Bots and Misinformation Spread on Social Media: Implications for COVID-19 (Preprint)

McKenzie Himelein-Wachowiak, Salvatore Giorgi, Amanda Devoto, Muhammad Mahbubur Rahman, Lyle Ungar, H. Andrew Schwartz, David H. Epstein, Lorenzo Leggio, Brenda Curtis

TL;DR

This paper investigates the role of social bots in spreading misinformation, particularly concerning COVID-19. It reveals that a significant portion of known Twitter bots are actively discussing COVID-19, raising concerns about their potential impact on public health and pandemic outcomes.

What Problem Does This Paper Address?

The COVID-19 pandemic has been accompanied by an unprecedented surge in online information, often referred to as an "infodemic." This includes a proliferation of misinformation, conspiracy theories, and "fake news." While human susceptibility to misinformation is well-documented, the role of automated accounts, or "bots," in amplifying and spreading this content, especially during critical public health events, is a significant concern. This paper aims to shed light on how bots contribute to the spread of misinformation, their origins, and, crucially, to quantify their involvement in the COVID-19 infodemic.

Methods

The authors employ a two-pronged approach:

  • Scoping Review: They conduct a comprehensive review of existing literature on social bots, covering their definitions, technical features that distinguish them from humans, and methods for their detection using machine learning. They also examine strategies bots use to spread information and explore their potential origins.
  • Secondary Analysis: They perform a quantitative analysis of publicly available datasets of known Twitter bots. Using the Twitter API, they retrieve tweets from these bots from 2020 and analyze the extent to which they discuss COVID-19, using a set of 15 COVID-19-related keywords.

Datasets

The study utilizes several publicly available datasets of known Twitter bots, sourced from the "Bot Repository." These datasets were originally identified through various methods and span from 2011 to 2019. The authors do not introduce a new benchmark dataset but rather leverage existing ones for their secondary analysis.

Key Insights and Ideas

  • Bots are not just spammers: The paper highlights that not all bots are malicious; some serve benign purposes (e.g., providing information) or even social good. However, a significant subset are designed to spread misinformation or engage in coordinated campaigns.
  • Sophistication of Social Bots: Social bots are evolving beyond simple spambots, employing more human-like behaviors and messaging to blend in and influence online discourse. This makes their detection increasingly challenging.
  • Human-Bot Interaction is Key: The paper emphasizes that the effectiveness of bots in spreading misinformation is amplified by human susceptibility. Humans often struggle to distinguish bots from genuine accounts and can unknowingly contribute to the spread of false narratives.
  • COVID-19 is a Prime Target: The COVID-19 pandemic has become a fertile ground for bot activity, with a substantial percentage of known bots actively engaging in discussions related to the virus. This suggests that bots are strategically leveraging major global events.
  • The "Infodemic" is a Bot-Amplified Phenomenon: The rapid spread of misinformation during COVID-19 is not solely a human issue; bots play a crucial role in amplifying these narratives, potentially impacting public health decisions and behaviors.

Key Results

  • Significant COVID-19 Bot Activity: The secondary analysis found that between 53% and 66% of bots in the analyzed datasets were actively tweeting about COVID-19 in 2020.
  • Bot Strategies: Bots employ strategies like rapid retweeting, mentioning influential users, and using specific hashtags to maximize the reach of their content, including misinformation.
  • Detection Challenges: While machine learning algorithms show higher accuracy in bot detection compared to human annotators, there's a continuous "arms race" as bots become more sophisticated. Human annotators, for instance, had an F1-score of 0.57, while automated methods achieved up to 0.92.
  • Potential for Real-World Harm: The paper links bot-driven misinformation to real-world consequences, such as shortages of medications (e.g., hydroxychloroquine) due to false claims about their efficacy against COVID-19, and a decreased likelihood of engaging in protective behaviors.

Limitations and Future Work

  • Limitations of Data Access: The study acknowledges limitations due to Twitter's API restrictions, which meant that deleted or private tweets and accounts were not accessible, potentially affecting the completeness of the data.
  • Validity of Claims Not Assessed: The analysis focused on whether bots were discussing COVID-19, not on the validity of the claims they were spreading. Future work could investigate the specific types of misinformation being propagated.
  • Origin and Motive Ambiguity: Determining the precise origin and motives of social bots remains a challenging task.
  • Evolving Threat Landscape: The paper highlights the ongoing "arms race" between bot creators and detection methods. As bots become more intelligent and human-like, detection techniques will need to become more sophisticated. The authors suggest that future research should focus on developing more precise detection methods and understanding the evolving strategies of bot creators, especially in the context of future global events.

Summary updated

Abstract

<sec> <title>UNSTRUCTURED</title> As of March 2021, the SARS-CoV-2 virus has been responsible for over 115 million cases of COVID-19 worldwide, resulting in over 2.5 million deaths. As the virus spread exponentially, so did its media coverage, resulting in a proliferation of conflicting information on social media platforms—a so-called “infodemic.” In this viewpoint, we survey past literature investigating the role of automated accounts, or “bots,” in spreading such misinformation, drawing connections to the COVID-19 pandemic. We also review strategies used by bots to spread (mis)information and examine the potential origins of bots. We conclude by conducting and presenting a secondary analysis of data sets of known bots in which we find that up to 66% of bots are discussing COVID-19. The proliferation of COVID-19 (mis)information by bots, coupled with human susceptibility to believing and sharing misinformation, may well impact the course of the pandemic. </sec>
BibTeX
@article{himeleinwachowiak2021bots,
  author = {McKenzie Himelein-Wachowiak and Salvatore Giorgi and Amanda Devoto and Muhammad Mahbubur Rahman and Lyle Ungar and H. Andrew Schwartz and David H. Epstein and Lorenzo Leggio and Brenda Curtis},
  title = {{Bots and Misinformation Spread on Social Media: Implications for COVID-19 (Preprint)}},
  journal = {Journal of Medical Internet Research},
  year = {2021},
  doi = {10.2196/26933},
}
Figure