GMU
Volgenau School of IT&E
CS Department
AIT Department

The Learning Agents Center conducts fundamental and experimental research on the development of learning agents for real-world problems, and supports teaching in the areas of intelligent agents, machine learning, knowledge acquisition, and artificial intelligence.

Major research areas include instructable agents, development of knowledge bases and knowledge-based agents, multistrategy learning and knowledge acquisition, domain modeling, knowledge representation, intelligent tutoring systems, natural language processing and cooperative problem solving.

Research vision: agents taught by normal users

The long term objective of our research is to develop artificial intelligence methods that will change the way Intelligent Agents are built, from “being programmed” by a knowledge engineer to “being taught” by a user that does not have prior knowledge engineering experience. These methods should allow a normal computer user, that is not a trained knowledge engineer, to build by himself or herself an intelligent assistant as easily as he or she now uses a word processor.

We believe that this research will contribute to a new revolution in the use of computers, probably even more important than the creation of personal computers. Indeed, it will allow every person to no longer be only a user of programs developed by others, but also a program/agent developer himself or herself.

General characterization of our research approach

Our original approach to developing agents by non-programmers, called Disciple, relies on building a series of increasingly more capable learning and reasoning agents that can be taught to solve problems in an application domain by a user that is an expert in that domain, but does not have knowledge engineering or computer experience. The agent will learn from the expert, developing its knowledge base that consists of an object ontology that defines the terms from the application domain, and a set of general problem solving rules expressed with these terms. This process includes importing ontological knowledge from existing repositories of knowledge, and teaching the agent how to perform various tasks, in a way that resembles how the expert would teach a human apprentice. This is a mixed-initiative process, premised upon a division of responsibility between the expert and the agent where each is accorded responsibility for those elements of knowledge engineering for which they have the most aptitude, and together they form a complete team for knowledge base development. The approach is based on several levels of synergism between the expert that has the knowledge to be formalized and the agent that is able to formalize it. At the highest level there is the synergism in solving complex problems, where the agent contributes routine and innovative problem solving steps and the expert contributes inventive and creative ones. At the next level down, there is the synergism between teaching and learning, where the expert helps the agent to understand the problem solving steps contributed by him or her, and the agent learns general problem solving rules that will allow it to apply similar steps in future problem solving situations. Finally, at the lowest level, there is the synergism between different learning strategies employed by the agent to learn from the expert in situations in which no single strategy learning method would be sufficient. In this way, the agent learns continuously from the expert, building, refining, verifying and improving its knowledge base.