Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Retweets, Receipts, and Resistance: Discourse, Sentiment, and Credibility in Public Health Crisis Twitter

Tawfiq Ammari, Anna Gutowska, Jacob Ziff, Casey Randazzo, Harihan Subramonyam

TL;DR

This study analyzes 275,124 COVID-19–related tweets surrounding the CDC to understand discourse, credibility, and engagement on Twitter. Using a mixed-methods pipeline (BERTopic for 71 topics, VADER sentiment, Iffy-index credibility, ERGM network modeling, and SHAP explainability), it reveals a largely top-down CDC presence with highly polarized public discourse around vaccines and masking. Rich media and verified accounts amplify content, including low-credibility sources, creating a fragmented information environment and limited two-way engagement. The authors propose a CDC AI Assistant and long-term participatory design and policy measures to foster trust, counter misinformation, and adapt messaging across diverse audiences during evolving health crises.

Abstract

As the COVID-19 pandemic evolved, the Centers for Disease Control and Prevention (CDC) used Twitter to disseminate safety guidance and updates, reaching millions of users. This study analyzes two years of tweets from, to, and about the CDC using a mixed methods approach to examine discourse characteristics, credibility, and user engagement. We found that the CDCs communication remained largely one directional and did not foster reciprocal interaction, while discussions around COVID19 were deeply shaped by political and ideological polarization. Users frequently cited earlier CDC messages to critique new and sometimes contradictory guidance. Our findings highlight the role of sentiment, media richness, and source credibility in shaping the spread of public health messages. We propose design strategies to help the CDC tailor communications to diverse user groups and manage misinformation more effectively during high-stakes health crises.

Retweets, Receipts, and Resistance: Discourse, Sentiment, and Credibility in Public Health Crisis Twitter

TL;DR

This study analyzes 275,124 COVID-19–related tweets surrounding the CDC to understand discourse, credibility, and engagement on Twitter. Using a mixed-methods pipeline (BERTopic for 71 topics, VADER sentiment, Iffy-index credibility, ERGM network modeling, and SHAP explainability), it reveals a largely top-down CDC presence with highly polarized public discourse around vaccines and masking. Rich media and verified accounts amplify content, including low-credibility sources, creating a fragmented information environment and limited two-way engagement. The authors propose a CDC AI Assistant and long-term participatory design and policy measures to foster trust, counter misinformation, and adapt messaging across diverse audiences during evolving health crises.

Abstract

As the COVID-19 pandemic evolved, the Centers for Disease Control and Prevention (CDC) used Twitter to disseminate safety guidance and updates, reaching millions of users. This study analyzes two years of tweets from, to, and about the CDC using a mixed methods approach to examine discourse characteristics, credibility, and user engagement. We found that the CDCs communication remained largely one directional and did not foster reciprocal interaction, while discussions around COVID19 were deeply shaped by political and ideological polarization. Users frequently cited earlier CDC messages to critique new and sometimes contradictory guidance. Our findings highlight the role of sentiment, media richness, and source credibility in shaping the spread of public health messages. We propose design strategies to help the CDC tailor communications to diverse user groups and manage misinformation more effectively during high-stakes health crises.

Paper Structure

This paper contains 55 sections, 5 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Dissemination Patterns
  3. Message Content, Rumor, and Trust
  4. Public's experience
  5. Related Work
  6. Twitter affordances and audience management
  7. Imagined Audiences Built with Hashtags
  8. Imagined Audiences Built with Visible Mentions
  9. Twitter Propagation Affordances: Retweets, Likes, Responses, and Quote-tweets
  10. Rich Media, Verified Users, and Misinformation on Twitter
  11. Rich Media on Twitter
  12. Verified Twitter accounts
  13. Misinformation and Distrust
  14. Digital Communication in Public Health Crises
  15. Emerging Infectious Diseases: Rolling uncertainty, risk, and public relations
  16. ...and 40 more sections

Figures (5)

  • Figure 1: Topic groupings into themes
  • Figure 2: This figure shows the sentiment distribution for two highly propagated topics—Topic 0 (vaccination discourse) and Topic 26 (masking in schools). Both exhibit bimodal sentiment distributions, indicating strong polarization. Positive sentiment clusters align with public health advocacy, neutral clusters contain logistical or institutional information, and negative clusters reflect misinformation, conspiracy theories, and ideological resistance.
  • Figure 3: This figure displays a CDC-generated infographic highlighting updated school safety protocols in response to the emergence of the Delta variant. It reflects a shift in guidelines recommending continued masking indoors, even for vaccinated individuals, representing what many saw as a departure from earliery messaging.
  • Figure 4: This figure juxtaposes two public-facing CDC messages: one (left) suggesting that vaccinated individuals no longer need to wear masks, and the other(right) provided a more nuanced recommendation masking post-vaccination in different contexts. Both were used as "receipts" by critics of the CDC (left) questioning CDC credibility and consistency and defenders of the original guidelines (right) arguing that the evolving COVID situation required nuance, which was present in CDC guidelines.
  • Figure 5: Mention co-occurrence network showing top accounts mentioned along with the CDC account.