Supporting mathematical learning by using a case retrieval system that controls similarity between example problems

摘要

In mathematical learning, learners achieve learning by repeatedly solving example problems as analogical instances. Learners must notice specific features related to the solution, but they are strongly influenced the surface representation of the problems. Therefore it has been reported that their performance remarkably decreases when they solve a problem whose surface representation is different from an original problem even though the deep structure related to its solution is almost the same. These attributes, both surface and deep structures, are considered as meta information that novice learners are not generally aware of. We implemented a case retrieval system that controls the meta similarity, and we performed experimental evaluation for verifying the utility of the system. Key words mathematical learning, surface feature, analogical reasoning, analogical instance, similarity