Expanding the Plausible Solution Space for Robustness in an Intelligent Tutoring System

摘要

The knowledge acquisition bottleneck is a problem pertinent to the authoring of any intelligent tutoring system. Allowing students a broad scope of reasoning and solution representation whereby a wide range of plausible student solutions are accepted by the system, places additional burden on knowledge acquisition. In this paper we present a strategy to alleviate the burden of knowledge acquisition for building a tutoring system for medical problem-based learning (PBL). The Unified Medical Language System (UMLS) is deployed as domain ontology and information structure in the ontology is exploited to make intelligent inferences and expand the domain model. Using these inferences and expanded domain model, the tutoring system is able to accept a broader range of plausible student solutions that lie beyond the scope of explicitly encoded solutions. We describe the development of a tutoring system prototype and report the evaluation of system correctness in accepting such plausible solutions. The system evaluation indicates an average accuracy of 94.59% when compared against human domain experts, who agreed among themselves with a statistical agreement based on Pearson Correlation Coefficient of 0.48 and p < 0.05.