Comparing Toxicity Across Social Media Platforms for COVID-19 Discourse

Nahiyan Bin Noor, Niloofar Yousefi, Billy Spann, Nitin AgarwalarXiv 2023 · OpenReview

TL;DR

This study quantifies and compares the toxicity of COVID-19 discussions across Twitter, Parler, and Reddit. It reveals that while Parler generally exhibits lower toxicity than Twitter and Reddit, Reddit hosts the most toxic content, particularly concerning vaccine discussions, often driven by anti-vaccine forums.

What Problem Does This Paper Address?

The proliferation of toxic content, misinformation, and hate speech on social media platforms has become a significant concern. While platforms like Twitter and Facebook have content moderation policies, others like Parler are designed with fewer restrictions, potentially becoming havens for harmful discourse. Understanding how toxicity manifests and differs across these platforms, especially during critical events like the COVID-19 pandemic, is crucial for developing effective mitigation strategies and fostering healthier online environments. This paper specifically investigates the landscape of COVID-19 discussions on three distinct platforms to identify patterns and comparative toxicity levels.

Methods

The researchers employed a computational approach using the Detoxify API, a Convolutional Neural Network-based tool, to analyze the toxicity of user-generated text. Detoxify assigns a probability score between 0 and 1 for various toxicity categories, including overall toxicity, obscenity, and insult. For this study, a score of 0.5 or greater was considered toxic. Before analysis, seed keywords and hashtags were removed from the data to prevent bias. The study focused on seven toxicity categories but reported on three (toxicity, obscenity, and insult) due to data availability. The analysis involved comparing toxicity scores across different platforms and over time.

Datasets

The study utilized data collected from Twitter, Parler, and Reddit between January 1, 2020, and December 31, 2020. The datasets were curated based on a list of seed hashtags related to COVID-19, including terms like "#fckcovid," "#fcklockdown," "#fckmask," and "#fckvaccine." For Twitter, data was accessed via the Twitter Developer API, with an acknowledgment that some tweets removed for misinformation might not have been included. Reddit data was collected using the Pushshift API and a custom Python script. A notable aspect is the inclusion of data from a previously banned subreddit that was known for spreading anti-vaccine misinformation.

Key Insights and Ideas

  • Platform-Specific Toxicity Hotspots: While Parler is often perceived as a more toxic platform, this study found it had lower toxicity than Twitter and Reddit for COVID-19 discussions. This challenges common assumptions and highlights the nuanced nature of online toxicity.
  • Reddit's Vaccine Discourse Problem: Reddit emerged as the most toxic platform overall, primarily driven by its vaccine-related discussions. This is attributed to the presence of anti-vaccine forums that actively spread misinformation.
  • Twitter's Conspiracy Theory Undercurrent: Even on a more moderated platform like Twitter, COVID-19 vaccination conversations revealed a significant presence of conspiracy theories among highly toxic users.
  • The Power of a Few: The study suggests that a small number of highly toxic users or conversations can disproportionately drive the overall toxicity levels on a platform, implying that targeted moderation could be highly effective.
  • Keyword-Driven Toxicity: Different keywords or hashtags are associated with higher toxicity on different platforms. For instance, "#fckcovid" was more toxic on Parler, while "#fckvaccine" was overwhelmingly toxic on Reddit.

Key Results

  • Overall Toxicity: Reddit exhibited the highest overall percentage of toxic content (toxicity score > 0.5) across all datasets, with approximately 37% of its content being toxic, compared to Twitter (just over 30%) and Parler (21.83%).
  • Vaccine Toxicity: The vaccine dataset on Reddit showed an exceptionally high toxicity rate of 86%, significantly exceeding that of Twitter and Parler for the same topic.
  • Lockdown Toxicity: Twitter had the highest percentage of toxic posts related to the "#f*cklockdown" hashtag (34.31% of tweets).
  • Obscenity and Insult: While toxicity varied, toxic content on all platforms often included obscene and insulting language. Twitter's content was more obscene, particularly in lockdown discussions, while Parler's COVID-19 content leaned more towards insults.

Limitations and Future Work

The authors acknowledge a potential limitation in their methodology: the seed hashtag stem "#f*ck" can be interpreted both positively and negatively, and the Detoxify model may struggle to distinguish semantic context, potentially misclassifying profane but non-toxic content.

Future work planned includes expanding the keyword list and collecting more data from these platforms, as well as exploring other popular platforms like TikTok and Facebook. The researchers also intend to delve deeper into the vaccine and lockdown topics, given their notably higher toxicity on Reddit.

Summary updated

Abstract

The emergence of toxic information on social networking sites, such as Twitter, Parler, and Reddit, has become a growing concern. Consequently, this study aims to assess the level of toxicity in COVID-19 discussions on Twitter, Parler, and Reddit. Using data analysis from January 1 through December 31, 2020, we examine the development of toxicity over time and compare the findings across the three platforms. The results indicate that Parler had lower toxicity levels than both Twitter and Reddit in discussions related to COVID-19. In contrast, Reddit showed the highest levels of toxicity, largely due to various anti-vaccine forums that spread misinformation about COVID-19 vaccines. Notably, our analysis of COVID-19 vaccination conversations on Twitter also revealed a significant presence of conspiracy theories among individuals with highly toxic attitudes. Our computational approach provides decision-makers with useful information about reducing the spread of toxicity within online communities. The study's findings highlight the importance of taking action to encourage more uplifting and productive online discourse across all platforms.
BibTeX
@misc{noor2023comparing,
  author = {Nahiyan Bin Noor and Niloofar Yousefi and Billy Spann and Nitin Agarwal},
  title = {{Comparing Toxicity Across Social Media Platforms for COVID-19 Discourse}},
  year = {2023},
  eprint = {2302.14270},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2302.14270},
}
Figure