Knowledge Graphs & eXplainable AI (XAI)

Seong Woo Ahn, Paul Bérard, Charlotte Sasson

Abstract: This paper explores the integration of Knowledge Graphs (KGs) into eXplainable AI (XAI) as a method to enhance the interpretability of AI-driven systems across diverse domains, including computer vision, recommender systems, natural language processing, predictive tasks, and healthcare. As AI systems continue to influence high-stakes industries, the demand for transparency and comprehensible decision-making grows. Knowledge Graphs offer a structured way to represent data relationships, facilitating insights into AI model decisions. We review the literature on KGs for XAI applications, highlighting their impact in improving performance and interpretability through semantic and relational insights. Additionally, a case study on explainable recommender systems demonstrates the utility of Knowledge-Based Embeddings (KBE) for producing personalized recommendations with transparent, user-friendly explanations. This study outlines a methodology for constructing KBE-enhanced collaborative filtering systems, emphasizing explainable recommendation paths. The paper concludes by discussing the promise and challenges of KGs in XAI, positioning them as essential tools for advancing AI transparency, trustworthiness, and user engagement.