A Framework for Improving Recommendation in Adaptive Metacognitive Scaffolding

Abstract

Metacognitive scaffolding is an important pedagogical support to encourage self-regulated learning, especially in computer-based learning environments. Currently, an e-learning system which is complemented by metacognitive scaffolding has been developed. The scaffolding includes a recommendation system given by a virtual pedagogical agent. However, an academic satisfaction evaluation on the system reveals that the recommendation is less helpful. Therefore, this paper proposes a framework for providing a better recommendation. Two types of recommendations are aimed, including recommendation for the definition of learning goal/sub-goal and learning strategy. Goal/sub-goal recommendation is generated by identifying students' level of prior knowledge by using text classification algorithm. On the other hand, the strategy-use recommendation is generated by modeling the fitness of previously used learning strategy using fuzzy inference system and analyzing students' interaction log with the learning strategy. Implementation of the proposed scheme produces a prototype of the recommendation system. An A/B testing is conducted to compare previous and newer recommendation systems. The test shows that most of the users prefer to use the improved recommendation system.