An Exploratory Analysis of COVID Bot vs Human Disinformation Dissemination stemming from the Disinformation Dozen on Telegram

TL;DR

The paper investigates how disinformation attributed to a well-known Disinformation Dozen spread on Telegram, comparing three user groups: the Disinformation Dozen, bot users, and human users, using a large Telegram dataset collected from January to June 2023. It combines temporal, topical, and network analyses to show that the Disinformation Dozen initiate disinformation but are not the primary spreaders; bots sustain engagement through replies, while humans propagate content through forwarding. A Telegram-adapted bot-detection method (based on BotBuster) is developed and validated against manual annotations, achieving substantial agreement, and a novel one-hop snowball data collection captures forwarding and discussion dynamics across thousands of channels. The study highlights platform-specific diffusion patterns on Telegram, revealing that humans act as bridging forwards and bots as active conversational participants, with implications for designing targeted countermeasures in mobile messaging environments.

Abstract

The COVID-19 pandemic of 2021 led to a worldwide health crisis that was accompanied by an infodemic. A group of 12 social media personalities, dubbed the ``Disinformation Dozen", were identified as key in spreading disinformation regarding the COVID-19 virus, treatments, and vaccines. This study focuses on the spread of disinformation propagated by this group on Telegram, a mobile messaging and social media platform. After segregating users into three groups -- the Disinformation Dozen, bots, and humans --, we perform an investigation with a dataset of Telegram messages from January to June 2023, comparatively analyzing temporal, topical, and network features. We observe that the Disinformation Dozen are highly involved in the initial dissemination of disinformation but are not the main drivers of the propagation of disinformation. Bot users are extremely active in conversation threads, while human users are active propagators of information, disseminating posts between Telegram channels through the forwarding mechanism.

Figures (7)

• Figure 1: Data Collection Pipeline
• Figure 2: User Group Identification Pipeline
• Figure 3: Overview of Methodology and Brief Results
• Figure 4: Temporal Post Frequency per User Group. The Disinformation Dozen posts least frequently, while Humans post most frequently. Bots and humans post with similar frequency patterns, while the Disinformation Dozen posts the least.
• Figure 5: Comparison of linguistic cues among the three user group