590AC_WK3_Lec_2_Modeling_Emotion_Generation_Part2(2).ppt

590AC_WK3_Lec_2_Modeling_Emotion_Generation_Part2(2).ppt -...

Info icon This preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: CMPSCI 590AC 3-2 Outline • • • • Intro Computational tasks in appraisal Theories of cognitive appraisal Building models of cognitive appraisal CMPSCI 590AC 3-2 Results of the Appraisal Process: “Emotion Object” Structure Type: Valence: Intensity: Variables affecting intensity: Descriptive detail: Duration: Cause: Goals affected: Direction: Coping potential: Other appraisal variables….: fear negative .90 Certainty, Importance of affected goals “aggressive dog approaching” 2 minutes (exp. decay) { aggressive dog | owner} { safety of self | safety of dog | delay } { dog | negligent owner | self } low CMPSCI 590AC 3-2 Factors Influencing “Emotion Object” Structure • How will the emotion be used? – If behavior is influenced, additional factors needed: • Cause (“dog” vs. “dog owner”) • Direction (“self” vs. “dog” vs. “dog owner” vs. “the entire world” vs. “innocent bystander”) • Which theory is being implemented? – If appraisal dimensions are mapped directly onto effects (behavioral & cognitive), need to represent them in ‘emotion object’ • Novelty ---> orienting response • Certainty [anger, happiness] --> heuristic processing CMPSCI 590AC 3-2 Knowledge & Data Required • Depends on: – Emotions and elicitors represented • Degree of domain-dependent data (how much is abstracted?) – Theory implemented & model resolution level • Level of detail of structures & processes – Complexity of emotion object • Type only? Intensity? Dynamics? Causes? Goals threatened?… • What data do we need? – – – – – Stimulus --> emotion mappings Stimulus --> appraisal variable --> emotion mappings Nature of dynamic mental constructs (goals, beliefs) Intensity & ramp-up & decay (emotion dynamics) Emotion ‘integration’ data CMPSCI 590AC 3-2 Knowledge & Data Required • Source of data – Literature & existing empirical studies & data (self-reports) – Concurrent empirical studies – Knowledge elicitation interviews • Types of data – – – – – – – Self-reports Observations Performance measurement for output & process measures … Neuromimaging (fMRI, PET scans) ?? Single-cell and cell-cluster recordings ?? EEG ??? CMPSCI 590AC 3-2 Problems With Obtaining Data • Assessing internal cognitive and affective states – Very difficult… impossible? • Transient, idiosyncratic, inaccessible via existing methods • Problems with self-report data – Not all data can be articulated – Not all articulated data reflect reality • Problems with physiological & behavioral correlates – Can we reliably assess emotional states? – Can we reliably assess internal cognitive states? • Situation assessments & beliefs • Expectations • Goals • Issues of privacy / ethics CMPSCI 590AC 3-2 Modeling Individual Computational Tasks - Stimuli-to-emotion mappings • Nature of the mapping process • Stages & function / stage or parallelism • Degree of variability • Intensity calculation • Emotion dynamics over time • Integrating multiple emotions • Within same time frame • …over time CMPSCI 590AC 3-2 Stimuli-to-Emotion Mappings • Nature of mapping – – – Domain stimuli --> Emotion Domain stimuli --> Appraisal variables --> Emotion Domain stimuli --> Dimensions --> Emotion • Representational requirements for: – Evaluating appraisal variables – Characterizing “emotion object” • “World” & “Self”…. Over time • “Mental constructs” & “Emotion object” • Appraisal structure (appraisal frames) CMPSCI 590AC 3-2 Domain Stimuli --> Appraisal Variables --> Emotion N V GR World Emotions ? GC This is the hard part!!! A C N … This can be difficult too… CMPSCI 590AC 3-2 Domain Stimuli --> Appraisal Variables N V GR World GC This is the hard part!!! A C N … CMPSCI 590AC 3-2 Representational Requirements… for Appraisal Variable Calculation • Novelty: – Compare previous state of [world & self] with new state – Maintain representation of previous state (for comparison) • Unexpectedness – Compare expected state with new state – Need explicit representation of expectations – whatever is not predicted is unexpected • Urgency – Predict