The Future of Education

The rapid expansion of online learning platforms in recent years has led to a significant increase in the volume of learner-generated feedback. This growth has created a strong need for automated methods that are capable of effectively analyzing learner experiences and perceptions. The objective of this paper is, therefore, to propose a multi-task artificial intelligence framework for the joint analysis of learner satisfaction and engagement in online courses. To this end, a dataset of 25,000 reviews extracted from Coursera was used. These reviews were posted by learners as feedback on a wide range of courses across multiple disciplines, including technology, business, personal development, and other academic fields.

The proposed model, based on Bidirectional Encoder Representations from Transformers (BERT), performs two classification tasks simultaneously: (i) learner satisfaction prediction, derived from user-provided star ratings and categorized into negative, neutral, and positive classes, and (ii) learner engagement detection, inferred using a rule-based labeling strategy that identifies patterns related to technical difficulties, content interaction, and general feedback behavior. By exploiting shared contextual representations, the model learns transferable linguistic features across both tasks and enables a deeper understanding of learner feedback.

To assess its effectiveness and generalization capability, the proposed model was evaluated on a separate set of previously unseen reviews. The results indicate consistent predictive behavior across different subsets of learner feedback and confirm its ability to generalize beyond the training data. These findings demonstrate the robustness of this approach and support its applicability to real-world educational contexts.

The proposed approach highlights the potential of AI technologies for the extensive analysis of educational data. By jointly modeling learner satisfaction and engagement, the approach provides a more comprehensive understanding of learner experiences. Thus, it supports data-driven decision-making and enables educators and platform administrators to enhance course design, improve learner engagement, and optimize the overall quality of online learning environments.