Representing human emotions in intelligent agents
The hysteretic computationally-based intelligent agent architecture provides a powerful framework for constructing agents capable of reasoning about and interacting with their environment, including other computationally-based agents. When asked to interact with and attempt to cooperate with human agents, a special problem is encountered. The emotions of the human contribute to reasoning patterns far more complicated than the symbolic logic used by the computational agents. In this thesis a series of systems is presented which attempt to extend the world model of a class of hysteretic intelligent agents to model human as well as computational agents. The class of agents considered uses a revisable belief system (such as a JRMS) to construct and maintain its world model. First, the possibilities of using an unmodified JRMS to represent human emotions are investigated. The later systems attempt to reduce unnecessary complexity by providing a filtering mechanism and by providing default inference rules.