Modeling emotion dynamics in intelligent agents

Abstract

Emotions were shown to have a leading role in the human decision-making process, and thus they play an important role in human intelligence. Intelligent agents' research produced many models of emotional agents. However, most of these models focused on agent's behavior rather than its internal states, and thus, these agents tend to produce emotions according to some rules or predefined domain knowledge. Throughout the history of research on emotions, memory and experience were emphasized to be the main sources of the beating heart of the emotional process. More recently, the term "emotional intelligence" was introduced. Its introduction further emphasized the importance of experience in the emotional process. We are proposing a model of an agent that produces emotions according to its experience. Moreover, it simulates the emotional intelligence process. Research on human psychology has long considered the notion of an emotion to be a matter of degree; however, most of the existing research on emotional intelligent agents treats emotion as a black-and-white matter. We used fuzzy rules to explore the capability of fuzzy logic in modeling emotions. Fuzzy logic helped us to capture the fuzzy and complex nature of emotions. Experience and learning also helped in evolving our model to create a better illusion of reality. We chose to simulate a PET rather than a human or a creature, because pets have a limited set of goals that can be defined in a simple structure. Humans are more complex than pets. The emotional process in human beings is not a clear-cut process by itself, but rather it is intermingled with many other processes in the brain, including planning, self-perception, etc. Therefore, in order to simulate a human, one will have to simulate the other processes as well, while in pets it was much simpler. However, we will argue that our model could also be used to simulate human emotions.