Twits, Toxic Tweets, and Tribal Tendencies: Trends in Politically Polarized Posts on Twitter

TL;DR

This work trains and open-source a DeBERTa-based toxicity detector with a contrastive objective that outperforms the Google Jigsaw Persective Toxicity detector on the Civil Comments test dataset, and finds that toxic comments are correlated with users who engage with a wider array of political views.

Abstract

Social media platforms are often blamed for exacerbating political polarization and worsening public dialogue. Many claim that hyperpartisan users post pernicious content, slanted to their political views, inciting contentious and toxic conversations. However, what factors are actually associated with increased online toxicity and negative interactions? In this work, we explore the role that partisanship and affective polarization play in contributing to toxicity both on an individual user level and a topic level on Twitter/X. To do this, we train and open-source a DeBERTa-based toxicity detector with a contrastive objective that outperforms the Google Jigsaw Perspective Toxicity detector on the Civil Comments test dataset. Then, after collecting 89.6 million tweets from 43,151 Twitter/X users, we determine how several account-level characteristics, including partisanship along the US left-right political spectrum and account age, predict how often users post toxic content. Fitting a Generalized Additive Model to our data, we find that the diversity of views and the toxicity of the other accounts with which that user engages has a more marked effect on their own toxicity. Namely, toxic comments are correlated with users who engage with a wider array of political views. Performing topic analysis on the toxic content posted by these accounts using the large language model MPNet and a version of the DP-Means clustering algorithm, we find similar behavior across 5,288 individual topics, with users becoming more toxic as they engage with a wider diversity of politically charged topics.

Figures (19)

• Figure 1: Estimated Political Orientation of Political Leaders and All Users Using CA-- We differentiate users' political leanings based on who they follow on Twitter.
• Figure 2: Examples of Tweet pairs at different similarities (0.735 left and -0.032 right).
• Figure 3: Topic analysis of Toxic Tweets---We determine the toxicity, embed, and cluster toxic tweets to identify the most polarized and toxic conversations on Twitter throughout 2022. We note that for this approach, we limit our analysis to English tweets. We utilize the whatlango Go library to determine the language of tweets.
• Figure 4: Partial dependencies with 95% Normal Confidence intervals between our fitted standardized dependent variables and user toxicity.
• Figure 5: The more toxic the users mentioned by a given user, on average, the more toxic the content of that particular user. Within the mention graph (the darker the purple the more toxic) of user interactions, toxicity has an assortativity coefficient of 0.071, suggesting that, to some degree, users who post toxic content have a slight tendency to mention and interact with other users who post toxic content.