How does Twitter account moderation work? Dynamics of account creation and suspension on Twitter during major geopolitical events

Abstract

Social media moderation policies are often at the center of public debate, and their implementation and enactment are sometimes surrounded by a veil of mystery. Unsurprisingly, due to limited platform transparency and data access, relatively little research has been devoted to characterizing moderation dynamics, especially in the context of controversial events and the platform activity associated with them. Here, we study the dynamics of account creation and suspension on Twitter during two global political events: Russia’s invasion of Ukraine and the 2022 French Presidential election. Leveraging a large-scale dataset of 270M tweets shared by 16M users in multiple languages over several months, we identify peaks of suspicious account creation and suspension, and we characterize behaviors that more frequently lead to account suspension. We show how large numbers of accounts get suspended within days of their creation. Suspended accounts tend to mostly interact with legitimate users, as opposed to other suspicious accounts, making unwarranted and excessive use of reply and mention features, and sharing large amounts of spam and harmful content. While we are only able to speculate about the specific causes leading to a given account suspension, our findings contribute to shedding light on patterns of platform abuse and subsequent moderation during major events.

Analyzing Mentions of Death in COVID-19 Tweets

“…To assess this, on Jan 31, 2023, we compared the percentage of active, deleted, and suspended Twitter users in our dataset (79.5%, 13.1%, and 7.4%, respectively) with that in a sample of tweets from April 2020/March 2021 provided by the Twitter Stream Grab project ¶ (74.3%/82.2%, 18.3%/10.9%, and 7.4%/6.9%, respectively). Given that (i) suspension is often linked to bot activity (Pierri et al 2023), and that (ii) the suspension percentages are similar in our dataset on a generic sample, this suggests that overall bot activity is not unusually prominent in our sample.…”

“…Both Sarker et al (2020) and Mackey et al (2020) studied self-reported COVID-19 symptoms and as well as recovery progression and compared them with those reported in studies conducted in clinical settings. Conceptually, the work by Paul, Dredze, and Broniatowski (2014) and Broniatowski, Paul, and Dredze (2013) is related to ours as they show that for monitoring flu activity using tweets processed by a firstperson classifier is helpful as it separates tweets discussing the flu in general from first-person reports of the flu.…”

A Longitudinal Study of Italian and French Reddit Conversations Around the Russian Invasion of Ukraine

“…Pierri et al [23] investigate the dynamics of account creation and suspension on Twitter during the Russian invasion of Ukraine and the 2022 French Presidential elections. The same authors carry out in [24] a longitudinal study of the diffusion of misinformation about the Russian invasion of Ukraine which originated from the low-credibility sources posted on Facebook and Twitter and from Russian state outlets, during the first months of the conflict.…”