A neuroscientific approach to emotion system for intelligent agents

Abstract

For an agent to be believable, emotion is an essential part of the human-computer interaction. In this paper, we propose a dynamic affective system, inspired from neuroscience. The system comprises four modules: an appraisal module, an emotion-generation module, an emotional expression module, and a long-term memory (LTM). The appraisal module gathers environmental stimuli and evaluates them to see if they are rewarding or punishing. Based on the results of the appraisal module and the agent's internal stimuli, the emotion-generation module generates the affective states of the agent. The emotional expression module combines and expresses emotional behaviors in accordance with the affective states. As a result, the proposed affective system can generate various emotions simultaneously and produces emotional life-like expressions. © IFIP International Federation for Information Processing 2006.