Integrating Motivation and Emotion into Decision Making

1
Integrating Motivation and Emotion into Decision
Making
Jerome R. Busemeyer Ryan K. Jessup Indiana
University jbusemey_at_indiana.edu http//mypage.iu.e
du/jbusemey/

• Modeling Integrated Cognitive Systems Systems,
• Saratoga Springs NY

2
What systems are we trying to integrate?

• Problem Solving
• Generate plans to accomplish goals
• Plans are action- event sequences courses of
  action
• Judgment
• Estimate likelihood of events that occur in a
  plan
• Evaluate importance of consequences produced
  along the paths of a plan
• Decision Making
• Select a course of action that has uncertain but
  important consequences
• E.g. Decide whether or not to pass a truck on a
  dangerous two lane highway
• Motivation
• Persistent needs that arouse and energize long
  term goals
• Hunger, Sex, Curiosity, Security, Power, ect.
• Emotions
• Temporary states reflecting current changes in
  motivation
• Joy (e.g. gain of power) vs. Anger
  (e.g. loss of power)
• Hope (anticipated gain) vs. Fear
  (anticipated loss)
• Affect
• State evaluation in terms of positive versus
  negative feelings
• Anger ? negative feeling, Joy ? positive feeling

3
What are the Bases of Motivational and Emotional
Experiences (E.g., Fear)?

• Neuro activation
• Brain Activation (Fearincrease,
  Sadnessdecrease)
• Neurotransmitter release (GABAinhibition,
  Dopamine reward)
• Hormonal response of the Endocrine system
• Adrenaline (epinephrine) tension anxiety flight
• Noradrenaline (norepinephrine) aggression fight
• Physiological reaction of autonomic nervous
  system
• Pupil size, heart rate, respiratory rate
• Galvanic skin conductance (perspiration), skin
  temperature
• Behavioral Preparation
• Facial expressions (Tomkins, Izard, Ekman) body
  posture
• Programmed reactions and coping responses (flight
  or flight)
• Cognitive Interpretation (more on next slide)
• Appraisal and interpretation of above reactions
• James-Lange theory (Schacter Singer, 1962
  Lazarus, 1991 Weiner, 1986)

4
(No Transcript)
5
Two System View of Motivation and Emotion
(Buck,1984 Gray, 1994 Ledoux, 1996
Levenson,1994 Sherer, 1994 Panksepp,1994
Zajonc, 1980)

• Subcortical Direct Route
• Fast, spontaneous, unconscious, physiological,
  involuntary reaction
• Thalamus ? Amygdala ? motor cortex, limbic
  circuit)
• Neocortical Indirect Route
• Slower, conscious, appraisal, coping response
  (indirect path through Thalamus ? Sensory Cortex
  ? Prefrontal Cortex ? Amygdala ? motor cortex,
  neocortical circuit)
• Integration of emotion and cognition
• Orbital (ventral medial) prefrontal cortex center
  for integration emotion and cognition (Damasio,
  1994)

6
Two System View of Decision Making (Epstein,
1994 Kahneman Frederick, 2002 Loewenstein
ODonoghue, 2004 Metcalf Mischell, 1999
Slovic Peters, 2000 Sloman, 1996,)

• Heart
• Emotional, Intuitive, Affective, based system
• Implicit, unconscious, automatic, associative,
  fast, parallel, non-compensatory, experiential,
  contextual
• Little demands on working memory
• Mind
• Rational, Analytic, Reasoning based system
• Explicit, conscious, controlled and deliberative,
  slow, serial, compensatory, comprehensive,
  abstract
• Large demands on working memory
• Heart is corrected by Mind at a cost of working
  memory (willpower).

7
Do we need to change decision theory for
emotional consequences?

• Regret effects (Zeelenberg Beattie, 1997,
  OBHDP)
• Preferences among gambles change depending on
  whether or not outcome feedback is given
  following choice (which provides an opportunity
  for regret).
• Decision weights (Rottenstreich Hsee. 2001,
  Psych Sci)
• Function is more inverse S-shaped (flat across
  the intermediate ranges of probabilities) for
  emotional outcomes.
• Discount Rates (Loewenstein Lerner, 2003)
• Higher discount rates are obtained using
  emotional consequences (e.g., cocaine vs. money
  for cocaine abusers)
• Decision Strategies (Luce, Bettman, Payne, 1997
  JEPLMC)
• Switch to non-compensatory strategies to avoid
  making difficult negative emotional tradeoffs.

8
Can emotions distort our decision processes?

• Emotional carry over effects (Goldberg, Lerner,
  Tetlock, 1999, European JSP Lerner, Small
  Loewenstein, 2004, Psych Sci)
• Anger from watching a murder movie spills over
  and influences judgments of punishments for
  unrelated crimes.
• Emotional films affects subsequent prices for
  gambles
• Emotions overwhelm reasons (Shiv and Fedorikhim
  (1999, JCR)
• When given a choice between a healthy and
  unhealthy snack, participants generally choose
  the health snack
• But with hunger aroused (tested before lunch) and
  healthy thoughts suppressed (by a working memory
  task), then the Unhealthy snack was preferred.

