A generic intelligent architecture for computer-aided training of procedural knowledge

Abstract : Intelligent Tutoring System (ITS) development is a knowledge-intensive task, suffering from the same knowledge acquisition bottleneck that plagues most Artificial Intelligence (AI) systems. This research presents an architecture that requires knowledge only in the form of a shallow knowledge base and a simulation to produce a training system. The knowledge base provides the basic procedural knowledge while the simulation provides context. The remainder of the knowledge required for training is learned through the interaction of these components in a state-space scenario exploration process and inductive machine learning. These knowledge components are used only at the interface level, allowing the internal representation to take any form that meets the interface requirements. A prototype of this architecture is implemented as a proof-of-concept to illustrate the viability of the key knowledge acquisition techniques.