112 11
. . . . . . .
. . .
. . . . . . . . . . . . . . . . . ... ..
A2)
(A3
A3)
Anger
.
12
(B B)
.
. .
2 1
.
.
.
(A3
A1)
.
.
S)
. ..
.. . . . . . .
.
A2)
11121 1 1
. . . .
111111
. . . ..
. .. . .
. ..
.. .. .
1 2
.
. . .. . . . . . . .
.
(A2
(A1
A3)
1
2 111 11211121 12111121112
1
.
. .. . . . .
A1)
(B B)
(A3
(A2
1.
. .
(B B)
A2)
A1)
12 2
2 12
Gain
Frame
(3 participants)
.
221 2 112112 1
. . .
.. .
. .. . .
.
B B)
. . . . . . . .. .
.
.
(A2
.
.
.. .
A1)
11 2
1 112 2
2
.
S)
221221 12 122 122 21 2
122 2 21 2 1
.. .
221212121221 211212 1
. . . . . . .. .
. . . . . . .. . . .
(B B)
A3)
.
.
(A1
1 12
1
1 1
.
2
.
.
2
.
A1)
.
1
.
(A1
A1)
(A2
(A1
. . . .
. . .
.
A1)
(A2
1
.
. . .
. .. . . . . . . . . ..
.
11
. .
11
11
..
12 2
1
.
. .
. .
1
. . .
1 2 2
11 2
1 2 1 12 11 1 21 1 1 22 1 1 1 1 1 2 2
1 12
1
.
.
.
.
.
. .
1
11111 21 1121 1111 1
11 . .
. . . .. ... .. .
1 21 11
. . . .
(A2
.
(A3
111 1111
.
.
. .
..
. .. . . . ... . .
. . . . .
111111
1
21 . . . . . .
A2)
2
2
.
222222 . . .
11
2 2
111
1 .. . .. . .
.
. .
. . . . .
1 11111
11111111
. . .. . .. . .
. . .. . . ..
2 1 2 2 2 21 121 11 1 11 1
1
1
A2)
2 1
1 2 1 1 2 2 11 211 12 1
121121211
.
. .
.
..
..
A3)
(A3
. .
A3)
8 participants
(400 trials)
..
Loss
Frame
(5 participants)
S)
11
11 .
.
. . . . ...
.. .. .. . . . . . . .
. .
A2)
2
.
11
111
. .
. ..
1
1
.
.
1
1
11
1 21 1 11 2
11 . . .
..
.
.
.
.. .
1 1 11
. .
. . .
.
(A3
A3)
S)
17
Data Visualization (Fear)
Video
(F1
F1)
Safe Option is Better Video
(F2
F2)
2 2 (F3
1
11
1
1 1 11.
. .
. . .
. .
(F1
F1)
21
1 .
2 1 11 . .
.
(F1
12 21 2
(F2
1 1
. . 2 1
.
. . . 11 . 1 1
.
.
. . . 11 111
.
F1)
(F2
.
. .. .. . .
. ..
F3)
.
222
2
.
F3)
(F3
F3)
S)
. . . . . . . . . . . . . .. . . . . . . . . . .
F1)
1
11122122
21222 . . . . . . . .
(F2
F2)
21
1 1
.
. . . . .. ..
.
F1)
(F2
1111111111
1 11
111
.. .. . .. .
1 2 2 2 2 21 2 2
. . .
....
.
.
. .
F1)1 2 22 1 1 2 2 21 12221
2
212 1 12221
. . .
.
. . . .
. . . . . . .
. . . . . .
. . . .
(F1
.
.
. . .
(F2
(F1
F1)
.
.....
. ... ..
B)
(F1
....
..
2 22
2 2 2 222
. . ..
. .. .
.. . . .. . ...
F3)
S)
Fear
(F3
F3)
. . . . .
10 participants
(500 trials)
S)
. . . . . . . . . . .. . . . . . ..