9
Does reasoning always improve decision making?

• Over-emphasis on Reasons (Wilson Schooler
  (1993, PSPB)
• Participants were asked to choose a poster to
  take home
• One group gave a list reasons prior to the choice
• Second simply used their intuitive feelings
• Six weeks later the group who focused on reasons
  were less pleased with their choice compared to
  the intuitive group.

10
Can we predict the effect of motivation on our
decisions?

• Hot-Cold Empathy Gaps (Loewenstein Lerner,
  2003, Read Van Leeuwen, 1998, OBHDP)
• When in a cold state, (not hungry), people under
  predict how they will feel in a hot state
  (hungry)
• When in a hot state (sexually aroused) people
  under predict how they will later feel when in a
  cold state (morning after effect)
• A person in a cold state (no pain) cannot predict
  how a person in a hot state (in pain) will react

11
Does mood bias information processing?

• Negative moods (as compared to positive moods)
  narrow the focus of attention and make people
  more vigilant and systematic in information
  processing (Isen, 1999 Schwarz, 1990)
• Pleasant moods enhance helping behavior (Baron,
  1997)
• Positive mood affects risk aversion. (Isen,
  Nygren, Ashby, 1998)
• Fearful moods generate pessimistic risk assements
  while anger produces less pessimistic risk
  assessments (Lerner Keltner, 2000)

12
Models of the Two System View

• Mind
• The decision maker retrieves weights and values
  from some fixed table (like reading a consumer
  report magazine).
• Utility is computed as the weights times values
  summed across outcomes
• Choose the action producing maximum utility
• Heart
• Collection of heuristic rules of thumb
• E.g. Lexicographic rule

13
Decision Field Theory A dynamic and stochastic
computational model of decision making

• Overview and Summary
• Busemeyer, J. R. Johnson, J. (2004)
  Computational models of decision making. D.
  Koehler N. Harvey (Eds.) Handbook of Judgment
  and Decision Making, Oxford UK Blackwell
  Publishing Co. Ch. 7, Pp 133-154.
• Decision Making Under Uncertainty
• Busemeyer, J., Townsend, J. T. (1993).
  Decision Field Theory A dynamic cognitive
  approach to decision making. Psychological
  Review, 100, 432-459.
• Multi Alternative Preferential Choice
• Roe, R. M., Busemeyer, J. R. Townsend, J. T.
  (2001) Multi-alternative decision field theory A
  dynamic artificial neural network model of
  decision-making. Psychological Review, 108,
  370-392.
• Price and Choice Preference Reversals
• Johnson, J. J. Busemeyer, J. R. (2004) A
  dynamic, stochastic, computational model of
  preference reversal phenomena. Revision under
  review for Psychological Review.
• Motivational basis of utility
• Busemeyer, J. R., Townsend, J. T., Stout, J. C.
  (2003) Motivational Underpinnings of Utility in
  Decision Making Decision Field Theory Analysis
  of Deprivation and Satiation. In S. Moore (Ed.)
  Emotional Cognition. Amsterdam John Benjamin

14
Example Dynamic Decision

• Walter is riding his motorcycle behind a truck on
  a dangerous two lane highway. The truck is loaded
  with old tires. Suddenly, the truck hits a bump
  and a tire bounces down, landing flat on the
  road directly in Walters path.
• What course of action should Walter choose?
• Screech to a stop to avoid the tire
• Swerve to the side and avoid the tire
• Speed up and ride straight over the top of the
  tire

15
Choice Process for Subject Controlled Stopping
Time
Random Walk / Diffusion Process
Threshold bound controls speed accuracy tradeoffs
16
Evolution of Preference
Connectionist Framework
M Motivational values W attention Weights V
input Valences V(t) C? M(t) ? W(t) P
Preference state P(th) S?P(t) V(th)
E V M?W an Expected Utility P(t) estimates
this over time
17
Evolution of Preference
time t, W(t) V input Valences V(t) C? M ?
W(t)
18
Evolution of Preference
time th, W(th) V input Valences V(th) C?
M ? W(th)
19
Evolution of Preference
time t2h, W(t2h) V input Valences V(t2h)C?
M ? W(t2h)
20
Evolution of Preference
P Preference state P(th) S?P(t) V(th)
M1
C
VA
PA
M2
S
VB
M3
PB
M4
VC
PC
M5
21
Multi-Alternative choice paradigm

