CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 1
Lecture 3Emotion Theory (Part 2)
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 2
“Hot-headed, irrational and swayed by
emotion – who’d want a human in control?”
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 3
Lecture 3Emotion Theory (Part 2)
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 4
Brief note on homework
▪ Lenient grading on HW1 (if you did it, got full credit)
▪ I’ll discuss results during next lecture on computer
models
▪ HW2 (pt 1) assigned after class. This is short
experiment to collect data needed in HW2 (pt 2)
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 5
Review: Introduced some key emotion theories
▪ Theory important– Makes predictions, proposes mechanisms, allows control
– Typically supported by substantial empirical research
▪ But no grand unified theory of emotion– Different theories explain different aspects of emotion
– Different theories make different claims▪ Emotion is discrete vs. emotion continuous
▪ Emotion an atom or vs. emotion a molecule or mixture
▪ Emotion follows from cognition vs. emotion precedes cognition
▪ But choice of theory has implications– E.g., for our labels in machine learning algorithm
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 6
Discrete or continuousRussell’s ‘80 circumplex model
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 7
Appraisal Models
Is emotion a cause or consequence of thought?
Event
Self-reported
Fear
Amygdala
activation
Increased
heartrate
Fight, Flight,
Freeze
Com
ponents
of E
motion
Construct
beliefs,
desires,
intentions
ThinkThink influence
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 8
James-Lange Perspective
Is emotion a cause or consequence of thought?
Event
Amygdala
activation
Increased
heartrate
Fight, Flight,
Freeze
Com
ponents
of E
motion“Mindless”
Automatic Evaluation
influenceThink
Shapes
beliefs,
desires,
intentions
Label
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 9
Emotion components are tightly-
coupled and can be treated as a circuit
linking stimuli and response
Emotions are defined by loose
configuration of different components
Phoebe Ellsworth, Klaus Scherer, Lisa
Feldman Barrett
Compound
Molecule or Mixture“I feel disgusted”
Increased skin
conductance
Vomiting
Dis
gu
st
Cir
cu
it
Insula
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 10
Emotions are defined by loose
configuration of different components
Emotions are defined by loose
configuration of different components
Phoebe Ellsworth, Klaus Scherer, Lisa
Feldman Barrett
Mixture
Molecule or Mixture
Increased skin
conductance
Vomiting
I fe
eld
igu
ste
d
Insula
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 11
In class experiment (tried to induce disgust)
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 12
Tried to experience of emotion
1
2
3
4
5
6
7
8
9
Positivity Arousal Dominance1
2
3
4
5
6
7
8
9
Positivity Arousal Dominance
Discrete
emotions?
Continuous?
1st person:
Experienced video
3rd person:
Guessed feelings
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 13
Results
1
2
3
4
5
6
7
8
9
Positivity Arousal Dominance
Watchers
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 14
Automatic analysis (Facet)
21 seconds (Joy)
disgust
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 15
Automatic analysis (Facet)
19 seconds (disgust)
25 seconds (surprise)
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 16
Automatic analysis (Facet)
29 seconds (Joy)
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 17
Automatic analysis (Facet)
12 seconds (confusion)
sadness
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 18
What’s going on here
Takeaway:
Emotional expression complicated
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 19
Outline
▪ Contrast top-down (appraisal) with bottom-up
(constructivist) theories
▪ In-class experiments on these themes
▪ Discuss dual process models– That emphasize disassociation between emotion and cognition
▪ Discuss more integrative models– That see emotion and cognition as two sides of larger process
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 20
Example: Thought precedes emotion
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 21
BREAKING: Biden’s Executive Order
Revokes Most Student Visas to fight COVID
National
By Tim Craig January 27, 2020 at 12:00 PM
WASHINGTON D.C. — Foreign students at US
universities had their future thrown into question
today by President Biden’s sweeping new executive
order revoking student visas from the Middle East,
and 30 Asian countries including China and India.
