A Methodology for Modeling and Representing Expert Knowledge that Supports Teaching-Based Intelligent Agent Development

The long term research goal of our research group is to change the way a knowledge-based agent is built, from being programmed by a knowledge engineer (based on what he or she has learned from a domain expert) to being directly taught by a domain expert that receives limited or no support from a knowledge engineer. The investigated approach, called Disciple (Tecuci, 1998), relies on developing a very capable learning and reasoning agent that can collaborate with a domain expert to develop its knowledge base consisting of an ontology that defines the terms from the application domain, and a set of general task reduction rules expressed with these terms. An important component of this research is the development of a general methodology for modeling and representing expert knowledge that supports teachingbased intelligent agent development. With respect to this methodology we formulate the following claims: 1) it is natural for the expert, 2) it is applicable to a wide variety of domains; 3) it identifies the tasks to be represented in the agent’s knowledge base, 4) it identifies the necessary concepts and features to be represented in the ontology, 5) it guides the rule learning process 6) it supports natural language generation of solutions by the agent and their abstract justifications. We will briefly illustrate this methodology with the Course Of Action (COA) challenge problem that consists of rapidly developing a knowledge-based critiquer that receives as input the description of a military course of action and assesses various aspects of the COA, such as its strengths and weaknesses with respect to the Principles of War and the Tenets of Army Operations. The domain expert is given a specific problem to solve (such as, to “Assess COA411 with respect to the Principle of Objective”) and solves it through task reduction. To perform this assessment, the expert needs a certain amount of information about COA411. This information is obtained through a series of questions and answers that help reduce the initial assessment task to simpler and better defined ones, until the expert has enough information to perform the assessment. Thus, to “Assess COA411 with respect to the Principle of Objective”, one has to consider the features that characterize the objective, and these are specified by