Incorporating Rationality in Intelligent Agents

An autonomous intelligent agent functions in interaction with its environment. The agent is capable of altering the state of its environment by acting on it. An essential requirement in any goal-oriented system is that these actions should be rational. In this thesis a design is presented for incorporating rationality into the architecture of the agent. The class of autonomous agents discussed maintain an internal model of their world in a revisable belief system (specifically a Justification-based reason Maintenance System, or JRMS). A theory of rational action is developed and represented symbolically as an independent control structure that filters candidate beliefs . The JRMS is extended to maintain information on the type of a belief, which is used by the control structure to ensure that the agent's actions are consistent with its theory of rational action.