Immigration lawyers are warning students not to
leave the country because of the risk that they will be barred from re-entry.
Students already in the country may be targeted for deportation as early as
this March.
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 22
Co
nse
que
nce
s?
Expecte
d?
Co
ntr
ol?
Bla
me?
Self-reported
Fear
Amygdala
activation
Increased
heartrate
Fight, Flight,
Freeze
Com
ponents
of E
motion
Be in US
Need Visa
“Thoughtless”
Automatic
Evaluation
Appraisal
Con
scio
us
“Effort
ful”
thought
In a sense, emotion a byproduct of thinking
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 23
Another example
▪ Imagine the following situation
– You are a foreign student in another country. You discover a mistake
you made with your visa. If it is not fixed, you will be deported
– You spend $10,000 on an immigration lawyer to help you
– The lawyer knows he has to file paperwork by January 15
– He knows if he misses this deadline, you will be deported
– He’s busy and decides not to submit the paperwork
– Immigration agents come to your house and deport you
▪ What are you likely to feel?
▪ Who are you mad at?
– Yourself?
– Immigration agents?
– Lawyer?
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 24
Some appraisals require social inferences (theory of mind) – e.g. Weiner; Shaver
e.g., Causal attribution involved in Anger
NegativeConsequence
Attributional Inferences
Cause Foreseen
Unforeseen
Intended
Unintended
Voluntary
Coerced
No Blame No Blame No Blame No BlameIncreasing responsibility but not blame
Blame
or
Credit
Such reasoning underlies how people make sense of complex social
activities involving the attribution of blame and feelings of anger
Appraisal theory perspective on anger
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 25
Preview: computer appraisal models(Mao&Gratch2005;Oh et al 2007; Melissen)
Scenario 1
1. The VP British Petroleum discusses plans for new oil well
2. The VP states the program will likely increase profits
3. The VP states the program will likely harm the environment
4. The CEO orders the program to be started
5. The VP executes the new program
6. The environment is harmed
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 26
Preview: computer appraisal models(Mao&Gratch2005;Oh et al 2007; Melissen)
Scenario 1
1. The VP British Petroleum discusses plans for new oil well
2. The VP states the program will likely increase profits
3. The VP states the program will likely harm the environment
4. The CEO orders the program to be started
5. The VP executes the new program
6. The environment is harmed
VP: Cause
CEO: Cause + Intent + Voluntary → Blame
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 27
Preview: computer appraisal models(Mao&Gratch2005;Oh et al 2007; Melissen)
Scenario 1
1. The VP British Petroleum discusses plans for new oil well
2. The VP states the program will likely increase profits
3. The VP states the program will likely harm the environment
4. The CEO orders the program to be started
5. The VP executes the new program
6. The environment is harmed
VP: Cause + Intent
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 28
Preview: computer appraisal models(Mao&Gratch2005;Oh et al 2007; Melissen)
Scenario 1
1. The VP British Petroleum discusses plans for new oil well
2. The VP states the program will likely increase profits
3. The VP states the program will likely harm the environment
4. The CEO orders the program to be started
5. The VP executes the new program
6. The environment is harmed
VP: Cause + Intent + ¬Voluntary → ¬Blame
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 29
Preview: computer appraisal models(Mao&Gratch2005;Oh et al 2007; Melissen)
Scenario 1
1. The VP British Petroleum discusses plans for new oil well
2. The VP states the program will likely increase profits
3. The VP states the program will likely harm the environment
4. The CEO orders the program to be started
5. The VP executes the new program
6. The environment is harmed
VP: Cause + Intent + ¬Voluntary → ¬Blame
CEO: Cause + Intent + Voluntary → Blame
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 30
Scenario 1E1 The vice president of Beta Corporation goes to the chairman
of the board and requests, “Can we start a new program?”
E2 The vice president continues, “The new program will help us
increase profits,
E3 but according to our investigation report, it will harm to the
environment.”
E4 The chairman answers, “Start the program anyway.”
