AI, qualitative data, and the case for statistical rigour

African Data for African Health: Building the AI Foundations the Continent Needs

🎙️ Episode Title African Data for African Health: Building the AI Foundations the Continent Needs --- 🧠 Episode Summary In this episode, Oliver Morgan speaks with Agnes Kiragga, Head of the Data Science Program at the African Population and Health Research Center in Nairobi, Kenya. Drawing on over two decades of experience building data systems and leading large-scale multi-country initiatives across the continent, Agnes reflects on what it takes to make AI work in African public health — not as a concept, but in practice, inside real institutions with real constraints. The conversation spans the full range of Agnes's work: from no-code AI platforms that bring real-time analytics to researchers without coding backgrounds, to AI-powered blood stock management systems in Cameroonian hospitals, to federated analysis platforms that allow data to be analyzed across borders without ever leaving its home institution. The episode addresses one of the most consequential gaps in global health AI: the underrepresentation of African datasets in the models being built for epidemic intelligence. Agnes argues that tools not trained on African data cannot capture the comorbidity burden, the linguistic diversity, or the epidemiological reality of African populations — and that the consequences range from missed diagnoses to wrong predictions during outbreaks. She also makes the case for a new generation of evaluation frameworks for large language models in African health contexts, where multi-turn, multilingual interactions are the norm, not the exception. Throughout, Agnes is clear-eyed about what is still missing — governance structures, infrastructure, skilled personnel — while pointing to the momentum building across the continent in training, tooling, and institutional capacity. Her core argument: AI for African health cannot be imported. It has to be built, evaluated, and owned from within. --- 💬 Guest Agnes Kiragga is Head of the Data Science Program at the APHRC in Nairobi, Kenya. With over 20 years of experience working with large and diverse health datasets across Africa, she leads the Data Science Without Borders project. She also leads the INSPIRE Network, a consortium of approximately 25 longitudinal Health and Demographic Surveillance System sites across Africa. Her work spans data harmonization, federated analysis, AI governance, and the responsible evaluation of AI tools in low- and middle-income country settings. She is a delivery partner for the Evidence for AI in Health initiative, which generates rigorous evidence on whether AI tools actually work in the contexts where they are deployed. --- 🌐 Resources and References - APHRC: https://aphrc.org/ - DSWB: https://dswb.africa; https://doi.org/10.3389/fpubh.2025.1695907 - INSPIRE Network: https://www.inspiredata.network/; https://aphrc.org/inspire/; https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2024.1329630/full - DataSHIELD: https://assets-eu.researchsquare.com/files/rs-9149238/v1/30a60413-de69-4c32-9237-577e7690944c.pdf?c=1776244497 - EVAH: https://www.povertyactionlab.org/initiative/evidence-ai-health-evah-initiative - AfriMed-QA: https://afrimedqa.com/ - OHDSI common data model: https://www.ohdsi.org/ - AIMS: https://aims.ac.za/ - Data Science Africa: https://www.datascienceafrica.org/ - Deep Learning Indaba: https://deeplearningindaba.com/ - IDRC AI for Development: https://www.idrc.ca/en/initiative/artificial-intelligence-development --- 🎵 Music Credits Track: ‘Nairobi Nights’. License code: WPJKGDX60OHV0I2I. --- ⚠️ Disclaimer This podcast is produced by the World Health Organization (WHO) as part of the Pandemic and Epidemic Intelligence Innovation Forum initiative. The views expressed by guests are their own and do not necessarily represent those of WHO or its affiliates. Content is intended for informational purposes only and does not constitute professional medical advice.

