Detection and Analysis of Interrupted Behaviors by Public Policy Interventions during COVID-19

Guimin Dong, Lihua Cai, Shashwat Kumar, Debajyoti Datta, Laura E. Barnes, Mehdi Boukhechba

Abstract: In most countries around the world, various public policies and guidelines, such as social distancing and stay-at-home orders, have been put in place to slow down the spreading of COVID-19. Relying on traditional surveys to assess policy impacts on community level behavior changes may lead to biased results, and limit fine-grained understanding of human behavior dynamics over time. We propose to leverage mobile sensing to capture people's behavior footprints amid the COVID-19 pandemic, and understand their collective behavior changes with respect to existing policies. Specifically, we propose to extract a rich set of behavioral markers from raw mobile sensing data, including mobility, social interactions, physical activities, and health states, and apply them in a generalized behavior change analysis framework to measure and detect community level behavior changes in an epidemic context. We present how to combine change point detection algorithm and interrupted time series analysis to automatically detect three different measurements of behavior changes (e.g., level, trend, and variance changes), and provide insights supported by statistical inference. A case study using a dataset that we collected from a large mobile sensing study conducted in the United States is shown to demonstrate the proposed framework and method.