A Second Pandemic? Analysis of Fake News about COVID-19 Vaccines in Qatar

Preslav Nakov, Firoj Alam, Shaden Shaar, Giovanni Da San Martino, Yifan Zhang

Introduction

As the world grappled with the COVID-19 pandemic, a secondary "infodemic" of misinformation threatened to undermine public health efforts, particularly regarding vaccine rollouts. This paper investigates this phenomenon within the unique multilingual context of Qatar, analyzing how misinformation, propaganda, and media framing differ across Arabic and English social media discourse.

Analyzing the Infodemic Pipeline

The researchers developed a comprehensive computational pipeline to dissect tweets from three distinct angles:

  1. Disinformation Analysis: Using a Support Vector Machine (SVM) classifier, the team evaluated tweets for verifiable claims, potential harm, and the need for fact-checking.
  2. Propaganda Detection: The study employed two specialized tools: Proppy, for overall propagandistic scoring, and Prta, for identifying 18 specific techniques like "Loaded Language" or "Fear-mongering."
  3. Framing Analysis: Utilizing the Tanbih system (a fine-tuned BERT model), tweets were categorized into 15 media frames, such as Economy, Health and Safety, or Morality.
Examples of tweet categorizations used in the study, including rumors, actions taken, and panic-inducing content.

Data and Linguistic Divergence

The study analyzed a corpus of 606 Arabic and 2,589 English tweets originating from Qatar between February 2020 and March 2021. The analysis revealed a stark contrast in how different linguistic communities engaged with vaccine information.

The Factuality vs. Propaganda Paradox

One of the most significant findings was the divergence between factuality and persuasion. English tweets were generally more "factual"—containing verifiable claims—yet they were significantly more propagandistic (approx. 33%) compared to Arabic tweets. This suggests that English-speaking users often leverage true facts to advance specific, biased agendas.

Framing and Concerns

The "angle" of the conversation also shifted by language. Arabic tweets, likely reflecting the concerns of Qatari locals, focused heavily on Health and Safety. In contrast, English tweets, likely from the large expatriate population, were dominated by Economic concerns, highlighting anxieties over job security and financial stability.

A typical health-related rumor found in social media discourse, claiming garlic water can cure COVID-19.

Key Results

  • Disinformation Prevalence: Nearly 50% of the Arabic tweets contained false information or rumors. English tweets, while more factual, often utilized humor or directed blame toward authorities.
  • Propaganda Techniques: In Arabic, "Doubt" was the primary persuasive tool. English users employed a wider array of techniques, including loaded language, exaggeration, and fear-mongering.
  • Model Performance: Automated classifiers achieved F1 scores between 0.41 and 0.73, illustrating the ongoing challenge of fact-checking short-form, informal social media content.

Limitations and Future Outlook

While providing valuable insights, the study notes limitations such as potential translation bias when analyzing Arabic propaganda through English-centric tools and the relatively small size of the Arabic corpus. Future research aims to expand this framework to "closed" platforms like WhatsApp and Facebook and to include more languages to determine if these linguistic patterns are global.

Summary updated

BibTeX
@inproceedings{nakov2021second,
  author = {Preslav Nakov and Firoj Alam and Shaden Shaar and Giovanni Da San Martino and Yifan Zhang},
  title = {{A Second Pandemic? Analysis of Fake News about COVID-19 Vaccines in Qatar}},
  booktitle = {RANLP},
  year = {2021},
  url = {https://openreview.net/forum?id=amABgm5Cfq0},
}
Figure