Public response to the coronavirus outbreak is giving researchers insight into panic diffusion dynamics..
A trio of researchers at the Luddy School of Informatics, Computing, and Engineering are using machine learning techniques to create models of COVID-19 panic diffusion dynamics while comparing various panic diffusion models at different stages of disease transmission and awareness.
Associate Professor of Information Science Xiaozhong Liu, Professor of Informatics Johan Bollen, and Associate Professor of Informatics and Computing Yong Yeol Ahn are leveraging purchasing data of millions of ecommerce users over the past decade to help characterize people’s panic status. Diffusion models can highlight the rate at which panic occurs in people and how that panic manifests in their actions. Knowing how quickly various populations react to developments can create a range of insights.
“We also can estimate the panic reason and model panic diffusion patterns across different communities,” Liu said. “We’ve collected data on 94 million active ecommerce users and their purchase logs for the past 10 years. Using this data, we can tell or estimate their demographic information, such as their gender, age, income, education and geolocation, to generate a very large social network. Using that data, we can build a sophisticated machine learning and statistical model to generalize the panic statuses of users.”
The study, which is believed to be the pioneering investigation to model panic diffusion using massive ecommerce data, will use machine learning to create panic diffusion models at various levels of panic, locations, and communities, and will provide generalizable insights for different research domains, such as psychology, sociology, public health, complex networks, and information science. A coronavirus case study can also be a crucial and urgent opportunity to investigate public panic control, public health policy decision-making, and disaster management. It further can help policy makers locate vulnerable populations and communities in a pandemic, allowing those areas to receive valuable assistance.
The team has previously studied collective emotional reactions during natural disasters and expanding the research to focus on the COVID-19 pandemic was a natural progression.
“In an increasingly connected world, handling mass fear, anxiety, and misinformation in a public health emergency can be as important as controlling the crisis itself,” Liu said. “During a pandemic, material and physical resources are stretched thin, and, often, the wild spread of fear and panic can be fatal for vulnerable populations and communities. We hope this study can help to model the panic diffusion in different communities to allow policy makers to provide targeted assistance.”
Future studies of the same data could be used to track and estimate other chronic diseases, such as depression and diabetes.
“The wealth of data that is generated in our everyday lives can provide tremendous insight into the behavior of people and make a real-world impact on public policy,” said Dennis Groth, interim dean of the Luddy School. “Studies such as this really showcase the powerful information that can be gleaned from raw data, and our researchers at the Luddy School are doing a fantastic job of using their expertise to make observations that can make a difference.