E5 The vice president executes the new program.
E6 However, the environment is harmed by the new program.
Appraise Representation
Intentional action
intend(x, p, t1) do’(p, x, A) t1<t3
(t2)(t1<t2<t3
intend(x, p, t2)) execute(x, A, t3)
Side effect
effect(A) intend(x, b, t1) by(b, A, e)
(t2)(t1<t2<t3 intend(x, b, t2))
t1<t3<t4 execute(x, A, t3)
occur(e, t4)
Coercion
coerce(y, x, p, t1) do’(p, x, A)
eeffect(A) t1<t2<t3
etc29(x, y, e, t2) coerce(y, x, e, t3)
BLAME = YES
Preview: computer appraisal models(Mao&Gratch2005;Oh et al 2007; Melissen)
Counterfactual Scenario 2E1 The vice president of Beta Corporation goes to the chairman
of the board and requests, “Can we start a new program?”
E2 The vice president continues, “The new program will help us
increase profits,
E3 AND according to our investigation report, it WON’T harm to the
environment.”
E4 The chairman answers, “Start the program.”
E5 The vice president executes the new program.
E6 However, the environment is harmed by the new program.
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 31
Appraisal theory perspective(emphasizes primacy of deliberative thought)
▪ Emotion is “goal relevant”
▪ It arises from how events impact goals
▪ The emotion prepares your body and mind to
address goal threats or opportunities
▪ The emotion is said to be “endogenous”– Meaning that it arises from thoughts about a task at hand
▪ Thinking determines emotion
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 32
Different example
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 33
James-Lange Perspective
Is emotion a cause or consequence of thought?
Event
Amygdala
activation
Increased
heartrate
Fight, Flight,
Freeze
Com
ponents
of E
motion“Thoughtless”
Automatic
Evaluation
influenceThink
Shapes
beliefs,
desires,
intentions
Emotion determines thought
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Yet another example (in class experiment)
▪ I’ll post Qualtrics survey
▪ You will listen to music
– May not play automatically on Macs
▪ Listen quietly for 2 minutes then complete survey
▪ Try to go with your first instincts (don’t use math)
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
This is the Education Building on USC
campus. Imagine that you have to walk to
the top floor using the stairwell
Don’t think carefully, don’t count how many
floors. Just from this image, guess how
many minutes it will take you to walk to the
top.
Have blank that will accep integers or real
numbers
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
What should have happened?
▪ Mozart induces happiness
▪ Happiness makes you energetic,
confident, (over) optimistic
▪ Mahler induces sadness
▪ Sadness makes you lethargic,
unconfident, realistic
▪ These impressions bias appraisals
– Particularly perceived control
Riener, Stefanucci, Proffitt & Gerald Clore (2011) An effect of mood on the perception of geographical
slant, Cognition and Emotion, 25:1, 174-182
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Thinking
James-Lange Perspective
Is emotion a cause or consequence of thought?
Event
Amygdala
activation
Increased
heartrate
Fight, Flight,
Freeze
Com
ponents
of E
motion“Mindless”
Automatic Evaluation
Shapes
beliefs,
desires,
intentions
influence
Make a
time
judgement
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
James-Lange perspective
▪ Characteristics of events automatically induce
emotional responses
▪ These responses can alter decision making
▪ The emotion is said to be “incidental” to what was
previously “in mind”
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Stepping back
▪ Some work suggests primacy of thinking
– Thought shapes emotion
▪ In that thought is (somewhat) under our control,
suggests that we can control emotion
Emotion Cognition
“Automatic” appraisals of
the contents of thought
Lazarus, R. S. (1984). On the primacy of cognition. American Psychologist, 39(2), 124-129.
