From black box to glass box: A survey on explainable AI for accountable e-governance

While Artificial Intelligence (AI) is rapidly transforming e-governance by enhancing efficiency and enabling data-driven policymaking, its prevalent "black box" nature poses significant risks to transparency, fairness, and public trust. Opaque algorithmic decisions in the public sector can undermine accountability and challenge principles of due process. This paper provides a systematic survey of the literature on Explainable AI (XAI) applications within the e-governance sector. We aim to map the current landscape, categorize existing work, and identify key trends and challenges to guide future research. We reviewed 42 papers sourced from key academic databases, including IEEE Xplore, ACM Digital Library, and Scopus, published between 2018 and 2024. Our survey organizes applications into a novel taxonomy based on the governance domain and the XAI techniques used. We identify key trends, such as the prevalence of post-hoc explanation methods like SHAP and LIME over intrinsically interpretable models, and highlight significant research gaps, particularly in the co-design of explanations with non-expert stakeholders. This survey serves as a foundational resource for researchers, policymakers, and practitioners aiming to develop trustworthy AI systems for the public sector. By structuring the field and outlining critical challenges, we provide a roadmap for advancing the development of accountable and transparent AI in governance.