Harnessing artificial intelligence for census in Nigeria: Advancing accuracy, efficiency, and governance outcomes

Abstract

Successful administration of national censuses in Nigeria has been a protracted agony plagued by inherent problems, including logistic, political, and methodological issues, which cumulatively have caused delays in enumeration, undercounting, and inconsistency of data. These defects diminish the credibility of demographic data needed for evidence-based governance, economic planning, and equitable resource allocation._. In this study, we explored opportunities for harnessing Artificial Intelligence (AI) to transform census activities in Nigeria through the injection of state-of-the-art computational approaches into the national enumeration exercise. We showcased a multimodal AI pipeline comprising Convolutional Neural Networks (CNNs) for population density estimation from satellite images, Natural Language Processing (NLP) pipelines for address standardization and matching in various languages, and unsupervised anomaly detection algorithms for real-time data quality verification. AI-based enumeration methods were simulated at both national and sub-national levels. CNN-generated heatmaps revealed population concentration trends in Lagos and other states and enabled the precise delineation of high-density urban agglomerations and underserved rural enclaves. The NLP tool generalized well to the linguistically diverse environments in Nigeria, with F1-scores greater than 0.90 for all but a few states for broken address reconciliation. Anomaly detection models built using Isolation Forest algorithms detected anomalous enumeration patterns as flags for potential undercounts or data manipulation. Population pyramid analysis for Lagos revealed an extremely young population structure, consistent with country-wide age trends. These findings provide empirical evidence that AI integration can promote census accuracy, operational efficiency and government effectiveness in Nigeria.