Modeling of emotion elicitation conditions for a cognitive-emotive architecture

Abstract To cater to the need of embodying emotional behavior in an autonomous agent, there is a need for modeling computationally apt definitions of emotions. A number of emotion theories have been developed that provide an understanding of human psychology and their emotional behaviors, but it is difficult to directly decipher a theory into a computational model of emotion. Nevertheless, these theories together can serve as the theoretical foundation for designing a model for emotion-eliciting conditions. In this study, the salient features of OCC, Scherer, and Roseman theories of emotions are identified, which complement each other. The features are unified and standardized to bring consistency in deriving the computationally apt definition of five emotions viz. Happiness/Joy, Sadness, Fear, Anger, and Surprise. The objective of this hybridization is to set a ground framework for appraising the emotion-triggering cues (e.g., an event) for a simple, flexible and tolerant computational model of emotions. The underlying emotion-eliciting processes are designed using Fuzzy Logic. Fuzzy rules are framed to model the conditions behind emotion elicitation. Furthermore, the ISEAR data set and the real test-case scenarios are used to validate the accuracy of emotion prediction and rule fulfillment respectively.