• Binary choices
• Add a New Brand
• Compare Conditions

quality
New Brand
Quality
Choice Probability
Economy
economics
BMW Saturn
22
Similarity Effect (Tversky, 1972, Psychological
Review)
PrX X,Y ? PrYX,Y PrXX,Y,S YX,Y,S Preference Reversal Violation of
Independence from Irrelevant Alternatives Rules
out Simple Scalable Class of Models (e.g.
Luces,1959) ratio of strength model) Explained
by Tverskys (1972) Elimination by Aspects model
Y
s
X
23
Compromise Effect (Simonson, 1989, Journal of
Consumer Research)
Pr C Y,C Pr
Y X,Y,C Preference Reversal Violation of
Independence of Irrelevant Alternatives Cannot
be explained by Tverskys (1972) Elimination by
Aspects Model Explained by Tversky Simonsons
(1992) Loss Aversion Model
Y
C
X
24
Reference Point Effects (Tversky Kahneman,
1991, Quarterly Journal of Economics)
Pr X X,Y,Ry X,Y,Rx Pr Y X,Y,Rx Violation of
Independence from Irrelevant Alternatives Not
explained by Tverskys (1972) Elimination by
Aspects Model Explained by Tversky Simonsons
(1992) Loss Aversion Model
Y
Ry
Rx
X
25
Attraction Effect (Huber, Payne, Puto, 1982,
Journal of Consumer Research)
Pr X X,Y Regularity Rules out Random Utility Models (e.g.
McFaddens (1982) generalized extreme value model
Explained by Tversky Simonsons (1992) Loss
Aversion Model
Y
D
X
26
Summary of Findings

• Similarity
• Pr(XX,Y,S)
• Attraction
• Pr (XX,Y,D) Pr (YX,Y,D)
• Reference Point
• Pr(XX,Y, RX)Pr(YX,Y, RX)
• Pr(XX,Y, RY)
• Compromise
• Pr (CX,Y,C) Pr (XX,Y,C)

C
S
D
27
Decision field theory predictions
X Pr (X) O Pr (Y) Pr (C)
X Pr (X) O Pr (Y) Pr (Rx)
X Pr (X) O Pr (Y) Pr (Ry)
X Pr (X) O Pr (Y) Pr (D)
X Pr(X) Pr O (Y) Pr (S)
28
Theoretical Requirements for a theory of
motivation and decision making

• Dynamic Model of Decision Making
• Describe the evolution of preferences over time
• Integrates traditional decision concepts
• Probabilities
• Multi-attribute Values
• Integrates traditional motivational concepts
• Need Stimulation and Attenuation
• Satiation Deprivation

29
Example Allocating time between work and
recreation

• Five Conflicting Motives
• Career Achievement
• Financial Security
• Rest and Relaxation
• Fun and Enjoyment
• Family Relations

30
Motivational Values Dynamic Control Problem
G Goal stimulation Q attribute Quantities
A Achievements on attributes A(th) F?A(t)
Q'?B(t) M(t) Q ? DiagN(t) V(t) C? Q ?
DiagN(t)?W(t) N attribute
Needs N(th)L?N(t)G(th)-A(th) P(th) S?
P(t) V(t) B(t) f( P(t) )
31
Motivational Values
Clark Hulls Drive X Incentive N drive Q
incentive Toates Feedback Control System N
state variable G control signal (N A) the
error B feedback controller Simon
(1967) Motivational control over attention
32
Model Example Recreational versus Work Related
Needs Over Time
Person A remains under control
Person B loses control
Stress Intervention at Time 50
33
Return to Effects of Emotion on Decision making
Decision weights (Rottenstreich Hsee. 2001,
Psych Sci) Function is more inverse S-shaped
(flat across the intermediate ranges of
probabilities) for emotional outcomes.
34
Process for Generating weights
G .10 .05 .85
12 90 98
Transitions q13 .10 q11 (1- q13) ? q12
(1- q13) (1-?) q24 .05 q22 (1- q24)? q23
(1- q24) (1-?)/2 q21 (1- q24) (1-?)/2 q32
.85 q33 (1- q32) ? q31 (1- q32) (1-?)
1 0 0 0 0 1 1/3 1/3 1/3 .10 .05 .85
Z start distribution Z z1 z2 z3
Two Free Parameters ??, z
35
General Solution for Weights
P input vector of objective probabilities for
each outcome W output vector of decision wgts
for each outcome Z Initial state vector Q
State transition matrix W Z(I Q)-1P
36
Example ? .2 and Z 1 0 0

• P .10 .05 .85
• ??
• W .22 .10 .68 .
• If ? 1 and Z P then W P
• In this way, the model can still recover the
  original probabilities

37
Solution for binary outcomesWin X with p
otherwise Y with q(1-p), XY
38
Fit of Process model to CPT Wgts
39
Effect of Emotional Outcomes on Decision Weights
Predicted by Increasing the Dwell time for
emotional consequences
Decision weights (Rottenstreich Hsee. 2001,
Psych Sci) Function is more inverse S-shaped
(flat across the intermediate ranges of
probabilities) for emotional outcomes.
40
Conclusions

• Motivation and emotion have an important and
  complex influences on decision making processes.
• Many decision theorists posit a dual system for
  decision making heart vs mind.
• Expected utility theory is used to model
  decisions based on the mind, however no formal
  model is presented for the decisions based on the
  heart.
• Decision field theory provides a formal model
  that integrates the mind and heart into a
  common dynamic system.
• In DFT, Motivation/Emotion moderates the amount
  of attention to consequences.
• This agrees with Simons (1967) hypothesis that
  motivation serves as a control mechanism for
  cognition.