The Disciple Learning Agent Shell and a Disciple Test Generation Agent

The Disciple Learning Agent Shell is a tool for developing intelligent agents where an expert teaches the agent how to perform domain-specific tasks in a way that resembles the way the expert would teach an apprentice, by giving the agent examples and explanations as well as by supervising and correcting its behavior. Disciple can be used by educators to build educational agents that will assist them in various ways. Such an educational agent that was built with Disciple generates history tests for middle school students, assisting the teacher in the assessment of students’ understanding and use of higher-order thinking skills. The agent can also provide intelligent feedback to the student in the form of hints, answer and explanations. The test generation agent and the process of building it are presented in (Tecuci and Keeling, 1998). We propose to demonstrate both the Disciple shell and the test generation agent. 1. Disciple Learning Agent Shell Disciple is an apprenticeship, multistrategy learning approach for developing intelligent agents where an expert teaches the agent how to perform domain-specific tasks in a way that resembles the way the expert would teach an apprentice, by giving the agent examples and explanations as well as by supervising and correcting its behavior. The central idea of the Disciple approach is to facilitate the agent building process by the use of synergism at several levels. First, there is the synergism between different learning methods employed by the agent. By integrating complementary learning methods (such as inductive learning from examples, explanation-based learning, learning by analogy, learning by experimentation) in a dynamic, task-dependent way, the agent is able to learn from the human expert in situations in which no single strategy learning method would be sufficient. Second, there is the synergism between teaching (of the agent by the expert) and learning (from the expert by the agent). For instance, the expert may select representative examples to teach the agent, may provide explanations, and may answer agent’s questions. The agent, on the other hand, will learn general rules that are difficult to be defined by the expert, and will consistently integrate them into its knowledge base. The Disciple approach is implemented in a tool, called Disciple Learning Agent Shell, that significantly reduces the involvement of the knowledge engineer in the process of building an intelligent agent. The Disciple Learning Agent Shell is implemented in Common LISP, and runs on Macintosh. It consists of four main domain independent components shown in the light gray area of Figure 1. They are: Exhibit at the International Conference on Intelligent Tutoring Systems, ITS'98, San Antonio, Texas, 1998. 2 • a knowledge acquisition and learning component for developing and improving the knowledge base, with a domain-independent graphical user interface; • a basic problem solving component which provides basic problem solving operations; • a knowledge base manager which controls access and updates to the knowledge base; • an empty knowledge base to be developed for the specific application domain. The two components in the dark gray area are the domain dependent components that need to be developed and integrated with the Disciple shell to form a customized agent that performs specific tasks in an application domain. They are: • a specialized problem solver which provides the specific functionality of the agent; • a domain-specific graphical user interface. Concept Editor Concept Browser Association Browser Dictionary Browser/Editor Rule Learner Rule Editor Rule Browser Rule Refiner Example Editor Explanation Grapher Transitivity and Inheritance Network Matching Rule Matching