Social Internet Research

Quantifying model uncertainty in agent-based simulations for forecasting the spread of infectious diseases and understanding human behavior using social media

Get Expertise  

  • thumbnail of Sara Del Valle
  • Principal Investigator
  • Sara Del Valle
  • (505) 665-9286
  • Email
Providing decision support to policy makers on consequence, mitigation, and response to the risks of infectious diseases.

Overview

An emerging epidemiological observation tool is the social Internet. That is, people use systems like Twitter and Facebook to share perceptions, desires, and concerns about daily life, and they use systems like Wikipedia and Stack Exchange to learn more about topics of interest. These activities leave traces, which we use to quantify disease spread and to understand disease-relevant human behavior (e.g., facemask use or vaccine acceptance), including the effectiveness of interventions. Our goal is to use this information as input to our models to better simulate disease dynamics in the presence of realistic human behavior.

For example, we asked whether facemask mentions on Twitter correlate with the influenza-like illness rate published by the Center for Disease Control and Prevention. They in fact do. The figure below shows tweet volumes from September 9, 2009 to May 6, 2010, where medical facemasks are mentioned and observed alongside the CDC ILI rate. The correlation is 0.92 and 0.90, respectively.

Graph showing Tweet volumes where facemasks are mentioned

Tweet volumes from September 9, 2009 to May 6, 2010, where medical facemasks are mentioned and observed alongside the CDC ILI rate.