Alternative Theoretical Perspectives on Emotion Representation & Modeling
- 1. Alternative Theoretical Perspectives on Emotion Representation & Modeling Eva Hudlicka Psychometrix Associates Blacksburg, VA [email protected] psychometrixassociates.com EmoSPACE2011 Workshop 9th FG 2011 March 21, 2011 Santa Barbara, CA Hudlicka EmoSPACE2011 1
- 2. Outline • Definition • Three Theoretical Perspectives • Similarities & Differences • Implications for Affective Modeling • Summary & Conclusions Hudlicka EmoSPACE2011 2
- 3. Emotions • Evaluative judgments of the: – World – Others – Self • … in light of agent’s goals & beliefs • …motivating & coordinating adaptive behavior Hudlicka EmoSPACE2011 3
- 4. Emotions Are Multimodal • Manifested across multiple, interacting modalities: – Physiological / Somatic (neuroendocrine - e.g., heart rate, GSR) – Cognitive / Interpretive (“Nothing is good or bad but thinking makes it so…”; appraisal, biasing effects) – Motor / Behavioral (expressive, action oriented) – Experiential / Subjective (conscious experience) Hudlicka EmoSPACE2011 4
- 5. Dominant Theoretical Perspectives • Discrete / Categorical (Tomkins, Izard, Ekman, Panskepp…) • Dimensional (Wundt, Osgood, Lang, Russell, Thayer, Watson, Tellegen, Mehrabian…) • Componential (focus on cognitive appraisal) (Scherer, Roseman, Reisenzein, Smith, Ellsworth, Frijda, Ortony…) Hudlicka EmoSPACE2011 5
- 6. Discrete / Categorical • A small number of ‘hardwired’ basic emotions – Joy, fear, anger, sadness, disgust, surprise.. • Characterized by distinct patterns of: – Triggering stimuli – Cognitive processing – Behavioral tendencies – Expression Hudlicka EmoSPACE2011 (Cohn, 2006) 6
- 7. Dimensional • Emotions defined in terms of 2 or 3 dimensions – 2D: Pleasure & Arousal – 3D: Pleasure & Arousal & Dominance • 4th dimension recently proposed: – Unpredictability (Fontaine, Scherer et al., 2007) (Breazeal, 2003 adapted Hudlicka EmoSPACE2011 from Russell, 1997) 7
- 8. Dimensional from de Groot & Broekens, 2003 – adapted from Mehrabian, 1995 Hudlicka EmoSPACE2011 8
- 9. 2 or 3 Dimensions? • 2D - cannot distinguish between emotions sharing same P & A values – Negative valence, positive arousal – Anger? Fear? • Need 3rd dim. to distinguish --> dominance – Anger: - val., + arousal, + dominance – Fear: - val., + arousal, - dominance Hudlicka EmoSPACE2011 9
- 10. Componential • Emotions characterized by synchronized activity across multiple modalities – Cognitive, Physiological, Expressive, Behavioral, Subjective • Cognitive modality & appraisal characterized by appraisal variables (dimensions) Hudlicka EmoSPACE2011 10
- 11. Stimuli Relevance Implications Coping Norms Appraisal Variables Novelty Valence Goal relevance Certainty Urgency Goal Emotion congruence Agency Coping potential Hudlicka Norms EmoSPACE2011 11
- 12. STIMULI FEAR Novelty high Valence low Goal high relevance Agency other Outcome high probability Goal low congruence Urgency v. high Coping potential low Hudlicka EmoSPACE2011 12
- 13. OCC Theory • Ortony, Clore, Collins (OCC): Cognitive Structure of Emotions (1988) • Emotions characterized by abstract evaluative criteria applied to: – Events (desirable / not…) – Acts by other agents (praiseworthy / not…) – Objects (attractive / not…) • ~22 emotions defined (including complex/social) Hudlicka EmoSPACE2011 13
- 14. Valenced Reactions Event-based emotions Attribution Attraction Desirability = low emotions emotions Praiseworthiness = low Fortunes-of-others Fortunes-of-self emotions emotions happy for, pity, degree of autonomy = high gloating.. expectation deviation = high Prospect-based Well-being emotions emotions distress reproach fear love,hate Hudlicka anger EmoSPACE2011 14
- 15. Outline • Definition • Three Theoretical Perspectives • Similarities & Differences • Implications for Affective Modeling • Summary & Conclusions Hudlicka EmoSPACE2011 15
- 16. Semantic Primitives: Atomic Structure of Emotions Discrete / Dimensional Componential Categorical PA or PAD ~6 Basic dims. Appraisal emotions variables (13) OCC evaluation criteria (11) Etc. Becker-Asano, 2005 Hudlicka EmoSPACE2011 16
- 17. (Size of) Affective Spaces small # of affective states accommodated large Hudlicka EmoSPACE2011 17
- 18. Which Emotions Can be Defined by the Semantic Primitives? • Discrete / Categorical – Basic emotions (joy, sadness, fear, anger, disgust, surprise) – Complex & social emotions? - Theoretical basis lacking • Dimensional – Larger set than basic - but not all emotions uniquely defined – Lacks cognitive differentiation • Componential (cognitive appraisal variables) – Very large space, accommodates many types of affective states & varying intensities Hudlicka EmoSPACE2011 18
