A brain process that computes the value of an experience --- Le Doux

Transcription

1 What is an emotion? The recognition of a change of feeling God s punishment for disobedience -- St Augistine A brain process that computes the value of an experience --- Le Doux A word we assign to certain configuration of bodily states, thoughts, and situational factors Feldman Barrett. 1

2 What is emotion NOT? From Scherer 2

3 Do emotion s exist? Are emotions natural kinds Are they reflected in the structure of nature or are they concepts we impose on a lose collection of related structures Basic/Discrete emotion view (e.g. Ekman, Le Doux ) Some of what we call emotions correspond to brain circuits Joy, Hope, Rfear, (Joy, Hope, Fear, Sadness, Anger, Disgust) Distinguished from more complex affective states Love, optimism, awe, remorse, shame E.g. Le Doux fear circuit 3

4 Do emotion s exist? Core Affect View (Russell, Feldman Barrett) Emotions don t exist as discrete circuits Emotions correspond differences in body state distinguished in terms of broad dimensions (e.g., valence and arousal) 4

5 What is an emotion? Component view 5 episode of massive synchronized recruitment of mental and somatic resources allowing to adapt to cope with stimulus event subjective appraised as being highly pertinent to needs, goals, and values of the individual. -- Klaus Scherer Components Cognitive elements: Goals, needs, values / Subjective appraisal Somatic elements: Physiological activation / Display / Action tendency Feelings: some authors distinguish these cognition Prototypical characteristics High intensity Synchronization across components Adaptive / Coping

6 Appraisal Theory Affective Computing: Lecture 2 Stacy Marsella Jon Gratch

7 Emotion Cognitive Emotion is a mental phenomena Arises from physiological & cognitive processes 7 Associated with key cognitive functions Focusing mental, sensory resources Influencing beliefs Informing decision-making Preparing action and reaction Learning and long-term adaptation

8 Emotion Cognitive Physical Emotion is a physical phenomenon Emotion processes associated with specific physical behaviors Facial expression Body language and posture Voice Behavioral dynamics 8

9 Emotion Cognitive Physical Emotion is a social phenomenon Emotion is a signal that influences the behavior of others Emotional contagion Social referencing Social 9

10 Unified Understanding Why our we interested in modeling emotion? Cognitive Goal is to develop unified computational framework for modeling, simulating, explaining and exploiting these phenomena Physical Social To inform intelligent system design By drawing on insights from emotion s cognitive/social function Simon, 1968; Minsky; Scheutz, &Sloman, 2001 To inform emotion theory By concretizing theory and developing methodological tools To drive applications Education, Training and Health Interventions 10

11 Application Areas: Social Learning Environments Gratch&Marsella04; Diaz04. User modeling Infer student motivational state (Conati&MacLaren, 2004) Entertainment Creating Emotion in Games 11

12 Social Learning Environments Explore emotionally charged social interactions in safety of VR Virtual Role Play Learner interacts with Virtual Humans

13 Virtual Humans Interact with users Computer-generated simulations of humans Play role of mentors, peers, competitors, etc. Communicate thru speech and gesture Coordinated gestures and non-verbal communication Reason about environment Behavior not pre-scripted Reason about possible responses to events Understand & express emotions Affects way they perform tasks & interact socially Affects gaze, face, gestures, posture

14 ICT Movie

15 Virtual Human Applications Education and Training Intelligent tutoring systems (Lester et al; Graesser) Health care provider training (Lok et al) Diagnosis & Assessment ADHD Assessment (Rizzo et. al) Health Interventions Mental health (Marsellaet al) Autism (Tartaro&Cassell) Informal Science Education Museum Guides (Boston Science Museum, Cooper-Hewett) Entertainment Façade (Mateas and Stern)

16 Theoretical Issues

17 Theoretical Views Categorical vs dimensional Categorical Tomkin s 8: fear, anger, anguish, joy, disgust, surpass, interest and shame Ekman s six basic emotions Culturally universal Dimensional: Russell s circumplex: Arousal & Valence (Pleasure) Russell &Mehrabian s PAD model: Pleasure, Arousal & Dominance Appraisal Theories Utility for computational work? Virtual Humans, Behavior animation, Speech

18 Theoretical Views (cont) Function of Emotion Stress on Social Coordination E.g., Parkinson Stress on Individual Regulation and Adaptation E.g., Lazarus Function of Emotional Expression Social Signal Reflection of internal state Part of adaptive behavior Example: Eyelid and eyebrow lift

19 Theoretical Issues (cont): Top down Emotion arises from cognitive processes Appraisal Theories Stimulus > Cognition/Appraisal > feeling Bottom up Physiological and perceptual processes drive Zajonc Stimulus > unconscious affect > feeling James Stimulus > response > feedback > feeling Pathways go both ways May differ for different emotions

