PROJECT SUMMARY/ ABSTRACT In the midst of the COVID-19 pandemic, a parallel `infodemic,' an abundance of reliable information and inaccurate misinformation, persists. There has also been a significant increase in misinformation exchange and consumption, largely on social media platforms, which threatens individual and public health. An important challenge remains to develop strategies to detect trusted and accurate `signals' amidst dynamic misinformation `noise.' This misinformation contributes to confusion, distrust, and distress around health behaviors such as vaccination, mask wearing, and social distancing. The racial disparities in morbidity, mortality, social, and economic consequences of COVID-19 are well documented; less studied are variations in the information- seeking and COVID-19 health decision-making specific to Black and rural communities. Public health information and campaigns have traditionally relied on theory-based surveys or interview methods to measure knowledge and attitudes to design health messaging. Rapid expansion of social media use and parallel advances in machine learning analytics provide a unique opportunity to track public views, knowledge, and attitudes simultaneously to translate novel analytic insights into precision public health communication with an intentional lens on Black and rural communities. This proposal aims to build a computational framework to uncover heterogeneity in attitudes and misinformation exposure towards COVID- 19 vaccination, model predictors of highly engaging and persuasive messages (including sources, linguistic choices, and content); and to use pragmatic qualitative methods to understand individual response to social media misinformation with a specific lens on race (Black and white individuals) and location (rural and urban). While we focus our message development process on COVID-19 vaccination as a timely and critical behavior, and compare targeting across four specific audiences (Black rural residents, white rural residents, Black urban residents, and white rural residents), our approach is highly adaptable across health topics and scalable to a number of precision-targeted audiences. We see a need for flexible and nimble methods for rapid, human-centered content generation that supports accurate, equitable, and effective precision public health messaging. Computational tools powered by machine learning, predictive analytics, and natural language processing married with patient-centered qualitative methods offer a powerful synergy to conventional approaches to public health campaigns to identify and combat misinformation. The findings from this study will directly inform broader public health action and future strategies so that they can be deployed in the current pandemic and in ongoing efforts to address racial disparities in chronic diseases, HIV, cancer, maternal mortality, and mental health.

PROJECT NARRATIVE Effective messaging to counter misinformation and encourage healthy behaviors, including COVID-19 vaccination, must accommodate differences in preferences, habits, and behaviors across minority populations experiencing health disparities. This proposal aims to build a computational framework to measure heterogeneity in information exposure and engagement during COVID-19; to model the predictors of highly engaging and persuasive messages (including sources, linguistic choices, and content); and investigate the response to social media misinformation through pragmatic qualitative methods to counter misinformation and promote healthy behaviors. While we focus our proposal on COVID-19 vaccination and compare misinformation exposure and attitudes based on race (Black and white) and urban vs. rural populations; the approach is adaptable and will establish a nimble and responsive paradigm for precision public health messaging beyond COVID-19.