Developing an Original Theoretical Model of Trustworthy AI Deployment for Public Sector Decision Support in Developing Countries

AI-based decision support systems in developing country public sectors introduce distinctive trust, accountability, and legitimacy challenges that existing trustworthy AI frameworks, developed in high-income country contexts, inadequately address. This study develops an original theoretical model of trustworthy AI deployment for public sector decision support in developing countries, grounded in Nigerian and broader Sub-Saharan African institutional realities. The model, designated the Contextualised Public Sector AI Trustworthiness Model (CPS-AITM), is developed through a three-stage research design. First, a systematic literature review of trustworthy AI frameworks (EU AI Act, NIST AI RMF, Singapore Model AI Governance Framework) identifies the assumptions about institutional capacity, judicial oversight, and data infrastructure that limit their applicability to low-income country contexts. Second, qualitative primary research through 48 expert interviews with AI practitioners, public sector officials, civil society representatives, and legal scholars across Nigeria, Ghana, and Kenya generates empirically grounded trustworthiness requirements. Third, a theoretical construction phase synthesises review and empirical findings into the CPS-AITM, which introduces four original theoretical constructs absent from existing frameworks: community legitimacy validation, infrastructure-contingent reliability modelling, customary accountability norms integration, and decolonial explainability standards. The model was submitted to an expert validation panel of 15 international AI governance scholars and found to address documented gaps in existing frameworks. CPS-AITM constitutes an original contribution to AI governance theory for the global south.

Keywords: trustworthy AI, public sector AI, developing countries, AI governance, theoretical model