20 20 Dimensional Appraisal Models AR Elliott EM Neal Reilly FLAME El Nasr Emile Gratch CBI Marsella Lazarus EMA Gratch/Marsella FearNot! Dias NAFO Marinier Cathexis Velásquez Frijda Armony Scheultz Mehrabian ALMA Gebhard MAX Becker MAMID Hudlicka ACRES Swagerman OCC Will Moffat Sloman Neuro/Ethology (e.g., LeDoux, Damasio) Cognitive Architecture Perspective HOTCO Thagard NML1 Beaudoin Minder1 Wright Alvila-Garcia, Canamero

21 Diagram comments Draw on Different Theories Grounded in different experimental techniques Appraisal Theories dominant Creating models from particular theories versus pulling from different theories Models of specific affective phenomena versus more general models Built for different Applications Different computational techniques Lineage 21

22 Cognitive Modeling the cognitive antecedents and consequences of emotion

23 Theories of cognitive emotion Top down theories Cognition influences emotion Magda Arnold Appraisal Theory (Arnold, Lazarus, Frijda, Scherer, Ortony et al.) Emotion arises from an evolving subjective interpretation of person s relation to their environment and informs cognitive and physical acts

24 Appraisal Theory Smith and Lazarus91 cognitive-motivational-emotive system Environment Desirability Appraisal Expectedness Controllability Goals/Beliefs/ Intentions Action Tendencies Causal Attribution Appraisal Variables Physiological Affect Response Coping Problem-Focused (act on world) Coping Strategy Emotion-Focused (act on beliefs)

25 Appraisal Theory Smith and Lazarus91 cognitive-motivational-emotive system Environment Desirability Appraisal Expectedness Controllability Goals/Beliefs/ Intentions Action Tendencies Causal Attribution Appraisal Variables Physiological Affect Response Take action Seek support Problem-Focused (act on world) Coping Emotion-Focused (act on beliefs) Form/drop goal Form/drop belief Form/drop intention

26 Appraisal External World Appraisal Internal goals, beliefs Appraisal = Situation assessment Compare beliefs, desires and intentions with external circumstances

27 Appraisal Characterize via appraisal variables Desirability Does this event help/hurt my goals Likelihood How likely is it that this event will occur Unexpectedness Was this event expected Causal attribution (causality, agency, blame/credit) Who deserves blame for causing the event Coping potential (controllability, adaptability) What chance do I have for dealing with this event Urgency Superset of criteria considered by intelligent systems Decision theory: desirability, likelihood Scheduling: desirability, urgency

28 Appraisal Emotions defined in terms of configurations of appraisal variables Undesirable, Uncertain Fear Desirable, Certain Joy Undesirable, Caused-by(Other) Anger-at(Other)

29 Coping Strategies Coping = Response strategy Characterized by ontology of coping strategies External World Emotion Internal goals, beliefs Coping Appraisal Appraisal Problemfocused Emotionfocused

30 Coping Strategies Problem-focused (act on the world) Action execution Planning Seek instrumental social support Analogous to: Deliberative or reactive problem solving Team negotiation

31 Coping Strategies Emotion-focused (act on belief) Denial Find positive meaning Resignation Shift blame Distancing Not typically considered by intelligent systems More than a decision (e.g. abandon current plan) Provides self-justification for why Related to motivational / explanatory coherence Leads to persistent change in behavior Helps find ecological niche

32 Appraisal Theory as architectural specification Provides high-level requirements How do we map this into an architecture How do we represent the person-environment relation? How do appraisal processes operate over this representation? What is the relation between the processes of Appraisal Cognition Coping How do these interactions interact/unfold over time?

33 EMA Gratch & Marsella

34 EMA Model of Appraisal and Coping Cognition as multiagent planning/envisionment Causal Interpretation: Working memory of plans, beliefs, desires, intentions Past Events Past Act Cause: Other Intend: yes Prob: 100% Present Goal Utility: 50 Probability: 100% Belief: False Inhibits Future Plans Future Act Cause: self Intend: yes Probability: 50% Goal Utility: 50 Probability: 50% Intend-that: True Facilitates Planning Perception Dialogue Action Cognitive Operations (inference)

35 Appraisal Past Present Future Past Act Cause: Other Intend: yes Prob: 100% Goal Utility: 50 Probability: 100% Belief: False Inhibits Future Act Cause: self Intend: yes Probability: 50% Goal Utility: 50 Probability: 50% Intend-that: True Facilitates Threat Desirability: -50 Likelihood: 100% Causal Attribution: Other Coping Potential: moderate Emotion: Anger(50) Challenge Desirability: 50 Likelihood: 50% Causal Attribution: self Coping Potential: Moderate Emotion: Hope(25)

