A Theoretical Investigation of Responsible AI Governance Mechanisms for Automated Decision-Making in Nigerian Social Protection Programmes

AI-assisted automated decision-making in Nigerian social protection programmes, including the National Social Investment Programme and state-level conditional cash transfer schemes, raises governance challenges including algorithmic discrimination, opacity, lack of appeal mechanisms, and the power asymmetry between implementing agencies and programme beneficiaries. This study develops an original theoretical investigation of responsible AI governance mechanisms for automated decision-making in Nigerian social protection. The investigation generates three theoretical contributions: a political economy analysis of AI governance failure modes specific to Nigerian social protection contexts, drawing on 60 fieldwork interviews with programme administrators, beneficiaries, civil society advocates, and AI developers across North Central, South West, and South East zones; an original Beneficiary-Centred AI Accountability Theory (BCAAT) that repositions beneficiaries as rights-holding stakeholders in AI governance rather than passive subjects, drawing on African ubuntu communal epistemology and international human rights law; and a Nigerian Social Protection AI Governance Framework (NSPAIGF) with five operational components: algorithmic transparency requirements proportionate to decision stakes, beneficiary appeal and explanation mechanisms, independent algorithmic audit processes, participatory system design protocols, and sunset review provisions. The NSPAIGF was validated through consultative workshops with 12 programme administrators, 10 civil society representatives, and 5 AI ethics scholars. The framework constitutes an original contribution to responsible AI governance theory for the global south with direct policy applicability.

Keywords: AI governance, social protection, automated decision-making, Nigeria, responsible AI