Analysis of Learning Curves for Weighted Constraint-based Tutoring Systems

Weighted constraints have been proven a useful means to enhance the quality of error diagnosis, e.g., in the domain of natural language parsing. In this paper, we propose an heuristic approach using constraint weights to analyze learning curves for weighted constraint-based tutoring systems as an addition (or alternative) to the commonly used analysis of pre-/post-test comparisons. Using a tutoring system for logic programming on the basis of weighted constraints as an example case, we report on the effectiveness of the system with respect to helping students solve logic programming problems. The study showed that the group learning curves developed positively as students were solving five exercises.