Personalizing Second Language Learning: Integrating AI with Learner Preference, Proficiency, and Engagement

Personalization has become a cornerstone of online learning platforms, offering tailored experiences that enhance learner engagement, satisfaction, and performance. Moving beyond one-size-fits-all approaches, personalized systems can provide flexible access, improve efficiency, and support both cognitive and non-cognitive development. In language learning, personalization affords increased motivation, self-efficacy, confidence, and technology acceptance, while reducing instructor workload. Despite these benefits, many language learning platforms remain non-personalized. This study explores the integration of personalization in an online language learning platform, simultaneously taking into account three latent learner variables: preference, proficiency, and engagement. Preference captures learners' thematic and content choices, proficiency reflects their level of knowledge and skills based on their performance in the online learning environment, and engagement measures their sustained interaction with the leraning materials and risk of dropout. By personalizing based on these learner variables, we aim to tailor learning materials that align with learners' interests, match their abilities, and foster sustained participation. We validate our approach in the use case of learning Dutch as a second language, utilizing data from the free online platform NedBox, targeted at newcomers in Flanders. Across the three personalization variables, the experiments reveal that lightweight recommender systems outperform deep models for preference prediction, DeepIRT offers the strongest yet interpretable proficiency estimates, whereas random survival forest, particularly when augmented with those proficiency estimates, offers the most effective modeling of learner engagement. In line with teachers' views on meaningful personalization, these models can enable relevant recommendations regarding learning materials, driven by learner preference and engagement, as well as proficiency-aligned entry points into exercises. The source code is available at https://gitlab.kuleuven.be/kor-itec/XAI4PEPOL.