Towards an explainable course recommendation system (xcrs)

The rapid expansion of online learning resources in Information Technology (IT) necessitates personalized guidance and transparent course recommendations. Learners face an overwhelming number of courses and career pathways, making it challenging to find tailored content aligned with their learning objectives. Existing recommendation systems often lack transparency and explanations for their suggestions. To address this gap, this study introduces the Explainable Course Recommendation System (XCRS)—the first to recommend both career roles and associated courses in an explainable manner. XCRS leverages large language model (LLM) embeddings from Google, OpenAI, MistralAI, VoyageAI, and Cohere to generate personalized suggestions based on users' knowledge, experiences, and learning interests. A unique dataset of online courses was collected and shared publicly to enrich the recommendation resource base. XCRS utilizes cosine similarity for precise vector matching and Retrieval-Augmented Generation (RAG) to enhance explanation quality, improving both transparency and relevance. Evaluated through both user and expert judgment studies, XCRS demonstrated high performance in transparency and efficiency, rated at 89.6% and 88.8%, respectively, as well as overall user satisfaction. Experts found recommendations well-aligned with user needs, validating the system's capacity to support informed decision-making in educational and career pathways. All resources needed to reproduce our results, such as code, datasets, and experimental setup, are openly provided.