Pedagogically-Informed Implementation of Reinforcement Learning on Knowledge Graphs for Context-Aware Learning Recommendations

Context-aware recommender systems are important tools to address the learning context in technology enhanced learning. However, the contextual factors of learning, as well as the mechanisms of integrating them into the recommendation, are mostly defined from a technical perspective, rather than a pedagogical one. In this paper, we introduce a new approach for generating pedagogically informed, context-aware, learning recommendations. We build on the context definition in situated and subject-oriented learning theories. Then, we utilize a knowledge graph structure to build the environment for a path exploration and ranking algorithm, which is influenced by agent exploration in reinforcement learning (RL), for creating sequential learning-path recommendations. Our design of the agent's reward function integrates learning-context factors in the recommender system. We evaluate the proposed solution qualitatively with domain experts, and quantitatively using semantic-similarity measures to compare our recommended paths to expert-curated learning content. Our evaluation shows an enriched recommendation based on the learners' context, as well as a better discovery of relevant educational content.