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Stepping back
▪ Some work suggests primacy of thinking
– Thought shapes emotion
▪ Some work suggests primacy of affect
– Emotion shapes thought
Emotion Cognition
Perception
“Automatic” stimulus
evaluation
Zajonc, R. B. (1984). On the primacy of affect. American Psychologist, 39(2), 117-123
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Stepping back
▪ Some work suggests primacy of thinking
– Thought shapes emotion
▪ Some work suggests primacy of affect
– Emotion shapes thought
▪ Suggests emotion is unconscious and difficult to
control
Emotion Cognition
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Stepping back
▪ Some work suggests primacy of thinking
– Thought shapes emotion
▪ Some work suggests primacy of affect
– Emotion shapes thought
▪ Together, seem to emphasize two distinct processes
Emotion Cognition
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 43
Most emotion theories are dual-process theories
Passion,
visceral reward
The Allegory of the Chariot
Abstract goals,
Deliberation
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 44
System 1 System 2Integration
Behavior
Most emotion theories are dual-process theories
Together, these systems determine behavior
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 45
System 1 vs System 2: Kahneman
A useful fiction
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 46
Affective system
▪ fast
▪ unconscious
▪ reflexive
▪ myopic
▪ effortless
Analytic system
▪ slow
▪ conscious
▪ reflective
▪ forward-looking
▪ (but still prone to error:
heuristics may be analytic)
▪ self-regulatory
▪ effortful and exhaustible
Commonalities between classification schemes
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 47USC CSCI 534 Affective Computing © Jonathan Gratch
Quick illustration of automatic processing
If you are a
woman:
Raise your hand
when you find the
face
If you are a man:
Raise your hand
when you find the
gloves
I will show a bunch
of pictures
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 48USC CSCI 534 Affective Computing © Jonathan Gratch
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 49
Mesolimbic dopamine reward system
Frontalcortex
Parietalcortex
Affective vs. Analytic Cognition
mPFC
mOFC
vmPFC
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 50
▪ The brain makes decisions (e.g. constructs value)
by integrating signals from multiple systems
▪ These multiple systems process information in
qualitatively different ways and in some cases
differentially weight attributes of rewards (e.g.,
time delay)
Dual Process Theories: more recent perspective
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 51
▪ Passions vs Interests (Smith)
▪ Id vs Ego vs Superego (Freud)
▪ Automatic vs Controlled (Schneider & Shiffrin, 1977; Benhabib & Bisin, 2004)
▪ Hot vs Cold (Metcalfe and Mischel, 1979)
▪ Impulsive vs Deliberative (Frederick, 2002)
▪ Unconscious vs Conscious (Damasio, Bem)
▪ Effortless vs Effortfull (Baumeister)
▪ Doer vs Planner (Shefrin and Thaler, 1981)
▪ Visceral vs Abstract (Loewenstein & O’Donoghue 2006; Bernheim & Rangel, 2003)
▪ Mesolimbic dopamine vs PFC & parietal cortex (McClure et al, 2004)
▪ System 1 vs System 2 (Frederick and Kahneman, 2002)
Variations on this theme
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 52
A battle in your mind
Dual process theories often adopt the analogy
of two independent entities fighting for control
of your mind
Research then tries to identify which “horse” wins
• Developmental factors: children and teenagers are emotional
• Dispositional factors: some people are naturally emotional
• Cultural factors: some societies are naturally emotional
• Situational factors: some situations evoke emotional thinking
(e.g., when hungry, sleepy, rushed
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 53
An example
Visceral
reward:
pleasure
Abstract
goal:
dietIntegration
Behavior
Would you like a piece of chocolate?
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 54
Individual differences (simple test)
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 55
A battle in your mind
Dual process theories often adopt the analogy
of two independent entities fighting for control
of your mind
Research then tries to identify which “horse” wins
• Developmental factors: children and teenagers are emotional
• Dispositional factors: some people are naturally emotional
• Cultural factors: some societies are naturally emotional
• Situational factors: some situations evoke emotional thinking
(e.g., when hungry, sleepy, rushed
But are these really independent systems?
Is there always just one winner?
