Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Scalable Computational Models of Emotion for Virtual.

Slides:



Advertisements
Similar presentations
Affective Facial Expressions Facilitate Robot Learning Joost Broekens Pascal Haazebroek LIACS, Leiden University, The Netherlands.
Advertisements

Computational Aspects of Emotion in Adaptive Behavior Joost Broekens, Walter Kosters, Fons Verbeek LIACS, Leiden University, The Netherlands.
Module 16 Emotion.
Consistent Dynamic-group Emotions for Virtual Agents. Abstract The use of computational models of emotion in virtual agents enhances the realism of these.
Emotion Psychology, 4/e by Saul Kassin CHAPTER 12: Emotion 4/12/2017
Joost Broekens, Doug DeGroot, LIACS, University of Leiden, The Netherlands Emergent Representations and Reasoning in Adaptive Agents Joost Broekens, Doug.
general psychology Firouz meroei milan Emotion 1.
Measuring emotions in an intergroup context and beyond Dr Roger Giner-Sorolla Department of Psychology Presentation given at the Research Methods Festival,
(How) Can Appraisal Theory be Formalized at a Meta-level? Joost Broekens, Doug DeGroot LIACS, Leiden University.
Module 11. Emotions  Whole-organism responses, involving: Physiological arousal Expressive behaviors Conscious experience.
Emotion. Definition of Emotion An evaluative response involving physiological arousal, subjective experience, and behavioral expression.
Modeling Emotion Frameworks Useful In Computation and Speech Frank Enos.
Creating Architectural Descriptions. Outline Standardizing architectural descriptions: The IEEE has published, “Recommended Practice for Architectural.
Affective Computing and Human Robot Interaction a short introduction a short introduction Joost Broekens Telematica Institute, Enschede, LIACS, Leiden.
Emotion: More Than a Feeling
The role of Confidence Factor in “Humanizing” the decision making of an AI Agent Syed Enam-ur-Rehman1 Mohammed Zeeshan Ozair2 1 Department of Computer.
Introduction to Affect and Cognition Copyright © 2005 by The McGraw-Hill Companies, Inc. All rights reserved. McGraw-Hill/Irwin Chapter 3.
1 IUT de Montreuil Université Paris 8 Emotion in Interaction: Embodied Conversational Agents Catherine Pelachaud.
February 6 1 EQ- How do we experience emotion? Agenda: 1.Daily Sheet 2.Review Motivation Quiz 3.Emotions Introduction 4.Basketball game? Table of Contents:
COMPUTATIONAL MODELING OF INTEGRATED COGNITION AND EMOTION Bob MarinierUniversity of Michigan.
1 From individuals to social and vice-versa André Campos Utrecht University, The Netherlands Universidade Federal do Rio Grande do Norte, Brazil Frank.
Computational Models of Emotion and Cognition Computational Models of Emotion and cognition Christopher L. Dancy, Frank E. Ritter, Keith Berry Jerry Lin,
Module 16 Emotions Kimberly, Diana, Kristen, JP, Chris, Michael, Chris.
Emotion Module 12. Emotions Whole-organism responses, involving: –Physiological arousal –Expressive behaviors –Conscious experience.
Emotion Module 12. What are emotions? full body responses, involving: 1. physiological arousal (increased heart rate) 2. expressive behaviors (smiling,
Chapter 8: Motivation and Emotion
MOTIVATION & EMOTION. HUNGRY? What motivates you to eat? Is it physiological (physical) factors or psychological (mental) factors? Could it be a combination.
Emotion. It is a big concept, not easy to define. Drever (1964) Emotion involves ‘bodily changes of a widespread character- in breathing, pulse, gland.
An Architecture for Empathic Agents. Abstract Architecture Planning + Coping Deliberated Actions Agent in the World Body Speech Facial expressions Effectors.
Cognitive Reasoning to Respond Affectively to the Student Patrícia A. Jaques Magda Bercht Rosa M. Vicari UNIVERSIDADE FEDERAL DO RIO GRANDE DO SUL BRASIL.
Computational Investigations of the Regulative Role of Pleasure in Adaptive Behavior Action-Selection Biased by Pleasure-Regulated Simulated Interaction.
THEORIES OF EMOTION. EMOTION is a set of complex reactions to stimuli involving subjective feelings, physiological arousal, and observable behavior.
IST Contribution lisbon Mind Races meeting, September 2005.
The human 3 of 3 U2Mvo&feature=player_embedded the human 3 of 31.
Human Learning Asma Marghalani.
V I B RAIN T: Bridging The Gap Between Higher-Level Cognitive Functions And System- Level Brain Structures. Abstract The Visual Brain Tool (V I B RAIN.
Bob Marinier Advisor: John Laird Functional Contributions of Emotion to Artificial Intelligence.
Achieving Believable Psychosocial Behaviour in Non-player Characters in Modern Video Games Christine Bailey, Jiaming You, Gavan Acton, Adam Rankin, and.
Chapter 2.2 Game Design. CS Overview This introduction covers: –Terms –Concepts –Approach All from a workaday viewpoint.
Lesson 12 - Communication
Emotion. Emotion  It is at the heart of who we are as people  It often is a reflection of our mental state  We are going to look at different theories.
Motivation and Emotion. Motivation Motivation - process by which activities are directed so that physical or psychological needs/wants are met. Extrinsic.
Mindful Shift Chapter 12. “Of all species on earth, we human have the capacity of mind change: we change our minds and that of others”
University of Windsor School of Computer Science Topics in Artificial Intelligence Fall 2008 Sept 11, 2008.
1 Lesson 4 Attitudes. 2 Lesson Outline   Last class, the self and its presentation  What are attitudes?  Where do attitudes come from  How are they.
Warm UP Identify the following topics in your own words
Emotion An introduction and approaches to understanding T. Curwen.
Module 16 Emotion.
Distributed Models for Decision Support Jose Cuena & Sascha Ossowski Pesented by: Gal Moshitch & Rica Gonen.
Motivation and Emotion. Motivation Motivation - process by which activities are directed so that physical or psychological needs/wants are met. Extrinsic.
Emotion Theories.
Emotional Intelligence
© 2011 The McGraw-Hill Companies, Inc. Instructor name Class Title, Term/Semester, Year Institution Introductory Psychology Concepts Emotions.
Copyright © Allyn & Bacon 2007 Emotion. What Is Emotion? Emotion – A four-part process consisting of physiological arousal, cognitive interpretation subjective.
Theories of Emotions Module 41.
Regulation of Emotion. Name the emotion Contempt Surprise Anger Happiness Disgust Sadness Fear.
EMOTION BY: JORDAN, MATT, DOUG, AND JORDAN. WHAT IS EMOTION? Emotion- a natural instinctive state of mind deriving from one's circumstances, mood, or.
ORGANIZATIONAL BEHAVIOR
Thinking About Psychology: The Science of Mind and Behavior Charles T. Blair-Broeker Randal M. Ernst.
The problem. Psychologically plausible ways of
Motivation, Ability, and Opportunity
Sport and Exercise Psychology
Motivation, Ability, and Opportunity
Measuring Social Life: How Many? How Much? What Type?
Chapter 5 Motivation and Emotion: Driving Consumer Behavior
Chapter 13 Emotion pt. 1: Facial Expressions and Theories of Emotions
MOTIVATION & EMOTION.
Computational Aspects of Emotion in Adaptive Behavior
Module 16 Emotion.
Presentation transcript:

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Scalable Computational Models of Emotion for Virtual Characters Joost Broekens, Doug DeGroot {broekens, LIACS, Leiden University, The Netherlands.

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Contents What’s an emotion? Psychological theories of emotion. Appraisal Theory and BDI Agents. Why use emotions in agents? Problem definition. FeelMe systems Context Sensitive Appraisal Banks Experiment Conclusion

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands What’s an Emotion Common emotions: fearful, angry, happy, sad, surprised, disgusted Short episode triggered by an (internal/external) event composed of: –subjective feelings –inclinations to act –facial expressions –cognitive evaluation, –(and some other things) Heuristic relating events to goals, needs, desires, beliefs of an agent. –Evaluates personal relevance and helps decision-making (Neurological evidence: Damasio) Communication medium. –Communicate internal state (Sociological evidence: Darwin, Ekman)

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands How is an emotion produced? James-Lange –Emotion results from the evaluation of the bodily reactions that are provoked by events. Schacter-Singer (Two Factor Theory) –Emotion results from the cognitive evaluation labeling the arousal of the organism. Arousal results directly from events. StimulusResponseFeedback StimulusArousal EmotionEvaluation Emotion

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Cognitive Appraisal Theory and BDI Agents Appraisal Theory (Frijda, Lazarus, Scherer, etc): –Emotion: is a result of the evaluation of the environment in relation to the agent’s goals, needs, beliefs and desires (=appraisal). –Evaluation in terms of appraisal dimensions: variables expressing a certain emotional aspect of a situation, e.g. valence or arousal. –Appraisal Assumption: evaluation is both necessary and sufficient for an emotion to occur. BDI based agents: –Agent’s thinking based on beliefs, desires and intentions. –Possess basics to which appraisal based emotions can be added. PerceptionAppraisalEmotion Beliefs/Desire/Goals/Etc.

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Why Use Emotions in Agents Virtual Agents (NPCs, Tutor agents) are enriched with emotions for, e.g., the following reasons: –Enhance sense of realism (VR Training). –Entertainment (Games) –Enhance communication between agent/robot and human (HCI). Examples: SIMS2, Mission Rehearsal Exercise (Marsella and Gratch), Kismet (Braezeal).

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Computational Models of Emotion in Virtual Agents Mostly Cognitive Appraisal Theory Based –Appraisal assumptions “evaluation in terms needs and goals of the agent is necessary and sufficient” makes Cognitive Appraisal Theory suitable for Virtual Agents based on BDI architecture. Integrated into BDI structure / Architecture of the agent. –Because emotions result from cognitive evaluation, computational models of emotion are integrated in BDI architecture. –Not built as extendible module or add-on but often deeply integrated in the BDI architecture.