future state to determine if need to act now • If ignored, will things get better or worse? – How far into the future? • …temporal representation also needed for… – “Prospect-based” emotions (hope, fear) – AND - predicting possible future states from past states… regret CMPSCI 590AC 3-2 Representational Requirements: Appraisal Variables Calculation • Goal relevance & congruence – Desired state of [the world & the self] • Goals • Goal hierarchies • Causal attribution Get dog treat Find owner Act cute – Causal structures – Plans – Other agents • Are they really responsible? Act cute Owner happy Owner gives treat CMPSCI 590AC 3-2 Representing State of [World & Self]: Mental Constructs goal Desired state situation Current state Match? expectation Expected state CMPSCI 590AC 3-2 Representing State of [World & Self]: Complex Structures Goal hierarchies Plans situation Current state goal Desired state expectation Expected state CMPSCI 590AC 3-2 Representational Requirements • Which aspects of [world & self] to represent? – Richness of mental constructs – Which attributes do mental constructs need? • Representing uncertainty • Criteria for goal matching – How close does a situation have to match the goal? – Does it depend on the context? – Does it depend on any affective factors? • Obsessives may need more exact match before goal satisfied CMPSCI 590AC 3-2 MAMID CMPSCI 590AC 3-2 MAMID Cues: State of the world (“medical task”, “in range”) Cues Attention Situations: Perceived state ( “can process task”) Situation Assessment Expectations: Expected state Expectation Generator (“will succeed”) Goals: Desired state Affect Appraiser Affective state & emotions: Positive valence High happiness Low anxiety (“win game”) Goal Manager Action Selection Actions Actions: to accomplish goals (“process task”) CMPSCI 590AC 3-2 Mental Construct Attributes Task domain attributes Object: State: Value: Originator: Recipient: Comparator: Rank: Time required: Capacity required: meta attributes Threat level: Desirability: Valence: Salience: Confidence: CMPSCI 590AC 3-2 CMPSCI 590AC 3-2 Appraisal Frames • Affective Reasoner (Elliot et al., 1992) • Structures representing appraisals from multiple perspectives – Stimuli (domain specific or appraisal variables) – Goals, values, preferences – Existing moods • Multiple frames represent: – Appraisals from multiple perspectives • Self, other.. – Appraisals over time • Collection of appraisal frames represents dynamic interpretation of an evolving situation – Appraisals of distinct affective factors • Emotions • Feelings • Moods CMPSCI 590AC 3-2 EMA’s “Causal Interpretation” (Gratch & Marsella, 2004) CMPSCI 590AC 3-2 Appraisal Variables --> Emotion N V GR Emotions ? GC A C N … This can be difficult too… CMPSCI 590AC 3-2 GENESE (Geneva Expert System on Emotion): (Scherer, 1993 ) • • • • One of the earliest models of appraisal (1993) Black-box model Maps 15 appraisal dimensions onto 14 emotions Current situation represented in terms of 15 appraisal dimensions – From answers to 15 questions about some situation in the past • Emotions represented in terms of specific values of these variables • Appraisal (= emotion identification) determined by Euclidian distance between the input vector and the emotion • Early tests achieved 73% accuracy in predicting subjects’ emotions CMPSCI 590AC 3-2 Representation & Reasoning Alternatives • Vector spaces (Scherer) • Connectionist (Velasquez) • Symbolic – Rules (Marinier, Jones, Henninger, Hudlicka…) – Belief nets (Hudlicka, de Rosis,…) • Complex symbolic structures (Elliot, Reilly, Gratch & Marsella) – Appraisal frames, causal plan structures • Spreading activation over networks of processes (Breazeal) • Decision-theoretic • • • • • – Decision trees – Decision theoretic formulations (Gratch & Marsella, Lisetti & Gmytrasiewicz) Finite state machines (Kopecek) Markov models (El Nasr) Blackboards and ‘specialists’ (Gratch & Marsella) Theorem proving (Zippora) Dynamical systems CMPSCI 590AC 3-2 Intensity Calculation: Factors Influencing Intensity • Types of stimuli – external; internal – past, present, future • • • Intensity of stimuli Likelihood / certainty of stimuli Types