Beyond the Genome: Tracking Climate-Driven Epidemic Risk with AI

🎙️ Episode Title Beyond the Genome: Tracking Climate-Driven Epidemic Risk with AI --- 🧠 Episode Summary In this episode of The Innovation Forum AI Podcast, Oliver Morgan speaks with Houriiyah Tegally, Associate Professor and Head of Data Science at the Centre for Epidemic Response and Innovation (CERI), Stellenbosch University. Drawing on her work tracking viral pathogens across Africa — including a pivotal role in identifying the Beta and Omicron variants of concern during the COVID-19 pandemic — Houriiyah reflects on what it takes to move from genomic data to actionable epidemic intelligence, and why that challenge is becoming more urgent as the climate changes. The conversation traces how her team has built a global consortium to forecast the risk of climate-amplified diseases like dengue and chikungunya, and how AI is helping them integrate the disparate data streams — genomic, ecological, climatic, epidemiological — that this work requires. Houriiyah explains what phylogenetics and molecular clocks tell us about outbreaks, how machine learning is accelerating that analysis, and where it is introducing new risks: overfitting, unexplainable outputs, and models that perform well in one context and fail in another. The episode is frank about the limits of what AI can currently deliver. Houriiyah describes firsthand experiences with large language model hallucinations in scientific data curation, and argues that AI tools in epidemic intelligence are still best understood as a junior research assistant, not a domain expert. She makes the case that the most important investments are not in the technology itself, but in the ecosystem around it: infrastructure, local capacity, and the co-development of tools with the public health decision-makers who need to act on their outputs. --- 💬 Guest Houriiyah Tegally is an Associate Professor and Head of Data Science at the Centre for Epidemic Response and Innovation (CERI), Stellenbosch University. Her research focuses on the genomic epidemiology and evolution of emerging viral pathogens in South Africa and across the African continent. She uses genomic surveillance, phylodynamics, geospatial data, and machine learning to track and predict infectious disease outbreaks. During the COVID-19 pandemic, she led the phylodynamic analysis of SARS-CoV-2 genomes from multiple African countries, work that was pivotal in the identification of the Beta and Omicron variants of concern. She co-leads CLIMADE, an international consortium forecasting the transmission risks and outbreak potential of climate-amplified pathogens, and is involved in the African Institute for Mathematical Sciences (AIMS) AI for Science programme, which trains mathematicians across Africa to apply AI methods to real epidemic research problems. --- 🌐 Resources and References - Visit CERI: https://ceri.africa/ - CLIMADE — Climate Amplified Diseases and Epidemics: https://climade.health/ - CERI Data Science: https://ceri.africa/data-science/ - The GEM Newsletter: https://ceri.africa/the-gem/ - Alignment-free viral sequence classification at scale: https://link.springer.com/article/10.1186/s12864-025-11554-5 - Craft: a machine learning approach to dengue subtyping: https://academic.oup.com/bioinformaticsadvances/article/5/1/vbaf224/8275733 - African Institute for Mathematical Sciences (AIMS) AI for science program: https://ai.aims.ac.za/ --- 🎵 Music Credits Intro and outro music from Podcastle Stock Audio. Track: ‘Nairobi Nights’. License code: 8HLBOKOIASQ8R7GE. --- ⚠️ Disclaimer This podcast is produced by the World Health Organization (WHO) as part of the Pandemic and Epidemic Intelligence Innovation Forum initiative: https://pandemichub.who.int/news-room/innovation-forum. The views expressed by guests are their own and do not necessarily represent those of WHO or its affiliates. Content is intended for informational purposes only and does not constitute professional medical advice.

