Emotions in man and machine

Emotions in man and machine
An exploration of an emerging field Some questions
History of Emotions
Many philosophers talked about emotion but emotions traditionally were thought: – Considered as “passions” (thus passive), – Examined mainly in the context of ethics.
What do we mean by emotion ?
You leave your home in a harry take your car and start driving in heavy traffic You hit the brakes, your face muscles contract, you blow the horn.
Components of emotion
You leave your home in a harry take your Environment
C gnition
Physiology
Behavior
What are emotions ?
How can one discriminate between emotions and other affective phenomena ? An emotion should be …
Relevant to major concerns of the organism What is not an emotion
Preferences (likes and dislikes)Attitudes (including love)Mo Affect dispositions (nervous, anxious, irritable, reckless, hostile, envious, jealous)Interpersonal stances (polite, distant, cold, warm, supportive)Aesthetic emotions ? Discriminating between emotions
Feeling Theories
– "we feel sorry because we cry, angry because we
that we cry, strike, or tremble, because we are sorry, angry, or fearful, as the case may be” Damasio, Somatic Markers Hypothesis, Perceptual Theories Appraisal Theories
Transformation of an event to an evaluated event is
often referred as appraisal.
Pleasantness, goal significance, coping potential, compatibility with goals Relevance, context, urgency, difficulty, action readiness Cognitive Theories
Emotions are complexes of beliefs, desires and feelings Social Constructivist Theories
Emotions are complexes of feelings and attitudes which like other experiences and behaviours are social products. Love as an altruistic sentiment can only be found in post renaissance (or medieval) texts and not ancient texts. Emotions like anger differ in western (individualistic) vs east cultures (interdependent) Not everyone is capable of experiencing each emotion. The evolutionary perspective
Evolutionary psychology sees the brain as a zoo of programs that were evolved to solve domain specific problems (foraging, mate choice, face recognition). This programs have sometime conflicting goals (e.g. flight and sleep). Emotions are therefore higher level programs that orchestrate these sub-programs e.g. switch on-off some programs when others are activated or modulate their parameters. Neuroscience Perspective
No single “emotion” circuit. Several circuits that med Three levels of circuits: reflexive (lower-brain), mid (mamalian) brain, high-brain (neo-cortex) Basic circuits: seeking/exploration, rage/anger, fear/anxiety, lust/sexuality, care/nurturing, panic/separation, play/joy Emotions are a part of a management system to coordinate each individual’s mu m ltiple plans and goals under constraints o Physiology of emotion
Emotions may “express” themselves through: Facial Expressions
Measured by high speed filming and electromyographyAt least some discrete emotions are associated wi Induced states in which individuals report positive and negative emotions are associated with distinctive facial actionsdisplays similar to those in adults can be found in neonates and congenitally blind, suggesting that these expressions are inherently linked with biological mechanisms.
Many emotional processes are not accompanied with visually perceptible facial expressions. May vary from one individual to an other. Expression can be used to (fake) hide or mask the actual emotion.
Contestants of the Eckman basic emotion theory say that the facial expressions reflect a communicative aspect and not the actual feeling. Vocal Expressions
– Temporal Measures (duration of words, pauses, – Combined Time-Frequency-Energy Measures: Complicated spectrum analysis techniques.
Acoustic correlates of emotion
Mainly indicated correlation with arousal High stress: >F0, >intensity, <utterance duration (faster speech) Anger: >F0, >intensity, increase in high frequency energy Sadness: < intensity, < F0, slower speech Human listeners accuracy is about 60% (actors). Autonomic Nervous System
Heart rate: > anger, fear and sadness. Anger, fear > happiness, fear > sadness Discrete emotions may not be differentiated by visceral activity alone. Negative emotions may be characterized by greater ANS activation Emotion machines
Do machines need emotions ?
“The release of emotion is what keeps us health. Em "That may be, Doctor. However, I have noted that the healthy release of emotion is frequently unhealthy for those closest to you." McCoy and Spock, "Plato's Stepchildren", stardate 5784.3 They need to have some sort of emotional intelligence to interact with humans Emotional Intelligence
The Affective Computing
Research Program
Machines can measure emotion/affect by observing
behavior, physiology
“Natural” HCI (Embodied conversational agents) Social sciences (psychology, psychiatry, neuroscience) Affective Computing and Perception of
Affect
Signal processing for measurement of behavioural/physiological expressions of emotion.
Psychophysiological signals (BP, CR, PPG, EEG) Classic pattern recognition (feature extraction, dimensional reduction, classification) A Survey of Affect Recognition Methods: Audio, Visual, and Spontaneous Expressions Zhihong Zeng, Member, IEEE, Maja Pantic, Senior Member, IEEE, Glenn I. Roisman and Thomas S. Huang, Fellow, IEEE How good they are ?
Maybe as good as humans (under controlled conditions)Wha Open areas in affect perception
Uncontrolled environmentsSubject independenceDyn Novel cues (posture, head movements, gaze, body,
para-linguistic etc.)
Multimodality
Use of context
Subtle, Spontaneous
Limitations due to behaviorist premises
Embodied Conversational Agents
Successful HM communication requires that thecommunication partner is attributed with a form ofpersonality (Cassell)Believable human like physical appearance and behavior (facial expressions, gestures etc.)Huge progress in computer graphics research.
But, basically “emotionalizing” virtual agents or robots by “tweaks and tricks”.
Recently, synthesis (and analysis) of emotions is driven by computational cognitive models. Cognitive Models of Emotion:
The OCC Model
Ortony, A., Clore, G., Collins, A., eds.: The Cognitive Structure of Emotions … some successful demonstrations
– Medical guidelines, patient education – Virtual training, therapy (progress evaluation) … and many open problems
“Internal” effect of emotions on cognitive appraisal Multiple emotions (combine and integrate) Emotion in
Large-Scale Architectures for
Human-level Intelligence
SOAR and ACT-R extensions include emotions to modulate learning “emotional states are not especially different from the processes that we call "thinking"; instead, emotions are certain ways to think that we use to increase our resourcefulness” Reactive, deliberate (“what-if”), reflective processes Emotions as a short if interrupt mechanism Conclusion
Emotion research: an emerging field in cognitive science understanding: richness, ambiguity, multi-modality Computer scientists should take a step back and see the big picture.
Real world applications should show the way.

Source: http://www.iti.gr/iti/files/document/seminars/Emotions_in_man_and_machine.pdf