SOAR WORKSHOP XIV MAY 22, 1999 Towards a model of fear in Soar* Ronald S. Chong Soar Technology, Inc. *This research was funded under Phase 1 of Army STTR 98-T004 awarded to Soar Technologies, Inc. Ronald S. Chong Towards a model of fear in Soar
WHY MODEL EMOTIONS? • More human-like behavior from computer-generated entities. � Increases the efficacy of training • Recent wide acceptance of the intrinsic role emotions play in cognition. � LeDoux, J. E. (1996). The Emotional Brain: The mysterious underpinnings of emotional life . New York, NY: Touchstone (Simon & Schuster). � Ortony, A., Clore, G.L. & Collins, A. (1988). The cognitive structure of emotions . Cambridge University Press. � Damasio, A. (1994). Descartes’ error. New York, NY: Avon Books. � Emotional and Intelligent: The tangled knot of cognition. Papers from the 1998 AAAI Fall Symposium. Technical report FS-98-03. Menlo Park, CA:AAAI Press. Ronald S. Chong Towards a model of fear in Soar
FEAR: SOME PHENOMENA AND OBSERVATIONS • Non-volitional freezing at the appearance of a perceived threat • Perceptual attention is narrowed to the perceived threat • Responses are biased towards those reducing the level of threat • Neurologic structures implicated in fear response Ronald S. Chong Towards a model of fear in Soar
NEUROLOGY OF FEAR (L E DOUX) Sensory cortex “high road” Mildly encoded stimulus features Raw Detailed representation Sensory thalamus Stimulus of the stimulus Data Crude stimulus features “low road” Amygdala Startle Physiological Reflex Changes Freeze Response Ronald S. Chong Towards a model of fear in Soar
COGNITIVE STRUCTURE OF EMOTIONS (ORTONY, CLORE & COLLINS) • “. . . an approach to the study of emotion that explains how people’s perceptions of the world—their construals—cause them to experience emotions.” • “. . . view emotions as valanced reactions to events, agents, or objects, with their particular nature being determined by the way in which the eliciting situation is construed .” • “A person’s appraisal of an emotion-inducing situation is based on three central variables: desirability, praiseworthiness, and appealingness . . .” • Several factors affecting the intensity of emotions � Global variables for all emotions: sense of reality, proximity, unexpectedness, arousal � Local variables for particular groups of emotions: likelihood, effort, realization, etc. Ronald S. Chong Towards a model of fear in Soar
SOMATIC MARKER HYPOTHESIS (DAMASIO) • “When the bad outcome connected with a given response option comes into mind , . . . your experience an unpleasant gut feeling.” • Somatic markers “force attention on the negative [positive] outcome to which a given action may lead, and functions as an automated alarm signal . . . [that] may lead you to reject [accept], immediately , the negative [positive] course of action . . . . and then allows you to choose from among fewer alternative s.” • How are they acquired? “When the choice of option X, which leads to bad outcome Y, is followed by punishment and thus painful body states, [a somatic marker is acquired.] Re-exposure . . . to option X, or thoughts about the outcome of Y, will now have the power to reenact the painful body state and thus serve as an automated reminder of bad consequences to come.” Ronald S. Chong Towards a model of fear in Soar
EMOTION/FEAR MODEL (AS OF 05/19/99 @ 2:33PM) • “Paper” model is inspired by LeDoux, Ortony, Damasio, Sherer and others. • Still very much under development Response “high road” Appraisal Emotion search Features Rational Response Intensity Control All Update Prime Response Perception & Overt Response Stimulus selection attention Features Innate Response Arousal Arousal Inhibition Features “Amygdala” “low road” Startle Physiological Reflex Changes Freeze Response Ronald S. Chong Towards a model of fear in Soar
SCENARIO A • A person is walking through the woods and comes to a large log across the path. He bends down to roll the log out of the way, and while doing so, uncovers and glimpses a long, slender, curvy object. • Assumptions: He is not a accustomed to the Appraisal “high road” sight of snakes, is not Features actively searching for snakes, and does not All Perception & have an expectation of Stimulus attention Features finding snakes. Upon Arousal seeing the object, he Features will initially exhibit a “Amygdala” startle response of a “low road” magnitude proportional to his Startle Physiological personal disposition Reflex Changes Freeze towards and experience Response with snakes. Ronald S. Chong Towards a model of fear in Soar
SCENARIO A, CONT’D • Response 1: He immediately runs away. Appraisal “high road” Features All Response Perception & Stimulus Overt Response selection attention Features Innate Response A Arousal r o u Inhibition s Features a l “Amygdala” “low road” Startle Physiological Reflex Changes Freeze Response Ronald S. Chong Towards a model of fear in Soar
SCENARIO A, CONT’D • Response 2: He freezes, recognizes that the object is a stick, and resumes his walk Response Appraisal “high road” Emotion search s e r u t a Rational Response e Intensity F l o r t n o C All Update Prime Response Perception & Stimulus selection attention Features e s Arousal Arousal n o p s n e Features R o i t e i t b a i n h n n I I “Amygdala” “low road” Startle Physiological Reflex Changes Freeze Response Ronald S. Chong Towards a model of fear in Soar
SCENARIO B • A husband has prepared a late dinner for himself and his wife. Thirty minutes ago, she called to say that she was leaving the office. She has yet to arrive, although her office is a leisurely five-minute walk across campus from home. • Assumptions: The husband cares for his wife and her tardiness is anomalous. Response Appraisal “high road” Emotion search s e r u t a e Rational Response Intensity F l o r t n o C All Update Prime Response Perception & Stimulus Overt Response selection attention Features e s A Arousal n o r p o u s n e s Features R o a i l t e i t b a i n h n n I I “Amygdala” “low road” Startle Physiological Reflex Changes Freeze Response Ronald S. Chong Towards a model of fear in Soar
PROOF-OF-PRINCIPLE DEMONSTRATION • A scenario in TacAir-Soar that would be expected to cause fear in a pilot: 1 v. 4. • The allied fighter, approaching from the west, is flying a CAP mission and will eventually confront the four enemy planes approaching from the east. Ronald S. Chong Towards a model of fear in Soar
PROOF-OF-PRINCIPLE DEMONSTRATION, CONT’D • Built a very simplified fear model • Very simple threat assessment � Am I alone? � Are there enemy present? � Am I/Are we outnumbered? All Response Appraisal Emotion Stimulus search Features Rational Response Intensity Response Overt Response selection Ronald S. Chong Towards a model of fear in Soar
PROOF-OF-PRINCIPLE DEMONSTRATION, CONT’D • In this scenario, the allied fighter will “bug-out” when the threat assessment exceeds a threshold. threat assessment 10 bug-out begins 8 threshold one bogey shot down 6 new bogey appears new bogey appears 4 new bogey appears “I’m alone” 2 new bogey appears time 0 take-off Ronald S. Chong Towards a model of fear in Soar
ISSUES • A knowledge-level implementation • Does not address how fear affects the way we perceive, think or plan • True threat assessment would be based on the following factors � What the likelihood of harm? � What is the expected harm? � How long to I have before I must act? � How effective could my actions be? Ronald S. Chong Towards a model of fear in Soar
FUTURE WORK • GET PHASE 2 FUNDING!! • Complete implementation of fear with a consideration of Ortony’s work • Explore an implementation of arousal • Develop a perceptual attention mechanism • Implement the emotion of anger � How offended are you at someone due to their actions? � How undesirable are the consequences of their actions? • Look for and implement architectural changes • Use a domain relevant to the Army • Get Phase 3 funding Ronald S. Chong Towards a model of fear in Soar
Recommend
More recommend