and magnitude of appraisal variables – Importance of goal – Degree of situation-goal congruence – Relative contribution of each variable varies by emotion type • Relationship between actual and expected – Expected events << surprising events (OCC, 1988) – If expected goal succeeds --> lower intensity than if goal was expected to fail (Reilly, 06) • Current emotional state • …..How should they be combined? CMPSCI 590AC 3-2 Intensity Calculations: Examples • Desirability x Likelihood (Gratch et al., EMA) – If likelihood (expectation, certainty) = 1, then Intensity = desirability • Desirability x (change in) Likelihood of success (Reilly, EM) – Asymmetry in desirability & failure (Reilly) • Separate variables needed for – Importance of success (goal desirability) » Joy, hope – Importance of avoiding failure (goal undesirability) » Distress, fear CMPSCI 590AC 3-2 Intensity Calculations: Examples EMA - Gratch & Marsella, 2004 CMPSCI 590AC 3-2 Issues in Intensity Calculation • Are intensity functions same for all emotions? • Asymmetric effects – “Joy from a goal succeeding is not necessarily symmetric with the distress felt when the goal fails” (Reilly, 06) – Desirability of success IS NOT EQUAL to Undesirability of failure – AND - there are individual differences in these reactions • “loss aversion”, sensitivity to negative vs. positive stimuli • Combining intensity of related emotions • Are different intensities required for multiple purposes? (Sonnemans & Frijda) – Felt emotion – Emotion effects on cognition – Emotion activation of behavior CMPSCI 590AC 3-2 • • • • Decay Rates Linear Logarithmic Exponential Some arbitrary monotonically decreasing function over time CMPSCI 590AC 3-2 Issues in Modeling Decay • Not well understood • Differences in decay functions of individual emotions? – Anger appears to decay more slowly • Individual differences – Trait-anxious – “Angry personality” • Differences in existing emotional context – Relates to how emotions are combined CMPSCI 590AC 3-2 Intensity & Decay: Theoretical Status “Appraisal dimensions that allow prediction of duration and intensity may be different from those that allow emotion differentiation.” “The current set of appraisal dimensions may be incomplete…” Scherer, 2001, p. 375 CMPSCI 590AC 3-2 Issues In Combining Multiple Emotions • Synergistic vs. opposing effects • Where are any contradictions resolved? – Appraisal process - does not allow derivation of multiple emotions – Cognition - goal manager selects most important goal – Behavior - only one action is selected • Which combination or conflict resolution functions and algorithms to use? CMPSCI 590AC 3-2 • Combining Intensities of Related Emotions Simple addition – Too intense and unrealistic – A few ‘low intensity’ emotions can result in a ‘high intensity’ reaction • Average – Final result can be less than most intense component - unrealistic? • Max (winner-take-all) – Ignores cumulative effects of multiple emotions • Joy or setbacks… • Logarithmic functions (Reilly, EM) – Linear at low-intensity ranges - good? • Sigmoidal function (Picard, 1997) Combining Opposing or Unrelated Emotions CMPSCI 590AC 3-2 • May not be an issue - depends on appraisal approach • Max – Assume strongest emotion is the most ‘valid’ • Average – May make sense with some emotions (+ / -) – Win lottery but house burns down ---> feel neutral? • Allow parallel effects and behavioral tendencies – Where are any contradictory influences resolved? – Fight and flee at the same time? CMPSCI 590AC 3-2 End of Lecture 3-2 • Questions? Comments? – Email me or post on Discussion board 35 ...
View Full Document

{[ snackBarMessage ]}

What students are saying

  • Left Quote Icon

    As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

    Student Picture

    Kiran Temple University Fox School of Business ‘17, Course Hero Intern

  • Left Quote Icon

    I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

    Student Picture

    Dana University of Pennsylvania ‘17, Course Hero Intern

  • Left Quote Icon

    The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

    Student Picture

    Jill Tulane University ‘16, Course Hero Intern