This study suggests that educational recommender systems should be explainable and extend beyond the commercially driven algorithms that primarily rely on user preferences and purchase behaviors. Instead, we propose a recommendation method that considers how and why people learn by employing the relative importance of various learner variables. To develop a recommendation algorithm, 100 adult participants used 4 to 6 foreign language learning mobile applications(apps), generating a dataset of 557 user perception reports. Using this data, we designed and developed a recommender system based on the importance weights of 14 learner variables, categorized into four groups: (a) demographic information, (b) motivational orientation for language learning (instrumental vs. integrative), (c) learning styles, and (d) learning experience. The results based on RandomForestRegressor model revealed that language learning motivation, learning styles (specifically information processing), and usage frequency were significantly more influential than general demographic factors in predicting learners’ evaluation of the apps. Furthermore, learners’ perception of the recommender system revealed that the recommender system was relevant and engaging, effectively meeting their needs and assisting them in selecting appropriate language learning apps. Overall, this study demonstrates the potential of educational recommender systems that consider learners’ motivation, experience, and learning styles. 

Keywords : Recommender System, Mobile applications, Mobile learning, Learner variables, Individual Difference