From Missing Censuses to Smarter Decisions: How AI Is Reshaping Population Data

This recording features Prof Andy Tatem's presentation and the Q&A session from Part 1 of the IUSSP & PAA Webinar Series on AI & Demography, moderated by Vissého Adjiwanou (Université du Québec à Montréal). The session brings together leading population scientists to explore how artificial intelligence and geospatial techniques (GeoAI) are being applied to population estimation, small-area mapping, and demographic modelling, while addressing critical methodological and ethical challenges. 00:00 Introduction & Opening Remarks (Vissého Adjiwanou) 00:35 Andrew Tatem: WorldPop, GeoAI & High-Resolution Population Mapping 08:20 - Warnings on GeoAI: Regional Biases, Foundation Models & Declining Censuses 14:00 - Q&A: Model Validation & Propagating Uncertainty 18:30 - Q&A: Prediction vs. Explanation & Demographic Theory 23:45 - Q&A: Complexities & Limits of Synthetic Demographic Data Key Presentation & Discussion Topics GeoAI & Spatial Disaggregation: Prof. Andrew Tatem introduces WorldPop and demonstrates how satellite imagery, street-view AI, and machine learning models map high-resolution population distributions for public health, disaster response, and resource allocation. Limitations & Risks of AI in Mapping: The panel cautions against using AI models trained on Global North data in the Global South, black-box foundation models lacking explainability, and relying on AI amid declining global census coverage. Model Validation Without a Gold Standard: Panellists discuss handling uncertainty, distinguishing design uncertainty from sampling uncertainty, and propagating confidence intervals. Prediction vs. Explanation & Theory: Prof Ridhi Kashyap, Dr Matthew Hauer, and the panel debate the balance between predictive power, demographic theory, model interpretability, and auditing AI systems for social equity and justice Synthetic Demographic Data: Dr Emilio Zagheni and Dr Matthew Hauer examine the mathematical hurdles of generating high-dimensional synthetic populations that preserve multivariate statistical relationships while maintaining privacy. Featured Speakers & Panellists Moderator: Vissého Adjiwanou — Associate Professor, Université du Québec à Montréal Andrew Tatem — Director of WorldPop & Professor of Spatial Demography and Epidemiology, University of Southampton Matthew Hauer — Head of Population Simulation at Aaru Inc. & Associate Professor, Florida State University Ridhi Kashyap — Professor of Demography and Computational Social Science, University of Oxford Emilio Zagheni — Director, Max Planck Institute for Demographic Research (MPIDR) #Demography #ArtificialIntelligence #GeoAI #WorldPop #PopulationResearch #DataScience #SyntheticData #IUSSP