Anticipation and Emotions for Goal Directed Agents

Abstract

Breakthrough challenges in the cognitive modeling are in providing artificial agents with abilities to operate in open systems where dynamism, partial knowledge and non-determinism exact agents not only to to quickly react to events but also to anticipate decisions, facing with uncertainty and unpredictability of future events. One of the key issues is then to enhance cognitive reasoning with emotions and affective abilities. The rationale behind the introduction of emotions into artificial agents is manifold. Since the grew up of cognitive sciences and applied psychology during 60s, the negative bias against emotions, and more generally against ‘irrational’ or ‘not reason-based’ responses, has been practically reversed. The functional value of emotions has now been widely acknowledged and their evolutionary contributes have been especially emphasized (Simon, 1967; Fridjda, 1986; Parrott and Schulkin, 1993; Lazarus, 1991). From the perspective of biological evolution, emotions can be considered psychological mechanisms that evolved to solve adaptive problems (Toby and Cosmides, 1990) (i.e. escaping threats or predators, finding food, shelter and protection, finding mates) and thus surviving and delivering one’s genes to one’s own offspring. According to this view, emotions mediate behaviors for organisms in order to enhance long term adaptation and to answer their recurrent ecological demands. Otherwise, emotions provide evolutionary solutions to many of the critical problems implied by agents’ situated interactions with their environments, for instance by enhancing proactiveness, by favoring the adaptive allocation of bounded computational resources, by providing anticipatory mechanisms for adaptation to mutable contexts, etc.