Brandon is a senior research methodologist in the Methodology & Quantitative Social Sciences department at NORC at the University of Chicago. A sociologist and data scientist by training, he focuses on what it takes to make data and AI systems trustworthy in public-sector institutions. He identifies where trust breaks down and builds the methods, governance, and infrastructure that earn it. With more than 14 years of experience applying computational methods to social science problems, he evaluates AI systems not only on technical performance but on whether they hold up to the institutional contexts, measurement standards, and communities that depend on them. Brandon is one of the primary designers of NORC’s AI governance framework.

Brandon's research examines how AI systems perform in data collection and statistical settings. His work has shown that large language models do not respond to surveys like human survey respondents. He has also documented how bias-detection tools can introduce data quality issues of their own when engineering choices made downstream of the statistical design reintroduce the bias the method was meant to control for. On the applied side, Brandon has built an LLM-based detector that identifies AI-generated and fraudulent survey responses with greater than 98 percent accuracy, and he has explored conversational AI agents that dynamically probe survey respondents and code their answers in real time, with experimental findings forthcoming in Survey Research Methods. He is Project Director for Measuring Large Language Model Understanding of Federal Statistical Data, a National Secure Data Service Demonstration project conducted with the National Center for Science and Engineering Statistics and the U.S. Department of Commerce. This work assesses whether large language models can reliably comprehend federal statistical data and informs how agencies make their data AI-ready.

Brandon's data infrastructure work focuses on making research data assets secure, accessible, interoperable, and governable. He has led four projects through America's Datahub Consortium that contribute to the design of a future National Secure Data Service, including the Federated Data Usage Platform, a prototype intended as a shared service for tracking how federal data are used. He serves as Infrastructure Lead and Chief Statistician for a national infrastructure effort for the Republic of Palau, integrating fragmented systems across education, labor, health, justice, and social services. He participates in the UNECE High-Level Group for the Modernisation of Official Statistics' AI-Ready Dissemination project, working on what it takes to make official statistics legible, reliable, and trustworthy when AI mediates access to them. These systems are how federal statistical agencies and national governments turn administrative records, survey data, and other fragmented sources into evidence policymakers can act on.

Brandon’s work has been published in peer-reviewed journals and conference proceedings across disciplines and supported by awards from the National Science Foundation, the National Institutes of Health, a Fulbright fellowship, the Government of France, and the Countway Library of Medicine (Harvard University / Boston Medical Library). He currently serves as Chair of the American Statistical Association’s Section on Text Analysis and has held advisory roles with the NIST Generative AI Public Working Group, the Data Foundation’s AI Working Group, AcademyHealth, and the UNECE High-Level Group on the Modernisation of Official Statistics. He has been elected to multiple section councils of the American Sociological Association and was previously assistant editor for the American Sociological Review. Brandon has served on program committees of leading conferences including the Conference on Empirical Methods in Natural Language Processing (EMNLP) and Widening NLP, and speaks regularly about where AI fits in public-sector research, where it fails, and what it takes to build systems that maintain trust.