(F3
F3)
S)
Loss
Frame
(5 participants)
1
21222 21221212222 22221122
. . . .. . . . . . .
.. ..
. . . . . . ... .
. .
. . ..
. . . . . .. . . . .
(F2
2
1212
1 21
.
2112212211 . . . 1212212
..
. .
(F3
222 2
S)
21
F2)
21
1 2 1 12 2
1 1 11 1111112111111
(F1
1 1 2 2 22
2
. . .
. . .
F3)
1 22 112 221211 1 111111 11 1 1 11
. . . . .. .
(F2
. .
. . .
2 211
F3)
..
.
(F3
F2)
12 21 111 11 12 1121 11 1
F1)
. . .
.
12
(F3
1 12 1 1
1
2111 1121 2. . . . .
1
.
.
2
.
..
.
...
1
.
.
1
.. .
Gain
Frame
(5 participants)
111111 1
. . .. . . .. . .
1 1111
. . .. .
. . . . .. . . .
.
(F2
. . . . . .
. . . . . .. . . . . .
..
Data in
F2)
2
.
F2)
22111111111
F1)
12 11 1 1 1
S)
1
12
..
2 21112122 . . . . . .
(F1
S)
. . .
1111111111 1111111111
..
..
. . .. .. .. .
.. . . .. .. . . . .
(F1
(F1
.
..
.
111 1122211 1111111111 111 11
. . . . . . . . . ... . . . .. . . . .. . . .
B)
1
.1
1
1
1 11 2 11 . 11121 1 2 1
. . 12 1
11 1 2. . . .
....
. .
.
.
..
(F3
F2)
2 1 2112122211111 2111221111
1
1 1 111 11111 2 1 2 .
.
1
.
.
F2)
.. ..
.
. . .
Risky Option is Better
F2)
(F3
F3)
S)
111212121
2112121212121121
. .. . .
F1)
1222111111111111111 111111
. . . . . . . . . . . . .. . . . . . . . . . . .
... . . .. . ...
.. ... .. ..
(F2
F2)
(F3
F3)
S)
18
11121111111111111111 11112
.. . . . . . . . . . . . . . . . .. . . . . . .
Preliminary
Data Analysis
19
Behavioral Responses: Speed
Standard Error of the Mean (SEM)
25
Trial
2 3
4
EDA
Time
5
6
20
Average Trial
Response Time (sec)
1
15
10
5
0
0.5
Betting
Surveys
1
Neutral
(N = 350)
1.5
*
2
Anger
(N = 399)
2.5
*
3
3.5
Fear
(N = 500)
People answer significantly faster in the negative
emotional states, and fearful people are significantly
faster than angry people.
* Statistically Significant (Two Sample T-Test)
20
Behavioral Responses: Performance
Advantageous
Disadvantageous
Emotions - Median Anticipatory Signal
0.8
Overall, people in the three
emotional conditions perform
similarly.
0.7
0.6
Safe
Option
Is Better
% Risky
0.5
0.4
% of
Selections
0.3
0.2
Risky
Option is
Better
0.8
0.5
0.1
0
0.8
Neutral
Emotions - Median Anticipatory Signal
Anger
*
Fear
*
0.7
% Risky
% Risky
0.6
Domains - Median Anticipatory Signal
0.7
0.4
Negative states are slightly better
when the safe option is the
optimal one, but they are slightly
worse when the risky option is the
optimal one.
Fearful people tend to perform
slightly better than angry people
0.6
0.3
0.5
0.2
0.4
0.1
0.3
0
0.2
0.1
*
Neutral
Neutral
*
Anger
Anger
Advantageous
21
Fear* * Statistically Significant (Two Sample T-Test)
Fear
Behavioral Responses: Risk Preference
Non-Risky Option
Risky Option
Emotions - Median Anticipatory Signal
0.8
Although people in the
neutral state significantly
choose riskier options,
people in the negative states
prefer non-riskier options.
