Presidential communications on Twitter during the COVID-19 pandemic: mediating polarization and trust, moderating mobility

Crisis communication from government officials requires carefully constructed strategies to maintain public trust, reduce uncertainty and improve compliance with non-pharmaceutical interventions during pandemics and other crisis-like scenarios. At the same time, communication can reduce or increase the polarization of public sentiment toward government actions. While data on government communications and public social media is widely available, the means to process the vast range of possible opinionated groups, populations, and themes are complex and computationally demanding. In this study, we investigate a system model consisting of novel measures to measure public trust and emotional polarization, government officials’ communications, and physical outcomes related to the COVID-19 pandemic in terms of compliance with restrictions, public health, and forward-looking economic conditions. We present methods for measuring emotional polarization in public Twitter discourse by constructing components with reduced dimensionality from the numerous sentiments in the unstructured text. We apply the methods to two months of data from the U.S.A. using the different states as separate subsystems. Our results show that the Presidential Communications on Twitter mediate the relationship between the pandemic public health outcomes with social media polarization and public trust. In contrast, some results still need to be more conclusive within the two-month dataset used in the evaluation. Most notably, polarization varies between user populations in different states.