Learning assistance mechanism using case-based reasoning

Abstract

An intelligent learning system can provide users with different learning materials and questions of fixed difficulty according to the learner's skill and knowledge. However, even the same question might have a different difficulty for different learners or at different learning stages. In this paper, we propose a learning assistance mechanism composed of the following functions: a question-setting assistance function that involves dynamically evaluating the difficulty of the question based on the knowledge state of the learner, and an answering assistance function that involves predicting the correctness of the answer, using case-based reasoning. In this mechanism, the knowledge state, treated as case components for each question, is extracted from the learner's knowledge model, and a case-base search is performed to find the correctness probability from analogous cases. All non-learned questions are dynamically arranged in the order of high correctness probability. Questions with high correctness probability have low difficulty, and conversely, those with low correctness probability have high difficulty. The teaching strategy for selecting a suitable question in accordance with the knowledge state of the learner is discussed. The mechanism collects a database of right answer cases and wrong answer cases from the answer history of the learner, for use in predicting his or her correctness to the current question. Based on the prediction result, the learner is offered assistance messages and a sorted keyword list to lead him or her towards successful learning. In case-base searching, an evaluation algorithm involving case reliability and relevant keyword importance is proposed to enhance reasoning accuracy. © 1998 Scripta Technica, Syst Comp Jpn, 29(5): 73–83, 1998