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Problem How to incrementally add sophistication to a computational model of emotion, while keeping it consistent with the emotions produced by the simpler version of the model? Why model-scalability: –Different mechanisms might give good emotion results in different situations (e.g. event-based emotion encoding / appraisal-based emotion). –Development benefits (selling upgrades/ debugging and evaluation). How to make scalable computational models of emotion? Why runtime-scalability: –Resources in games are limited, but very variable (high-end/low-end PCs). –Ideally users can trade-off the amount of emotional detail for e.g. frame- rate (analogous with e.g. Graphical detail). –Other situations,e.g. viewing distance.

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands FeelMe system, a Dynamic Modular Approach Used FeelMe system (DeGroot) to work on this problem. FeelMe system (limited overview). DSS: Decision Support System AS: Appraisal System, emotionally evaluates the environment EMS: Emotion Maintenance System, maintains emotional state BMS: Emotional expression AS EMS BMS DSS Agent's environment/Agent's internal state PerceptionAppraisalEmotion

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Appraisal System / Emotion Maintenance System. AS continuously emotionally evaluates the situation as constructed by the Decision Support System. AS interprets the situation in terms of appraisal dimension values. AS sends a stream of n-dimensional vectors of these values to the EMS. Such a vector is called an appraisal result. EMS maintains the emotional state as a point in the n-dimensional space of appraisal dimensions. (e.g. a point in the pleasure, novelty space) EMS assumes appraisal results are changes, or “deltas” to the emotional state. EMS integrates these appraisal results with the existing emotional state. –EMS is “pushed” towards a direction. I.E: Signal-based approach AS EMS BMS DSS Agent's environment/Agent's internal state

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Multiple “Appraisal Banks” What if multiple independent appraisal subsystems could send appraisal results? –Appraisal results can be produced by multiple appraisal mechanism, including event encoding (e.g. events have a fixed emotional meaning). –Modular and scalable approach to appraisal? Context Sensitive Appraisal Banks (i.e. appraisal modules) How to integrate the result of different concurrent “appraisal banks”? AS ASM EMS BMS DSS Agent's environment/Agent's internal state AS EMS BMS DSS Bank 1 Bank n Agent's environment/Agent's internal state

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Constraints for the Integration of Results from Appraisal Banks EMS integrates appraisal results: –Appraisal-results defined at interval scale –Banks together must produce non-zero positive and negative values. Context Sensitivity in Appraisal Banks –Sensitive to mutually exclusive situations. Either predefined (based on context), or Dynamic (based on appraisal of the current situation of the agent). –Appraising on different detail level: One bank: big picture Second bank: details AS EMS BMS DSS Bank 1 Bank n Agent's environment/Agent's internal state

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Experiment: PacMan To find out if such integration of appraisal bank information is feasible using the signal-based approach FeelMe system, and To find out if this integration permits the use of these banks as separate appraisal modules that add value to each other. –Allowing incremental design  model-scalability –Dynamic appraisal adaptation during run-time  run-time scalability PacMan (Chow) simulation: compare the traces of the resulting emotional state while playing one level of PacMan (controlled by the human) of an Appraisal Systems with 1, or with 2 appraisal banks. AS EMS Bank 1 AS EMS Bank 1 Bank 2

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands PacMan’s Appraisal System Appraisal Banks relate to PacMan’s following goals: –Points: all events related to gathering points. –Survival: all events related to survival. The output of Appraisal Banks is dependent on the appraisal intensity of other banks but is independent from the appraisal mechanism of other banks. –Survival bank inhibits the effect of points bank. –Rational: survival more basic and important than points –Example of mutual exclusive banks, dynamically based on the current situation as evaluated by the appraisal banks.. AS EMS: Integrate Survival Points Inhibition metric

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands PacMan’s Emotional State Based on three appraisal dimensions (Mehrabian): –Pleasure: related to goal congruency. –Arousal: related to novelty and attention needed for the event. –Dominance: related to the influence PacMan has on its environment, i.e. PacMan’s power. EMS simply integrates the appraisal results from the Appraisal Banks

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Results Emotional state (emotion) behaves more sophisticated to the conditions of the environment in the 2-bank (survival, points) configuration (e.g. eating a ghost is positive in the 2-bank case). Emotional state (emotion) is still consistent with the emotional state of those situations in which the 1-bank (survival) produced emotions that were meaningful.

Joost Broekens, Doug DeGroot, {broekens, LIACS, Leiden University, The Netherlands Conclusions Context Sensitive Appraisal Banks in a dynamic, signal-based approach facilitate model-scalability (i.e. the ability to incrementally add sophistication to a computational model of emotions while staying consistent with the earlier version of that model) –If: appraisal banks are mutually exclusive either pre-programmed or dynamic (experiment), and –If results from appraisal appraisal banks can be integrated in a meaningful way. Runtime-scalability can be achieved using the appraisal bank setup. –Switching the “point” appraisal bank on/off would only add/remove emotional sophistication, and not result in inconsistent emotional state behavior. –Triggering the emotion system at a (slightly) different rates did not introduce differences in the resulting emotion (see paper).