AI, qualitative data, and the case for statistical rigour

🎙️ Episode Title AI, qualitative data, and the case for statistical rigour --- 🧠 Episode Summary In this episode of The Innovation Forum AI Podcast, Oliver Morgan speaks with Adam Kucharski, Professor of Infectious Disease Epidemiology at the London School of Hygiene & Tropical Medicine and Co-Founder of WholeSum. Drawing on decades of experience working with messy, incomplete data — from contact surveys and outbreak investigations to real-time modelling during major epidemics — Adam reflects on a persistent frustration in public health: there is often far more information available than can be rigorously analysed. The conversation explores what AI tools now make possible for qualitative public health data — community narratives, open-ended survey responses, field reports — and what rigour actually requires before those outputs can be trusted. Adam explains key concepts including labelled training data and ground truth, and unpacks why collapsing human disagreement into a single consensus label can quietly undermine a model's usefulness. He also discusses the hidden assumptions embedded in AI models when they are applied in contexts different from where they were trained, and why understanding those assumptions matters as much as understanding the model's performance metrics. The episode closes with Adam's vision for tools that can go deeper into qualitative signals — not just classifying broad topics or sentiments, but extracting the underlying structures that link scattered observations to meaningful public health insights — while maintaining the reproducibility and statistical accountability that the field demands. --- 💬 Guest Adam Kucharski is a Professor of Infectious Disease Epidemiology at the London School of Hygiene & Tropical Medicine, where his research focuses on developing statistical and computational methods to extract reliable insights from incomplete and noisy data. He has advised multiple governments during outbreaks including Ebola and COVID-19, and has contributed to large-scale studies of social behaviour, population immunity, and transmission dynamics. He is also Co-Founder of WholeSum, a startup developing hybrid AI tools to bring statistical rigour to the analysis of qualitative text data at scale. --- 🌐 Resources and References - WholeSum: https://www.wholesum.tech/ - WholeSum pre-seed announcement: https://tech.eu/2026/01/05/wholesum-raises-730k-to-advance-qualitative-data-analysis-platform/ - When is the 'ground truth' not quite the whole truth?: https://kucharski.substack.com/p/when-is-the-ground-truth-not-quite - Inference of epidemic dynamics in the COVID-19 era and beyond (Cori & Kucharski, Epidemics, 2024): https://www.sciencedirect.com/science/article/pii/S1755436524000458 - How our concepts of what we can prove are shifting: https://www.theguardian.com/books/2025/mar/29/epidemiologist-adam-kucharski-proof-the-uncertain-science-of-uncertainty - Adam Kucharski's Substack — Understanding the Unseen: https://kucharski.substack.com --- 🎵 Music Credits Intro and outro music from Podcastle Stock Audio. Track: ‘Nairobi Nights’. License code: FJAMWGPDVHFKSGC2. --- ⚠️ Disclaimer This podcast is produced by the World Health Organization (WHO) as part of the Pandemic and Epidemic Intelligence Innovation Forum initiative: https://pandemichub.who.int/news-room/innovation-forum. The views expressed by guests are their own and do not necessarily represent those of WHO or its affiliates. Content is intended for informational purposes only and does not constitute professional medical advice. --- 📲 Listen and Subscribe The Innovation Forum AI Podcast is available on YouTube, Spotify, Apple Podcasts, and Amazon Music. You can find a written summary of this episode here: https://substack.com/@omorgan? Follow, rate, and share to help us reach more public health professionals exploring the future of AI.

AI, Genomics, and the Future of Public Health Surveillance

🎙️ Episode Title AI, Genomics, and the Future of Public Health Surveillance --- 🧠 Episode Summary In this episode of The Innovation Forum AI Podcast, Oliver Morgan speaks with Túlio de Lima Campos, Bioinformatics Core Facility Coordinator at Fiocruz in Brazil. Túlio explains how combining genomic sequencing with artificial intelligence is transforming what public health surveillance can see and do. Drawing on his work tracing arboviral outbreaks — including dengue, Zika, chikungunya and Oropouche fever — he discusses how genomic data reveals transmission dynamics that case counts alone cannot capture, and how machine learning can be layered on top to detect anomalies, prioritize signals, and anticipate risk. The conversation covers the practical meaning of ML-driven surveillance, lessons from foundational work on essential gene prediction in model organisms, and the realistic role of large language models in public health workflows. Túlio also addresses the critical challenge of data quality, the underrepresentation of Latin America in global genomic databases and AI models, and the governance, infrastructure and workforce investments needed to build genuinely integrated epidemic intelligence systems. --- 💬 Guest Túlio de Lima Campos is a Public Health Technologist and Coordinator of the Bioinformatics Core Facility at Fiocruz, a leading public health agency in Brazil. His work focuses on developing bioinformatics solutions for public health, applying AI and machine learning to genomic surveillance, and fostering scientific collaboration across the region. He has led research on arboviral genomic epidemiology, viral computational workflows, and regional capacity building in AI for the biosciences. --- 🌐 Resources and References - Fiocruz - Oswaldo Cruz Foundation: https://portal.fiocruz.br/en - Genomics of pathogens and vectors: ****https://www.frontiersin.org/journals/genetics/articles/10.3389/fgene.2024.1483676/full - Revisiting Key Entry Routes of Human Epidemic Arboviruses into the Mainland Americas through Large-Scale Phylogenomics: https://onlinelibrary.wiley.com/doi/full/10.1155/2018/6941735 - ViralFlow v1.0—a computational workflow for streamlining viral genomic surveillance: https://academic.oup.com/nargab/article/6/2/lqae056/7682253 - Genomic and phenotypic characterization of the Oropouche virus strain implicated in the 2022–24 large-scale outbreak in Brazil: https://onlinelibrary.wiley.com/doi/full/10.1002/jmv.70012 - Harnessing model organism genomics to underpin the machine learning-based prediction of essential genes in eukaryotes – Biotechnological implications: https://www.sciencedirect.com/science/article/pii/S0734975021001282 - BiotrAIn — AI capacity building initiative for Latin America: https://www.ebi.ac.uk/training/our-partnerships/biotrain --- 🎵 Music Credits Intro and outro music from Podcastle Stock Audio. Track: ‘Nairobi Nights’. License code: KTWK5CG66SP0Q0EU. --- ⚠️ Disclaimer This podcast is produced by the World Health Organization (WHO) as part of the Pandemic and Epidemic Intelligence Innovation Forum initiative: https://pandemichub.who.int/news-room/innovation-forum. The views expressed by guests are their own and do not necessarily represent those of WHO or its affiliates. Content is intended for informational purposes only and does not constitute professional medical advice. --- 📲 Listen and Subscribe The Innovation Forum AI Podcast is available on YouTube, Spotify, Apple Podcasts, and Amazon Music. You can find a written summary of this episode here: https://substack.com/@omorgan? Follow, rate, and share to help us reach more public health professionals exploring the future of AI.