0.7
0.6
0.5
% Risky
Gain
Frame
0.4
0.3
% of
Selections
0.2
0.1
0
Neutral
0.8
% Risky
Emotions - Median
Anticipatory Signal
Anger
Fear
Domains - Median Anticipatory Signal
In the loss frame, people
prefer the riskier options. The
difference is significant for
the neutral and fear settings.
0.5
0.7
0.4
0.6
0.3
0.5
% Risky
*
0.7
0.6
0.8
Loss
Frame
*
0.2
0.4
0.1
0.3
0
0.2
0.1
0
*
Neutral
Neutral
Anger
Anger
Non-Risky
Risky
*
Fear * Statistically Significant (Two Sample T-Test) 22
Fear
Behavioral Responses: Pleasantness
100
90
90
80
80
Pessimistic vs Optimistic
100
70
60
50
40
30
Angry people in the loss
frame perform slightly
better than angry
people in the gain
frame.
70
60
50
40
30
20
20
10
10
0
100
Average of Pleasantness
the Outcomes
Ratings on
Pessimistic vs Optimistic
Loss Frame
0
N
A
F
N
90
80
80
70
60
50
40
30
60
50
40
30
20
10
10
0
Neutral
N
Anger
A
Fear
F
F
70
20
0
A
100
90
Pessimistic vs Optimistic
Average of the % of
Selections
Advantageous
Pessimistic vs Optimistic
Gain Frame
N
A
Neutral Anger
F
Fear
As expected, the overall
pleasantness ratings on
the outcomes are slightly
lower in the loss frame.
Moreover, angry people
are surprisingly
unpleased even though
they obtained slightly
higher outcomes.
Preprocessing for EDA Analysis
Baseline Removal
Smoothed
Minimum Sliding
Window over 10
minutes
Filtering
Loss-pass
filter
EDA
(0.16 Hz cutoff
frequency)
Normalization
Scale each
subject
between
0 and 1*
Feature
Extraction
Normalized
Area under the
Curve
0.2
0.1
0.2
µS
0
0.1
0
0.3
5
10
15
20
25
30
35
40
5
10
15
20
25
30
35
40
5
10
15
20
25
30
35
40
Original Signal
Low-pass filtered signal
Baseline
Corrected signal
0.2
0.3
0.1
0.2
0
0.1
Minutes
*(Lykken, D.T., Venables, P.H, 1971)
24
Anticipatory Responses: SMH
Iowa Gambling Task
Two-Armed Bandit Task
-4
2
ACTIV ACTIV
ACTIV ACTIV
2
1
0
0
1-8
9-17
Trials
N
9-17
Trials
0
18-25
9
1-8
9-17
Trials
1-8
9-17
Trials
18-25
N
18-25
7
6
N
95
84
3
7
2
6
1
4
Risky Option is Better
8
Total # of Options
Total # of Options
Total # of Options
Hunch
Conceptual
Period
N (n: 625)
1
18-25
7
PrePrePunishment Hunch
N (n: 625)
-4
2
0
1-8
8
50
-4
1
1
1-8
1-8
9-17
Trials
9-17
18-25
18-25
3
Total # of Options
Average Activation
Safe Option Is Better
2
9
Total # Selections
x 10
N (n: 625)
x 10
Safe Option Is Better
x 10
N (n: 625)
x 10-4
6
N
9
5
4
8
3
7
2
6
1
5
0
*Trials
4
1-8
1-8
*
9-17
9-17
Trials
*
18-25
18-25
*
3
The SMH hypothesis (higher EDA responses before disadvantageous
Advantageous
selections) seems plausible when the Safe Option is optimal and it
Disadvantageous
* Statistically Significant
might be delayed when the Risky Option is the optimal one.
