From Large-scale Mobility Data to Behavioral Intelligence
Friday, April 17, 2026 · 11:30 AM – 12:00 PM · Auditorium 2
The widespread adoption of mobility sensors has enabled, for the first time, continuous and unobtrusive observation of offline human behavior at population scale. These data extend decades of online traces such as clickstream and social media, enabling a 360-degree view of human behavior and social interaction. This shift is transforming how we study and act on problems of immediate societal and economic importance, including epidemic dynamics, health outcomes, disaster resilience, urban systems, consumer behavior, and inequality in resource access. However, realizing this potential in practice remains challenging: mobility data are noisy, sparse, and privacy-sensitive. Extracting stable, interpretable behavioral structure from large-scale trajectories is nontrivial.
I present a research-driven, end-to-end framework for transforming large-scale location data into actionable behavioral intelligence. Drawing on systems and studies leveraging billions of mobility records, I will cover: • Data engineering and pipeline design: processing data for robust downstream modeling • Behavioral representation learning: combining clustering, sequence modeling, network analytics, and NLP to uncover latent behavioral structure • Population-scale pattern discovery: identifying stable behavioral regimes across heterogeneous populations • Causal inference in observational settings: leveraging natural experiments to estimate behavioral responses to shocks and interventions • Privacy-preserving analytics: designing interpretable methods that balance data utility with individual privacy risk
These techniques are illustrated through real-world applications, including disaster response, cultural influences on collective action, privacy behavior shift during crises, and business applications such as measuring disintermediation on two-sided platforms. Across these settings, the framework enables measurable improvements in behavioral prediction, intervention targeting, and decision-making in public health, policy, and business contexts.
This talk is designed for ML engineers, data scientists, and applied researchers working with large-scale behavioral or spatiotemporal data; basic familiarity with machine learning is assumed. Attendees will leave with concrete techniques for mobility data modeling, reusable pipeline and system design patterns, rigorous causal evaluation strategies, practical lessons from real-world deployments, and a roadmap of rich future research directions.
About the Speaker
Natasha Foutz
McIntire School of Commerce, University of Virginia
Professor Foutz is an expert in AI-powered entertainment marketing and mobility intelligence. Her research leverages big data through machine learning and econometric, statistical, and experimental methods. Her work is featured in books and top-tier journals, and she serves as an area editor and editorial board member for several academic publications. Professor Foutz teaches marketing analytics, entertainment marketing, and marketing models across undergraduate, MBA, EMBA, and PhD programs. She has received numerous honors for her research, teaching, and service, including Best Paper Awards, the Mallen Award for Lifetime Contribution to Motion Picture Studies, the UVA Outstanding Researcher Award, the Best Area Editor Award, the Meritorious Service Award, the Springer Nature Author Service Award, and the UVA All-University Teaching Award.