ARCHITECTING EDUCATIONAL RESILIENCE: EXPLORING AI-ENABLED KNOWLEDGE MANAGEMENT SYSTEMS AS LEARNING CONTINUITY INFRASTRUCTURE IN RIVERS STATE, NIGERIA

Abstract

Learning poverty in the Global South stems not only from policy gaps but from a deeper evidence failure: governments and partners cannot effectively invest in what they cannot systematically observe. This paper presents Read Cycle Nigeria’s two-phase community needs assessment across Rivers State as an empirical foundation for theorising AI-enabled Knowledge Management Systems (AI-KMS) as structural educational resilience infrastructure. Phase 1, conducted in 35 primary schools across nine Local Government Areas (LGAs), directly assessed 4,372 pupils and revealed chronic infrastructure collapse, severe teacher shortages, acute resource scarcity, and critical WASH deficits. Gender-disaggregated findings identified barriers affecting girls, including domestic labour demands, inadequate menstrual hygiene facilities, and reported intra-household resource inequalities. Phase 2 is extending the assessment across a wider set of LGAs, incorporating gender-disaggregated indicators and AI-KMS readiness metrics such as connectivity, device ownership, and community knowledge flows. Fieldwork also surfaced a critical implementation constraint: in several LGAs, schools were identified but could not be assessed due to the absence of trained personnel on the ground, a gap that itself evidences the need for a structured community assessor model. The paper builds a synthetic theoretical matrix from five established frameworks, Nonaka and Takeuchi’s knowledge creation theory and its later Ba refinement, Alavi and Leidner’s KMS framework, Wenger’s communities of practice, Bejinaru’s knowledge dynamics and its thermodynamics reformulation, and Ostrom’s theory of polycentric governance, the paper conceptualises AI-KMS as potentially supporting continuous knowledge renewal in volatile, low-infrastructure environments because such systems improve through ongoing interaction rather than requiring stable conditions, they can sustain knowledge flows where centralised, static systems fail. This capacity for contextual adaptation, grounded in community-generated evidence rather than externally imposed curricula, addresses the epistemological asymmetry that globally designed AI systems routinely reproduce in Global South educational environments. Critically, ReadCycle field evidence further specifies that a viable AI-KMS in this context must be offline-first, multilingual, community-governed, and gender-responsive by design. The paper concludes with a replicable scaling model for Nigerian states, positions trained community assessors as decentralised governance nodes within a polycentric AI-KMS architecture, and sets out the empirical, methodological, and technical limitations that must be addressed before the architecture can be validated at scale.