From human emotions to robot emotions

摘要: The main difficulties that researchers face in understanding emotions are only because of the narrowmindedness our views on emotions. We not able to free ourselves from notion necessarily human I will argue if animals have emotions, then so can robots. Studies neuroscience shown animal models, though having limitations, significantly contributed functional and mechanistic aspects suggest one functions is achieve multi-level communication simplified but high impact information. way this function achieved brain depends species, specific emotion considered. classical view ‘computed’ by specialized centers, such as ‘limbic system’, criticized. an ensemble well-known neurobiological phenomena, together referred neuromodulation, provide a useful framework for how arise, maintained, interact with other behavior cognitive processing. This suggests new ways which robot be implemented fulfill their function. There many inherent extremely difficult study account for. start listing few them, it may more fruitful focus rather than what are. do fact at least functionally, even we might empathize them. lead conclusion that, robots could well. briefly 1 New address: Duke University, Biomedical Engineering Department Center Cognitive Neuroscience, P.O. Box 90281, Durham, NC 27708-0281. fellous@duke.edu. author wishes thank Dr. Michael Arbib constructive comments previous versions manuscript. Paper presented AAAI Spring 2004 symposium Architectures Modeling Emotion: Cross-Disciplinary Foundations. E. Hudlicka L. Canamero. Copyright © 2004, American Association Artificial Intelligence (www.aaai.org). All rights reserved. open window neural bases offer thinking about implementing robotemotions. Why study? A difficulty studying here significant individual differences, based experiential well genetic factors (Rolls, 1998; Ortony, 2002; Davidson, 2003a, b; Ortony et al., 2004). My fear sight bear very different experienced park-ranger who has better sense bear-danger knows react. also another had same amount exposure bears, prone risk-taking behaviors. For perceptual fear, commonalities expression, experience underlying neurophysiological mechanisms large enough between individuals, studied models (LeDoux, 1996). However, issue differences overwhelming love or depression. Naturally occurring often arise sequences, time varying intensity, outlast stimuli elicited After Halloween prank, yield anger perpetrators followed shame oneself being angry neighborhood kids Halloween. falling roller coaster ride becomes less intense towards end ride, curves speeds similar. It therefore oversimplification speak emotional ‘state’, intrinsically dynamical phenomena widely constants (from seconds hours days moods, months years depression love). makes since (and cognitive) contexts accounted Laboratory studies try limit control these factors. must question extent controlled settings resemble naturally problem acute if, below, primary roles