Common Ground in Crisis: Causal Narrative Networks of Public Official Communications During the COVID-19 Pandemic

• In Event: 40221 - Section on Mathematical Sociology Flash Talks
  In Refereed Roundtable (90 minute): Table 4: Common Ground in Crisis: Causal Narrative Networks of Public Official Communications During the COVID-19 Pandemic

Abstract

Public communication surrounding the COVID-19 pandemic has been argued to be plagued by a combination of an overabundance of information, misinformation, and mixed messages that proved costly to the public in terms of time, funds and, more importantly, preventable deaths. During periods of threat, the public looks to officials to understand the situation, its causes, and its ramifications. A strategy officials may employ in their communications is the usage of causal narratives, or structured narratives that discuss an event or entity (a cause) and its impact (an effect). As causal narratives supply a highly condensed, structured method of imparting information to audiences, they are particularly well-suited in relaying essential, clarifying, potentially life-saving information through terse messaging platforms. However, their usefulness may be moderated by the welter of messages from different agencies, possibly leading individuals to encounter an aggregate narrative (i.e., discourse) that is incomplete, incoherent, or inconsistent.

This paper examines the causal narratives utilized by organizations in public communications during the first fifteen months of the COVID-19 pandemic, reconstructing the discourse of claims articulated by responding agencies to the public and examining the degree to which it provided clear, consistent, and actionable information regarding the evolving threat. We utilize large language models (LLMs) to extract causal narratives from a massive corpus of approximately 900,000 Twitter messages from officials, evaluating the accuracy of LLMs for extracting complex semantic features and experimenting with methodologies of improving LLM accuracy. We take on a relational, network approach to this work by constructing a semantic network of causal narrative concepts that represents a broader discourse. We examine the structure of the semantic causal narrative network, evaluate the consistency of causal narrative usage across different entities and across time, and examine the effect of differing narrative structures on public engagement with messages containing causal narratives.

Authors

• Sabrina Mai, University of California-Irvine

• Scott Leo Renshaw, Carnegie Mellon University

• Jeannette Sutton, University of Kentucky

• Carter T. Butts, University of California-Irvine