Explainable clustering: Methods, challenges, and future opportunities

Abstract In recent years, artificial intelligence (AI) has increasingly relied on subsymbolic techniques like machine learning (ML). Despite their widespread use, these techniques often lack transparency, leading to potential distrust. The field of eXplainable artificial intelligence (XAI) addresses this issue by making intelligent systems observable, explainable, and accountable. While much research has focused on explainability in supervised learning, there is a growing need to explore it in an unsupervised setting, especially given the challenges of unlabeled data in high volume. Clustering is an unsupervised ML strategy that groups data based on similarity. However, its reasoning often lacks transparency. This article reviews state-of-the-art explainable and/or interpretable clustering methods, categorizing them based on explanation generation techniques and highlighting the importance of making clustering results interpretable. We also discuss the challenges and opportunities in this domain and suggest future research directions, particularly the interpretability of advanced AI techniques like neural networks and large language models in the context of clustering. Our contributions include a comprehensive categorization of explainable clustering research and potential future research avenues to enhance the transparency and trustworthiness of clustering methods.