A Mathematical Theory of Knowledge for Intelligent Agents

Knowledge is a property that measures the degree of awareness of an agent about a target in an environment. The goal in conventional intelligent and cognitive agent development is to build agents that can be trained to gain knowledge about a target. The definition and operations of this knowledge associated to the agent is not clear, whereas these are required for developing a reliable, scaleable and flexible agent. In this paper, we take into account such requirements, and present a concise theoretical framework for the design of cognitive and rational intelligent agents, their properties and operations on targets. We present many illustrative examples and an experiment to show how the incorporation of different cognitive properties to an agent enables the agent to act rationally with improve generalization performance in an interactive environment.