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Buyer beware
▪ Psychological research often overemphasizes
independence of these systems
▪ This independence is exaggerated by experimental
designs that try to emphasize differences in thinking
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Reductionism
▪ Psychological studies typically try to show importance
of a single mechanism
▪ Can overestimate importance of that mechanism and
underestimate contribution/interaction with other
mechanisms
IV →DV
Race
Gender
Degree
Sunny?
Hungry
Sleepy
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Mechanically Separate emotion and cognition
▪ Ventral Medial/Orbital Prefrontal
Cortex damage
Phineas Gage
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Mechanically Separate emotion and cogition
▪ Ventral Medial/Orbital Prefrontal
Cortex damage
– Show can “turn off” emotion and
“thinking” preserved
– And show that “thinking” “needs”
emotion▪ Severe impairments in judgment and decision-
making in real-life
– But keep in mind this sort of
“separation” never occurs in real world
Transcranial magnetic stimulation
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Empirically separate Em & Cog “in the moment”(Clore, Schwarz)
▪ Emotions inform decisions
▪ But many experiments separate emotion from decision
– Induce an emotion:▪ Play happy/sad/angry music
▪ Read happy/sad/angry stories
– Make people perform an irrelevant task▪ Buy something
▪ Play ultimatum game
– Show logically irrelevant emotion biases decision making
▪ But how often is emotion irrelevant to the task?
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
▪ One CAN separate emotion and cognition
– Its seductive▪ Reflect longstanding theoretical and folk distinctions
▪ Consistent with some data
– But I’ll argue this data has limited ecological validity (e.g., see also
Gigerenzer)
▪ It is fun (and publishable) to show people are irrational
▪ But how often does this occur in real word
– But this leads to impoverish understanding of both▪ Cognition w/o emotion is a broken thing
Maybe dual processes are a cognitive illusion
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 62
Are these really separate systems?
▪ Much psychological work focuses on one system or
the other
▪ But some work tries to show how these systems
are more tightly linked
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
emotion AND cognition
▪ The majority of everyday “thinking” involves
– Acting in a dynamic and evolving world
– Juggling multiple goals and preferences
– Confronting opportunities and threats
▪ Emotion evolved hand and hand with cognition
– Two sides of the same system
▪ Attempts to separate them leads to anomalous behavior
▪ Yet that is what much of emotion psychology implicitly
or explicitly strives to do
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Appraisal Models
Is emotion a cause or consequence of thought?
Event
Self-reported
Fear
Amygdala
activation
Increased
heartrate
Fight, Flight,
Freeze
Com
ponents
of E
motion
Construct
beliefs,
desires,
intentions
ThinkThink influence
influence
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 65
Some theories show how these processes linked
Emotion CognitionIntegration
Behavior
Perception
Imagination
Perception and imagination
may use same neural circuits
(Damasio’s “somatic marker
hypothesis”)
Cognitive emotions strongest
when thoughts are vivid and
personal
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
EmotionAction
Tendencies“Affect”
PhysiologicalResponse
EnvironmentGoals/Beliefs/
Intentions
Another example (Lazarus, Smith, Gratch & Marsella)
Desirability
Expectedness
Controllability
Causal Attribution
Appraisal
Antecedents of Emotion
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
EmotionAction
Tendencies“Affect”
PhysiologicalResponse
EnvironmentGoals/Beliefs/
Intentions
Appraisal theory explains antecedents of emotion
Desirability
Expectedness
Controllability
Causal Attribution
Appraisal
Antecedents of Emotion
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
EmotionAction
Tendencies“Affect”
PhysiologicalResponse
EnvironmentGoals/Beliefs/
Intentions
Appraisal theory explains antecedents of emotion
Desirability
Expectedness
Controllability
Causal AttributionAntecedents of Emotion
Controlled thinking process
(e.g., planning, deliberation)
auto
matic
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Emotion
CopingStrategy
ActionTendencies
“Affect”Physiological
Response
EnvironmentGoals/Beliefs/
Intentions
“Coping” theory explains emotion’s consequence
Desirability
Expectedness
Controllability
Causal Attribution
Consequences of Emotion
Controlled thinking process
(e.g., planning, deliberation)
Automatic
thinking “bias”
(action tendency)
auto
matic
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Emotion
CopingStrategy
ActionTendencies
“Affect”Physiological
Response
Withdraw /
Act on self
Emotion-Focused
EnvironmentGoals/Beliefs/
Intentions
Coping
Approach /
act on world
Problem-Focused
