Title: Detecting Communities in and the Evolutionary Structure of Google Search Trends Related to COVID-19

Abstract:

COVID-19 has substantially altered the ways in which people interact, both in person and online. But people’s views on COVID-19 have not been stable since the outbreak of the virus. Using a suite of open source code and tools, I mine Google Trends web-search data to understand how people think about COVID-19 and the coronavirus. More specifically, I fit a variety of network models and also compare three major community detection algorithms (edge betweenness, label propagation, and greedy optimization of modularity) to mine queries related to the initial search for “coronavirus”, from the first Google search in the US (January 16, 2020) to the most recent at the time of writing (July 30, 2020). In addition to exploratory data analysis, network results across a variety of time intervals suggest people’s Google searches over the development of this pandemic have significantly and constantly shifted as more information is daily made available. For example, common related queries at the outset were “China” and whether “cocaine kills” coronavirus, whereas in March in the United States common related queries were “stimulus check” and in April included “Tom Hanks” across several communities.

More recently related queries have evolved to include “thanking” healthcare workers and specific hotspot “counties” like “Dallas” and “Sonoma”. In sum, the tone and structure of the search space beyond merely “coronavirus” has significantly evolved as the pandemic itself and its impact on society has also evolved. These results suggest that to better understand COVID-19 and the widespread social effects, not only does the massive repository of open source technology help, but also these tools should be leveraged to continue digging deeper into this ever-evolving global crisis. And in the spirit of open science, all code is housed at the author’s Github.