2
2
1
1
0
1-8
9-17
Trials
18-25
0
1-8
9-17
Trials
18-25
Main Limitations of the Analysis
1) Reduced number of participants (35 part. were excluded)
2) Consecutive tasks distort EDA responses
InitTrial SelectOption
(0.00 sec) (1.65 sec)
1 2 3
AnswerExperience
(6.28 sec)
4
5
Too short to
display
anticipatory
responses?
AnswerPrediction
(5.66 sec)
6
Cognitive load of
the first survey?
Average EDA
response
(N: 1250 trials)
BetClick
GetOutcome
(1.86 sec)
(0.64 sec)
Betting
~4 sec.
AnswerConfidence
(4.15 sec)
Answering
Surveys
~16 sec.
26
Conclusions
• People in the negative states bet faster than
people in the neutral state.
• Fearful people bet faster and performed
slightly better than angry people.
• Although most of the people preferred
riskier options, angry and fearful people in
the gain frame preferred safer options.
• Angry people performed slightly better in
the loss frame.
• Angry people were less pleased in the loss
frame even though they obtained relatively
higher outcomes.
• Although the SMH seemed plausible in the
Two-armed Bandit Task, further analysis is
required.
Time Distribution
Deliverables
Data
Analysis
Readings
Data
Synchronization
27
References I
Ahn, H.I. (2010). Modeling and Analysis of Affective Influences on Human Experience, Prediction,
Decision Making, and Behavior. MIT PhD Thesis.
Ariely D., & Loewenstein G. (2006). The Heat of the Moment: The Effect of Sexual Arousal on Sexual
Decision Making. J. Behav. Dec. Making, (19), 87-98
Bar-Anan Y., Wilson T & Gilbert (2009) . The Feeling of Uncertainty Intensities Affective Reactions.
Emotion 9, (1), 123-127
Bechara A., Damasio H., & Tranel D. (1997). Deciding Advantageously Before Knowing the
Advantageous Strategy. Science.
Damasio, A. R., Tranel, D., & Damasio, H. (1991). Somatic Markers and the Guidance of Behavior:
Theory and Preliminary Testing.
Lerner, J. S., Dahl, R. E., Hariri, A. R., & Taylor, S. E. (2007). Facial Expressions of Emotion Reveal
Neuroendocrine and Cardiovascular Stress Responses. Biol Psychiatry; 61:,253-260
Lerner, J. S., & Keltner, D. (2000). Beyond Valence: Toward a Model of Emotion-specific Influences
on Judgment and Choice. Cognition and Emotion, 14(4), 473–493.
Lerner, J. S., & Keltner, D. (2001). Fear, Anger, and Risk. Journal of Personality and Social Psychology,
81(1), 146–159.
Lerner, J. S., Small, D. A., & Loewenstein, G. (2004). Heart Strings and Purse Strings: Effects of
Emotions on Economic Transactions. Psychological Science, 15, 337–341.
28
References II
Loewenstein, G. & Prelec, D. (1992). Anomalies in Intertemporal Choice: Evidence and an
Interpretation. Quarterly Journal of Economics. 573-597
Lykken, D.T. & Venables, P.H.(1971) Direct Measurement of Skin Conductance: A Proposal for
Standarization. Psychophysiology 8(5), 656–672
MacGregor, Slovic P, Peters P & Finucane M. (2005) Affect, Risk, and Decision Making. Health
Psycholoy, 24 (4) S35-S40
Peters E., Vastfjall D., Garling T. & Slovic P. (2006). Affect and Decision Making: A “Hot” Topic.
Journal of Behavioral Decision Making, 19, 79-85
Pham M., Hung I. & Gorn G. (2011). Relaxation Increases Monetary Valuations. Journal of
Marketing Research, 48
Read & Lweeuwen (1999). Predicting Hunger: The Effects of Appetite and Delay on Choice.
Organizational Behavior and Human Decision Processes, 76(2), 189-205
Shafir, E., Simonson, I., & Tversky, A. (1993). Reason-based Choice. Cognition, 49, 11-36.
29

similar documents