Studying differential COVID-19 responses to trustworthy information will contribute new perspectives and improve our fundamental understanding of the human aspects of information trustworthiness during a pandemic. Although many U.S. residents are heeding restrictive COVID-19 orders and warnings from both official guidance and public health information sources, some individuals are not following official guidance. Simultaneously, the pandemic has spread through major U.S. cities and states in an uneven manner. Moreover, COVID-19 messaging is complicated by evolving guidance and contagion information, and media framing complicates information transmission, shifting public understanding of official guidance. Understanding and evaluating how a population is reacting to these multiple sources of information is a crucial scientific challenge with implications for improving future public policy and more effective crisis mitigation. To better understand differential population responses to this crisis, this interdisciplinary research project will gather Twitter data in Louisiana and Washington State to analyze differential sentiment and language use in responses to health information and public guidance, across assumed age and linked geographic conditions. It will also assess how these differences change given time, location, health, economic, social conditions, and sources of information. Assessing the systematic variation of sentiment and language use across age and then across time, space, context, and information sources will provide vital insight to targeting effective public messaging by public health and government officials during a crisis.
This project will collect three waves of Twitter and associated data to study differential sentiment and language use in responses to health information and public health guidance concerning COVID-19. This interdisciplinary project seeks to advance knowledge about how disparities in information consumption, as represented by sentiment and language use, vary across sources of information and guidance, across time, geography, socio-demographic characteristics, and health conditions. This project will compare sentiment and language use patterns among a population of Twitter users based on the content of their tweets, and analyze how these characteristics evolve through the recognition of the emergency, the peak of the crisis, and the mitigation of the pandemic in the U.S. The project will use computational methods to understand sentiment patterns and language use on Twitter and link tweets and relevant entities to corresponding longitudinal data about trustworthy information sources, including news media sources and official emergency guidance, policies, and orders. The tweets and entities will be classified for relevance (individual, organization, or bot) with derived age and location. Using timestamps and derived location, the project will associate tweets with daily disease-specific rates and annual demographic and socio-economic information. Data about the surrounding information context, including reporting trends and framing of the crisis, sources of health information, and information-seeking behavior about COVID-19 will provide tools for assessing the validity and reliability of our inferences about the patterns of sentiment and language use. Project findings will contribute to the literature on computational methods, disaster communication, misinformation during crisis events, and disaster resilience, informing future guidance on policy and emergency messaging during a crisis.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
