This research will apply case-based reasoning techniques to the construction of intelligent tutoring systems for diagnostic problem solving. It is believed that when a tutoring system's expertise and curriculum are anchored in the same set of cases, the tutor will be better able to select appropriate training cases, to assess student performance on these cases, and to remediate when errors are detected. This work will build on the foundation of the Protos automated knowledge acquisition system. Complete tutoring systems will be developed in at least two domains. Evaluation studies will compare the tutor's effectiveness to that of classroom instruction. Evaluation in multiple domains will permit cross-domain comparison to separate general results from domain idiosyncrasies. The project will contribute to three areas of intelligent tutoring research. First, it will address the modeling of expert and student knowledge in "weak-theory" domains. Second, it will address the diagnosis and remediation of student errors. Third, it will address curriculum generation and sequencing. In the long term, synthesis of the lessons learned from research in intelligent tutoring systems, expert consultation systems, and automated knowledge acquisition will contribute to a deeper understanding of diagnostic expertise.
