Extending domain expertise through student modeling

The acquisition of knowledge is one of the largest obstacles in the development and application of knowledge-based systems and is referred to as the knowledge acquisition bottleneck. Machine learning techniques such as induction have been successfully employed to develop knowledge acquisition tools. However these systems require an explicit training mode and are dependent upon a set of hand-picked examples. Knowledge acquisition techniques which operate during a system's normal interactions with its users are needed.; This research focuses on the development of a knowledge acquisition technique to facilitate the development of complex knowledge-intensive applications such as intelligent instruction systems. The I{dollar}sp3{dollar} or Intelligent Individualized Instruction system is being developed to enhance the quality and efficiency of engineering education. Its objective is to provide the next generation of self-sufficient computer instruction systems to allow a student to obtain competency. The I{dollar}sp3{dollar} system will provide the individualized learning tools necessary to acquire and build the required knowledge bases (Biegel, 1993).; An Intelligent Learning System is an instruction system where both the student and the system acquire knowledge. Within an Intelligent Learning System, the student model is responsible for inferring the student's knowledge of the domain. It does this by observing and analyzing the problem-solving actions taken by the student. Therefore, the diagnostic abilities of the student model make it an ideal vehicle for learning new problem solving strategies from the student. I{dollar}sp3{dollar}/SM is a student model that opportunistically acquires knowledge. The acquired knowledge is problem-solving strategies that extend the Domain Expert knowledge base. By acquiring additional domain knowledge, the I{dollar}sp3{dollar}/SM facilitates the knowledge acquisition process for Intelligent Learning Systems such as I{dollar}sp3{dollar}, and reduces the initial amount of required domain knowledge.