- 19. What Do the Different Semantic Primitives Characterize? Basic emotions Categorical / Discrete World & Relationship of Subjective Felt Self w/ World Experience Componential Dimensional Appraisal variables PA or PAD / OCC criteria dims. 19 Hudlicka EmoSPACE2011
- 20. Emotion Modalities Emphasized • Discrete / Categorical – Multimodal patterns (cognitive, motivational, behavioral,subjective) characterizing distinct basic emotions • Dimensional – Subjective felt experience - physiology & subjective experience >> cognition – Characterize ‘core affect’ (Russell) • Componential (appraisal) – Interpretation of stimuli & the stimuli-agent relationship - cognitive & interpretive structure >> physiology – Characterize emotions proper Hudlicka EmoSPACE2011 20
- 21. Outline • Definition • Three Theoretical Perspectives • Similarities & Differences • Implications for Affective Modeling • Summary & Conclusions Hudlicka EmoSPACE2011 21
- 22. (Some) Implications for Affective Modeling • Modeling core affective processes – Modeling emotion generation – Modeling emotion effects Hudlicka EmoSPACE2011 22
- 23. Domain EMOTION GENERATION MODELS Stimuli Emotions Discrete / Categorical Dimensional + -D P=x P A=y D=z = + + A Componential (appraisal) Appraisal Variables Novelty =x Valence =y = Goal Congr.=z Agency =w Etc. OCC (appraisal) Evaluative criteria Desirability Praiseworthiness Attractiveness … etc. Hudlicka EmoSPACE2011 23
- 24. Modeling Emotion Generation • Emphasis on cognitive modality - cognitive appraisal (other modalities ignored / minimized) • Appraisal theories (componential perspective) provide best theoretical support • Emotions defined in terms of abstract (domain- independent) features: – Vectors of appraisal variables – Vectors of OCC evaluation criteria • Identifying values of these features may be non trivial Hudlicka EmoSPACE2011 24
- 25. EMOTION EFFECTS MODELS Multimodal Emotions Manifestations of Emotions Discrete / Categorical Dimensional physiology expressions P=x & gestures A=y D=z speech cognition Componential Appraisal Variables Novelty =x Valence =y Goal Congr.=z Agency =w Etc. Hudlicka EmoSPACE2011 action 25
- 26. Modeling Emotion Effects • More challenging > emotion generation: – Cannot easily ignore multiple modalities – Effects on behavior, expression, (less) cognition • Theoretical support << emotion generation • PAD representation facilitates modeling of expressive manifestations • Supporting data for mapping PAD & appraisal variables onto effects not always available – More data available at discrete / categorical level Hudlicka EmoSPACE2011 26
- 27. Benefits of Dimensional Representations • Model large # of emotions • Facilitate integration of multiple emotions … but theory still lacking • Continuous representations facilitate smooth transitions among emotions & different intensities – …more realistic affective expression • Enable parsimonious representation of shared qualities of different emotions – …high arousal --> rapid movement, high speech pitch Hudlicka EmoSPACE2011 27
- 28. Which Perspective & Space Should You Choose? • ….it depends • Which emotions will you need? • Which affective processes will you model? – Appraisal variables good for emotion generation modeling (via cognitive appraisal) – PAD dimensions good for dynamics of emotion expression –… • Are the data available? – For each dimension / variable? – For each modality of interest? Hudlicka EmoSPACE2011 28
- 29. Outline • Definition • Three Theoretical Perspectives • Similarities & Differences • Implications for Affective Modeling • Summary & Conclusions Hudlicka EmoSPACE2011 29
- 30. Summary of Continuous Characterizations of Emotions • Different spaces defined by different underlying dimensions (semantic primitives) – PAD vs. appraisal variables (& OCC evaluative criteria) • Emphasizing different aspects & modalities of emotions – Felt experience vs. stimulus <--> agent relationship – Physiology vs. cognition • Different pros/cons for particular affective computing tasks • “the search for the optimal low-dimensional representation of the emotion domain remains open” (Fontaine, Scherer, Roesch, Ellsworth, 2007) Hudlicka EmoSPACE2011 30
- 31. Conclusions • Different perspectives / spaces should be viewed as complementary – Decision re: perspective choice guided by specific research & applied objectives – Multiple perspectives may be used in a single architecture …supporting different tasks • It’s too early for a “unified theory” of emotions – Unified theory of emotions may not exist – “Emotion is too broad a class of events to be a single scientific category, and no one structure suffices.” (Russell & Feldman Barrett, 1999) Hudlicka EmoSPACE2011 31
- 32. Thank you Questions? Hudlicka EmoSPACE2011 32