TOWARDS AN INTERPRETABLE AND PRIVACY-PRESERVING MUSIC RECOMMENDATION SYSTEM FOR TEENAGERS USING EXPLAINABLE AI

Abstract

The growing dependence of teenagers on music streaming platforms has raised significant concerns about bias, lack of transparency, and declining trust in recommendation systems. Biased or opaque recommendations can reduce user satisfaction, reinforce inequalities in content exposure, and hinder responsible personalization. Current music recommendation research primarily focuses on collaborative filtering and deep learning-based models, which, while effective in improving accuracy, are typically centralized, privacy-intrusive, and difficult to interpret. As a result, challenges related to fairness, interpretability, and privacy preservation remain unresolved. Addressing these limitations is essential to build trustworthy and user-centric music recommendation systems for teenagers. Federated Learning (FL) emerges as a promising paradigm to overcome privacy and scalability challenges by enabling decentralized training across multiple user devices without sharing raw data. While FL preserves user confidentiality, it still produces complex models that lack transparency. To ensure accountability and fairness, Explainable AI (XAI) offers interpretability by clarifying how input features influence recommendations, thereby enhancing user trust and addressing bias.

In this research, we propose an integrated FL–XAI framework for unbiased music recommendation tailored to teenage users. The framework combines the privacy-preserving strengths of FL with the transparency of XAI methods, such as SHAP and LIME, to deliver interpretable, trustworthy, and equitable recommendations. Experimental results demonstrate that the proposed model achieves an accuracy of 89% with a miss rate of only 11%, significantly outperforming prior methods [43–50]. This approach not only reduces systemic bias and protects user data but also provides meaningful justifications for recommendations, making it more reliable and socially responsible for real-world deployment.