Fiona Kastel leads the Data Innovations Group at 3ie and provides research, program management, and data analytics support for multiple programs and initiatives. She designs and conducts impact evaluations and supports evidence synthesis products, including conducting a geospatial impact evaluation of an agricultural intensification program in Niger, and an impact evaluation of the European Bank for Reconstruction and Development’s COVID-19 Solidarity Package. Prior to joining 3ie, Fiona worked on projects studying crime and resilience in Trinidad and Tobago and education and upward mobility in the U.S. She has a Masters in Public Affairs, specializing in Data Analysis, from Brown University and Bachelor of Science in Quantitative Analysis of Markets and Organizations with emphases in Cognitive Science and Finance from the University of Utah.

Topic: The International Initiative for Impact Evaluation (3ie) is advancing the use of artificial intelligence to strengthen development evidence. Our Data Innovations Group is testing machine learning and foundational AI models to overcome persistent challenges in impact evaluations, such as incomplete ground data, fragile environments, and hard-to-measure outcomes.

Through case studies in Sierra Leone, Guinea, Bangladesh, and Malawi, we demonstrate how AI methods, ranging from support vector machines and logistic regression classifiers to convolutional neural networks and SAMGeo, can reliably detect agricultural and aquaculture changes from satellite imagery. These innovations enable us to measure crop yields, fishpond expansion, and food security outcomes with greater accuracy and at scale, supplementing and enhancing traditional impact evaluations.

By embedding AI techniques into rigorous evaluation designs, 3ie is helping policymakers and funders answer pressing questions about what works in development, with faster, more granular, and cost-effective evidence.