Using AI to Strengthen Infectious Disease Surveillance Systems

🎙️ Episode Title Using AI to Strengthen Infectious Disease Surveillance Systems --- 🧠 Episode Summary In this episode of The Innovation Forum AI Podcast, Oliver Morgan speaks with Swapnil Mishra, from the National University of Singapore (NUS). Swapnil explains how AI and infectious disease modelling can move beyond academic exercises and become embedded within real public health workflows. Drawing on examples from dengue surveillance, vector-borne disease risk assessment, mobility data integration, and behavioral modelling, he discusses how AI can help public health agencies interpret surveillance data, combine fragmented datasets, and make decisions under uncertainty. The conversation explores the trade-offs between mechanistic models, Bayesian approaches, and machine learning; the role of hierarchical modelling in understanding local heterogeneity; and why separating transmission dynamics from reporting bias is essential for responsible epidemic intelligence. Swapnil also emphasizes that AI should augment public health professionals rather than replace them, and that the real transformation lies in strengthening institutional capacity for infectious disease decision making. --- 💬 Guest Dr. Swapnil Mishra is an Assistant Professor at the Saw Swee Hock School of Public Health at the National University of Singapore (NUS) and Deputy Director of the Centre for Epidemic Research and Modelling (CERM) and the AI for Public Health programme (AI4PH). His research focuses on infectious disease modelling, Bayesian statistics, and machine learning applied to epidemic and pandemic intelligence. He previously contributed modelling advice during the COVID-19 pandemic and works at the intersection of epidemiology, data science, and public policy. --- 🌐 Resources and References - NUS Saw Swee Hock School of Public Health: https://sph.nus.edu.sg/; https://www.linkedin.com/school/saw-swee-hock-school-of-public-health/?originalSubdomain=sg - Centre for Epidemic Research and Modelling: https://cerm.nus.edu.sg/ - Artificial intelligence for public health can harness data for healthier populations: https://www.nature.com/articles/s44360-025-00005-w.pdf - Estimating dengue force of infection from age-stratified surveillance data in Java, Indonesia: https://royalsocietypublishing.org/rsif/article/22/232/20250445/356293 - Incorporating human mobility to enhance epidemic response and estimate real-time reproduction numbers: https://journals.plos.org/ploscompbiol/article?id=10.1371%2Fjournal.pcbi.1013642 - Mitigating risks of malaria and other vector-borne diseases in the new capital city of Indonesia: https://www.nature.com/articles/s41467-024-54891-x --- 🎵 Music Credits Intro and outro music from Podcastle Stock Audio. Track: ‘Nairobi Nights’. License code: XUUYW3CDJ38EEGHM. --- ⚠️ Disclaimer This podcast is produced by the World Health Organization (WHO) as part of the Pandemic and Epidemic Intelligence Innovation Forum initiative: https://pandemichub.who.int/news-room/innovation-forum. The views expressed by guests are their own and do not necessarily represent those of WHO or its affiliates. Content is intended for informational purposes only and does not constitute professional medical advice. --- 📲 Listen and Subscribe The Innovation Forum AI Podcast is available on YouTube, Spotify, Apple Podcasts, and Amazon Music. You can find a written summary of this episode here: https://substack.com/@omorgan? Follow, rate, and share to help us reach more public health professionals exploring the future of AI.