A Comparative Study of Explainable AI Techniques for Bias Mitigation and Trust in E-Learning Recommendation Systems

Abstract

The adoption of Artificial Intelligence (AI) in e-learning has significantly improved personalized learning by tailoring recommendations to individual learners. However, inherent biases in AI-driven recommendation systems and their opaque “black-box” nature undermine fairness, transparency, and user trust. This study conducts a comparative analysis of various Explainable AI (XAI) techniques to address these challenges, focusing on bias mitigation and enhancing trust in e-learning recommender systems. The research investigates multiple XAI frameworks, including model-agnostic and model-specific methods, to evaluate their effectiveness in identifying and mitigating biases while providing interpretable recommendations. By integrating explainability into machine learning models, this study ensures that learners receive transparent and fair suggestions tailored to their needs, accompanied by clear justifications. Empirical evaluations are conducted on benchmark e-learning datasets to assess the performance of the proposed techniques in terms of accuracy, bias reduction, transparency, and user trust. Results indicate that XAI techniques not only improve recommendation fairness but also enhance learner engagement and confidence in the system. This comparative study highlights the importance of adopting explainable, ethical, and bias-aware AI methods for e-learning platforms. The findings provide actionable insights for educators, developers, and policymakers to design equitable and trustworthy recommendation systems, advancing the effectiveness and inclusiveness of AI-powered education.