Authors: Sara Saroj Pathan, Mrunal Jitendra Palaskar, Chetan Hiraman Satpute, S. N. Jadhav
Abstract: The rapid expansion of e-learning platforms has enabled large-scale access to education; however, most existing systems continue to employ static content pacing strategies that fail to accommodate individual learner differences. Such one- size-fits-all approaches often result in learner disen-gagement, inefficient knowledge acquisition, and high dropout rates. This work presents an AI-driven personalized learning pace optimizer that integrates Reinforcement Learning (RL) with Self-Paced Learning (SPL) to dynamically adapt instructional pacing based on a learner’s evolv-ing knowledge state. SPL is used to structure educational content from easy to hard, providing pedagogical stability and robustness to noisy learner data, while RL models the pacing decision as a sequential optimization problem. A multi-objective reward formulation is adopted to balance learner engagement, knowledge retention, and learning efficiency. The proposed approach pro-vides a technically robust and pedagogi- cally safe architecture for adaptive e-learning systems and serves as a foundation for future real-world deployment and evaluation.
DOI: http://doi.org/10.5281/zenodo.20927889