Attempts to characterize How emotion shapes cognition
Consequences of Emotion
e.g., Call
immigration
lawyer
e.g., go home
and get a job
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
(act on world) (act on self)
Emotion
CopingStrategy
ActionTendencies
“Affect”Physiological
Response
Problem-Focused Emotion-Focused
EnvironmentGoals/Beliefs/
Intentions
Resignation
Distancing
Wishful Thinking
Take action
Seek support
Coping
Attempts to characterize How emotion shapes cognition
These models emphasize that coping
responses are largely automatic
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
▪ Coping responses help explain common decision bias– Principle of rationality
– Desires (i.e., emotion) shouldn’t change beliefs (and vice versa)e.g., Just wanting something shouldn’t make it true
– Preferences fixed over time
Participate in an
Election
Lose
Win
p=.8
p=.2
Utility= 20
Intention(vote) EU=4
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Wishful Thinking“we will win Georgia”
Participate in an
Election
Lose
Win
p=.8
p=.2
Utility= 20
Intention(play) Sad
HopeJoy
Distancing“it doesn’t matter
who wins”
Fear
Utility= 10p=.6
p=.4
Resignation “I’m not going to vote”
• Coping serves to “confound” beliefs and desires
▪ Emotion-biases on decision making (Loewenstein & Lerner, 2003)
▪ Cognitive dissonance (Festinger57)
▪ Motivated inference (Kunda87)
– Little attempt to computationally model (Marsella&Gratch; Dias)
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 74
Coping can impact many aspects of appraisal process
▪ E.g., James Gross emotion regulation theory
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 75
▪ E.g., James Gross emotion regulation theory
– Emotion arises from appraisal
– But emotion regulation seeks to modulate aspects of the appraisal process
– This theory relies on interdependence of appraisal and emotion over time
Coping can impact many aspects of appraisal process
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 76
People differ in automatic coping styles
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 77
Locus of control: self vs. other
▪ Internalizes blame, externalizes credit: “god’s plan”
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 78
Stability of control: changeable vs. constant
▪ Views situation as unchangeable across time
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 79
People differ in “coping style”
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 80
People differ in “coping style”
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 81
Cognition can also impact coping
▪ Coping styles tend to be automatic
▪ But people can learn new patterns of coping
▪ Cognitive behavior therapy tries to help people
relearn adaptive coping patterns
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch
Some appraisal theories join both perspectives(Clore & Schwartz; Gratch and Marsella)
Endogenous
Emotion
Incidental
Emotion
Allow that prior emotions or incidental influences can shape appraisals
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 83
Coping shapes beliefs, desires and intentions
▪ Thought → Appraisal
▪ Appraisal → Emotion: – I’m afraid because I might lose
▪ Incidental influences (music) → Emotion
▪ Emotion → Coping– I don’t care about winning anyway
▪ Coping → Thought– I’m much happier now that I don’t care about wining
▪ This is a cycle
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 84
Integrated model also explains individual
difference in emotional responding
▪ Different individual goals– UCLA fan vs USC fan
▪ Different appraisal styles
▪ Different coping / regulation strategies
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 85
Appraisal theory
▪ For much of the rest of the class I’ll emphasize
appraisal theory (esp. work of Smith and Lazarus)– Emphasizes emotion as both an antecedent and consequence of
cognition▪ Provides detailed description of factors that elicit emotion (appraisal)
▪ Provides detailed description of how emotions can shape sequent cognition
▪ Thus can unify appraisal and constructivist approaches
– Relatively easy it translate into a computer program
– Serves as the theory underlying my own work on affective computing
– Can help explain several aspects of emotion▪ Why a given situation might produce a given emotion
▪ Why an emotion might influence subsequent decisions
▪ Why an expression might shape another’s decision
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 86
Summary
▪ Different theories of emotion– Emotion coherent/basic circuit vs. collection of loosely related systems
▪ Discussed dual process view of emotion– Emotion cause by thinking vs. automatic unconscious response
– Both perspectives seem active in many situations
▪ Introduced Lazarus’ appraisal model– Incorporates both deliberative and automatic processes into single
view
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 87
Some comments on projects
▪ Main source of your grade
▪ Project is intended to give you a deep and “hands on”
understanding of affective computing.