36 Coping Past Present Future Past Act Cause: Other Intend: yes Prob: 100% Goal Utility: 50 Probability: 100% Belief: False Inhibits Future Act Cause: self Intend: yes Probability: 50% Goal Utility: 100 Probability: 50% Intend-that: True Facilitates Threat Desirability: Likelihood: 100% 50% Causal Attribution: Other Coping Potential: moderate Low Emotion: Sadness(50) Anger(50) Challenge Desirability: 50 Likelihood: 50% Causal Attribution: self Coping Potential: Moderate Emotion: Hope(25) Resignation (abandon goal)

37 Modeling Appraisal & Coping Appraisal as evaluation of the causal interpretation Define appraisal variables in terms of features of interpretation Fast, automatic Coping: Operators that suggest ways to change interpretation Sequential, deliberate, mediated by focus of attention Problem-focused Take Action, Make Plans Emotion-focused Denial/Wishful Thinking Change belief / likelihood Find silver lining Change utilities Shift blame (socially) Dialogue moves Change causal attribution Distancing/acceptance Drop goal / intention Avoidance Change topic Add goal (e.g., go to party) Key: A Content & Process Model

38 Example: Modeling Naturalistic Emotion

39 Emotional Dynamics Surprise Fear Anger Empathy Humor/Relief

40 Collaboration with Klaus Scherer s Geneva Emotion Group Scherer s Sequential Checking Model Appraisal Relevance Environment EMA Working Memory Implication Coping Potential Normative Significance Action Appraisal Frames Appraisal Coping Affective State Control Signals Inference

41 Scherer s Sequential Checking Model Appraisal Relevance Dynamics in the world Environment Marsella & Gratch, JCSR 2008 EMA Working Memory Implication Coping Potential Normative Significance Action Appraisal Frames Appraisal Coping Affective State Control Signals Inference

42 Scherer s Sequential Checking Model Appraisal Relevance Dynamics in the world Environment Marsella & Gratch, JCSR 2008 Dynamics in perceived world relationship Working Memory Implication Coping Potential Normative Significance Action Appraisal Frames Appraisal Coping Affective State Control Signals Inference

43 Scherer s Sequential Checking Model Appraisal Relevance Dynamics in the world Environment Marsella & Gratch, JCSR 2008 Dynamics in perceived world relationship Working Memory Implication Coping Potential Normative Significance Action Dynamics through action Appraisal Frames Appraisal Coping Affective State Control Signals Inference

44 Bird Flies In Personal Health Desire: Self (100) Satisfied: True Probability: 80% Appraisal Perspective: Self Unexpectedness: High Controllability: Low Blame/Credit: unresolved Seek Info Surprise Coping: Seek More Information Virtual Actor s Appraisal of Bird Flying in Window

45 Bird Flies In Bird Attack Responsibility: Bird Move Away Threatens Personal Health Desire: Self (100) Satisfied: False Probability: 60% Appraisal Perspective: Self Desirability: Undesirable Controllability: Low Blame/Credit: Bird Avoid Threat Fear: 48 Coping: Avoid Further Inference: The Bird will attack

46 Bird at safer dx Bird Flies In Bird Attacking Responsibility: Bird Threatens Whack Bird Threatens Personal Health Desire: Self (100) Satisfied: False Probability: 50% Appraisal Perspective: Self Desirability: Undesirable Controllability: Medium Blame/Credit: Bird Attack Threat Anger: 48 Coping: Plan Re-Appraisal & Further Planning

47 Emotion as Meta-control system Environment Causal Interpretation (Goals, Beliefs, Causal Relations, Plans, Intentions) Action Appraisal Plan Inference Dialogue Appraisal Frames Affective State Belief Formation Coping Explanation Control Signals Dynamics a function of coping/cognitive processes and evolving situation

48 This is the Seeley G. Mudd Psychology building How many minutes will it take to climb the stairwell to the top? 48

49 Mood induction Experimental manipulations can impact a person s mood Happy/Sad Music Imagined situations Movies Weather These manipulations seem to impact unrelated decisions Sad music will make hills seem higher, math problems harder Performing experiment in disgusting room will increase sense of moral outrage This has been called the Misattribution of emotion (Clore) People use feelings to inform whatever decision is in mind Cognitive heuristic? 49

50 Views on the influence between emotion and cognition Direct Influence Emotion as Mediator Emotion as Information (Affect as information) Elicitor (e.g. Appraisal) Appraisal Other factors Appraisal Other factors causes Emotion Brain/Body Configuration Cognitive/Behavioral Changes causes Cog/Behav Changes Zajonc causes Feeling causes Cog Changes 50