Disciple-RKF/COG: agent teaching by subject matter experts

We are addressing the knowledge acquisition bottleneck in the development of knowledge-based systems by elaborating the Disciple theory and methodology that enables subject matter experts to build such systems by themselves, with limited assistance from knowledge engineers (Tecuci 1998). The investigated solution consists of developing a very capable learning agent shell that can perform many of the functions of a knowledge engineer. As an expert system shell, the learning agent shell includes a general problem solving engine that can be reused for multiple applications. In addition, it includes a multistrategy learning engine for building its knowledge base (KB) which has two main components: an object ontology that defines the concepts from a specific application domain, and a set of task reduction rules expressed with these concepts. The subject matter expert and the agent engage into a mixed-initiative reasoning process during which the expert is teaching the agent his problem solving expertise, and the agent learns from the expert, building, verifying, and improving its KB. Over the years we have developed a series of increasingly more capable learning agent shells from the Disciple family. The most recent family member, DiscipleRKF/COG, represents a significant advancement over its predecessors. It implements a more powerful plausible version space representation that allows all the types of knowledge from the KB (not only the rules, but also the objects and the tasks) to be learned with similar methods. Moreover, the partially learned knowledge pieces are represented at several levels of formalization, from natural language to formal logic, facilitating expert-agent communication, mixed-initiative problem solving, and learning. As a consequence, Disciple-RKF/COG incorporates new tools that allow a subject matter expert to perform additional knowledge engineering tasks, such as scenario specification, modeling of his problem solving process, and task formalization. Disciple-RKF/COG was used and evaluated in several courses at the US Army War College, with very promising results, being made part of their regular syllabi.