This study investigates the ethical and effective integration of Explainable Artificial Intelligence (XAI) into public sector audit systems in developing economies, using the Ghana Audit Service as a focal case. It responds to the critical governance challenge posed by opaque “black-box” AI models, which threaten transparency, legal defensibility and institutional accountability in high-stakes public audits. By addressing the interpretability gap, the study aims to ensure that AI-generated findings are not only technically robust but also publicly justifiable and democratically aligned.
A sequential explanatory mixed-methods design was employed. Quantitative experiments compared auditor trust, decision accuracy and perceived legal defensibility between XAI-enhanced and standard AI systems. Qualitative interviews and focus groups explored contextual factors, cognitive processes and governance implications of XAI adoption in public auditing.
Auditors using XAI-enhanced systems reported significantly higher trust, improved decision accuracy and stronger perceptions of legal defensibility compared to those using noninterpretable AI. Crucially, perceived explainability emerged as a key mediator between AI use and audit performance, while prior AI-related training significantly amplified trust in XAI outputs. These findings underscore the importance of interpretability and capacity-building in driving effective AI adoption. The study also revealed institutional and socio-technical enablers and barriers influencing XAI integration within Ghana’s public audit context.
The findings have significant implications for digital governance, especially in developing countries. By demonstrating that XAI can enhance trust, accuracy and legal defensibility in audits, the study supports policy reforms to embed explainability into AI governance frameworks. For practice, it underscores the need for structured auditor training and interdisciplinary collaboration between technologists, auditors and legal experts. The proposed framework can guide the Ghana Audit Service and similar institutions in adopting AI systems that meet both performance and accountability standards, ultimately strengthening transparency, reducing corruption risks and safeguarding democratic oversight in high-stakes public sector audits.
To the best of the author’s knowledge, this study provides one of the first empirical examinations of XAI integration in public sector auditing within a developing-country context. It delivers a context-aware framework that combines technical explainability with institutional governance safeguards, offering practical guidance for policymakers, audit institutions and AI developers to enhance transparency, fairness and public trust in AI-assisted audits.
