A Theoretical Framework for Trustworthy AI Deployment in Nigerian Public Sector Decision-Making Systems

AI systems are being considered for deployment in Nigerian public sector decision-making contexts including loan eligibility at the Bank of Industry, agricultural subsidy allocation, and welfare beneficiary selection under social protection programmes, yet no theoretical framework exists for evaluating trustworthiness along the dimensions of fairness, accountability, transparency, and robustness that are specifically adapted to Nigerian legal, cultural, and institutional contexts. This study developed an original theoretical framework for trustworthy AI deployment in Nigerian public sector contexts, making contributions to both AI ethics theory and practical certification methodology. The framework development followed a rigorous five-stage process: systematic review of global trustworthy AI frameworks from 2019 to 2024 (62 frameworks reviewed); primary research through structured consultations with 35 Nigerian stakeholders across government officials, AI researchers, civil society organisations, legal scholars, and citizens affected by algorithmic decisions; theoretical grounding in African Ubuntu ethics, Nigerian constitutional law, and international AI ethics principles; framework construction; and expert validation with 20 specialists. The original Nigerian Trustworthy AI Framework (NTAF) proposes six trust dimensions: Ubuntu-informed fairness (considering communal impact rather than only individual fairness metrics), constitutional accountability alignment, multilingual transparency (requiring decision explanations in relevant Nigerian languages), adversarial robustness certification, data sovereignty compliance, and participatory oversight mechanisms. Original contribution is made through the Ubuntu Fairness Formalisation, providing the first mathematical definition of communal fairness criteria grounded in African philosophical ethics. NTAF was applied to three case study AI systems from Nigerian government programmes, demonstrating practical operationalisability. Expert consensus confirmed its original theoretical contribution.