Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Morbid Realities of Social Media: An Investigation into the Misinformation Shared by the Deceased Victims of COVID-19

Hussam Habib, Rishab Nithyanand

TL;DR

The paper tackles the real-world harms of COVID-19 misinformation by analyzing Facebook posts from victims who propagated misinformation before dying. It introduces a data-driven workflow combining OCR-extracted text from about 20.8K posts, a FastText-based embedding with TF-IDF weighting, and clustering to $k=44$ groups to identify 14 misinformation themes, complemented by entity-sentiment analysis and source characterization. The study finds politicization, dominated by right-wing elites and biased external domains, as the primary driver of these narratives, while Facebook's soft moderation was only 6% complete yet showed consistency across similar content. These findings highlight platform responsibility and suggest policy and design changes to reduce the spread and impact of deadly misinformation in future infodemics.

Abstract

Social media platforms have had considerable impact on the real world especially during the Covid-19 pandemic. Misinformation related to Covid-19 might have caused significant impact on the population specifically due to its association with dangerous beliefs such as anti-vaccination and Covid denial. In this work, we study a unique dataset of Facebook posts by users who shared and believed in Covid-19 misinformation before succumbing to Covid-19 often resulting in death. We aim to characterize the dominant themes and sources present in the victim's posts along with identifying the role of the platform in handling deadly narratives. Our analysis reveals the overwhelming politicization of Covid-19 through the prevalence of anti-government themes propagated by right-wing political and media ecosystem. Furthermore, we highlight the failures of Facebook's implementation and completeness of soft moderation actions intended to warn users of misinformation. Results from this study bring insights into the responsibility of political elites in shaping public discourse and the platform's role in dampening the reach of harmful misinformation.

The Morbid Realities of Social Media: An Investigation into the Misinformation Shared by the Deceased Victims of COVID-19

TL;DR

The paper tackles the real-world harms of COVID-19 misinformation by analyzing Facebook posts from victims who propagated misinformation before dying. It introduces a data-driven workflow combining OCR-extracted text from about 20.8K posts, a FastText-based embedding with TF-IDF weighting, and clustering to groups to identify 14 misinformation themes, complemented by entity-sentiment analysis and source characterization. The study finds politicization, dominated by right-wing elites and biased external domains, as the primary driver of these narratives, while Facebook's soft moderation was only 6% complete yet showed consistency across similar content. These findings highlight platform responsibility and suggest policy and design changes to reduce the spread and impact of deadly misinformation in future infodemics.

Abstract

Social media platforms have had considerable impact on the real world especially during the Covid-19 pandemic. Misinformation related to Covid-19 might have caused significant impact on the population specifically due to its association with dangerous beliefs such as anti-vaccination and Covid denial. In this work, we study a unique dataset of Facebook posts by users who shared and believed in Covid-19 misinformation before succumbing to Covid-19 often resulting in death. We aim to characterize the dominant themes and sources present in the victim's posts along with identifying the role of the platform in handling deadly narratives. Our analysis reveals the overwhelming politicization of Covid-19 through the prevalence of anti-government themes propagated by right-wing political and media ecosystem. Furthermore, we highlight the failures of Facebook's implementation and completeness of soft moderation actions intended to warn users of misinformation. Results from this study bring insights into the responsibility of political elites in shaping public discourse and the platform's role in dampening the reach of harmful misinformation.
Paper Structure (16 sections, 3 figures, 3 tables)

This paper contains 16 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Data Collection
  3. Narratives in COVID-19 Misinformation
  4. Covid-19 misinformation themes
  5. Covid-19 misinformation entities
  6. Takeaways
  7. Sources of misinformation
  8. Public figures as sources of misinformation
  9. External domains as sources of misinformation
  10. Takeaways
  11. Moderation of Misinformation
  12. Completeness of interventions
  13. Consistency of interventions
  14. Takeaways
  15. Related Works
  16. ...and 1 more sections

Figures (3)

  • Figure 1: Representation of the most common progressions through narratives along with the percentage of users with similar progression. Colors of the nodes represent the cluster each topic belongs to based on the Louvian community analysis.
  • Figure 2: Percentage of posts with external domains in each topic category along with average bias and factual reporting score of the domains present. The size of the bubble is proportional to the percentage of posts with the topic containing an external domain.
  • Figure 3: Percentage of misinformation labels present in each topic.