▪ Ideal group size is 4 (maybe 5) students.
▪ Projects can involve building affectively aware software or
proposing and piloting an empirical study involving humans
interacting with affective technology.
▪ List of pre-existing software tools available to students, and
summaries of prior student projects built with these tools can
be found
– http://people.ict.usc.edu/~gratch/CSCI534/Tools-and-projects-2020.pdf
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 88
Some comments on projects
▪ Expectations– not expecting/requiring projects to make novel scientific advance over
the state-of-the-art in affective computing research (of course, this
would be great). Projects that replicate prior findings in order to give
you a better grasp of a theory, algorithm, or phenomena are
fine/expected).
▪ Project Milestones– Feb 22: in class discussion of potential projects and teams
– Feb 25 (midnight): One paragraph tentative project proposal (list team)
– Mar 3: 5min in-class group project plan presentation
– April 26,28: 20min in class final presentation
– May 7 (11:59p): Final written report
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 89
Some comments on projects
▪ For mid-term presentations– Prepare between 5 minutes of material. Be prepared for heavy
questioning from me and the class.
– Should include group members
▪ Project proposal presentation should address:– What is your general area of interest (e.g., emotion recognition,
emotion modeling, emotion synthesis)?
– What is the problem you hope to address within this area?
– How do you plan to address this problem?
– How would you know if you succeeded?
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 90
Some comments on projects
▪ Final project should address the following points– Why this area is interesting (e.g., potential applications, open scientific
questions)?
– What are related approaches? are they inadequate? i.e., is your
proposal an advance over state-of-the art?
– What (if any) of the theoretical perspectives on emotion (introduced in
class) does your approach build on, question, test?
– If you are performing some empirical test of success▪ What are your hypotheses
▪ What are your independent, dependent, and control variables
▪ What are your results and are they statistically significant
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 91
Some comments on projects
▪ Final team presentations should be 15min.
▪ Writeups should be minimum of 5 pages – (12pt single-column, single space, including title, authors, figures and references).
▪ Both presentation and writeup should at least briefly address – Team-members and role played by each member.
– What is the problem you hope to address within this area?
– Why is this interesting (e.g., potential applications, open scientific questions)?
– What are related approaches? are they inadequate?
– What is the evidence you succeeded (e.g., theoretical or empirical arguments)
– What (if any) of the theoretical perspectives on emotion (introduced in class) does your
approach build on, question, test?
– If you are performing some empirical test of success
▪ What are your hypotheses?
▪ What are your measures
▪ Describe the population of subjects. How were they recruited? Is this between subjects? Within
subjects? Do subjects know the hypotheses? (hope not)
▪ What are your results and are they statistically significant
– Anything surprise you?
CSCI 534(Affective Computing) – Lecture by Jonathan Gratch 92
Speed Dating Exercise
▪ For rest of class, do couple rounds of speed date
▪ I’ll put you in breakout rooms
▪ Take turns introducing yourself– Who you are: I’m a senior in computer science…
– Why you took the class
– Tentative project ideas: I want to make a horror game…
▪ I’ll jump around rooms
▪ I’ll reassign groups after few minutes