Abstract
Over the past couple of years, anecdotal evidence has emerged linking coordinated campaigns by state-sponsored actors with efforts to manipulate public opinion on the Web, often around major political events, through dedicated accounts, or "trolls." Although they are often involved in spreading disinformation on social media, there is little understanding of how these trolls operate, what type of content they disseminate, and most importantly their influence on the information ecosystem. In this paper, we shed light on these questions by analyzing 27K tweets posted by 1K Twitter users identified as having ties with Russia's Internet Research Agency and thus likely state-sponsored trolls. We compare their behavior to a random set of Twitter users, finding interesting differences in terms of the content they disseminate, the evolution of their account, as well as their general behavior and use of Twitter. Then, using Hawkes Processes, we quantify the influence that trolls had on the dissemination of news on social platforms like Twitter, Reddit, and 4chan. Overall, our findings indicate that Russian trolls managed to stay active for long periods of time and to reach a substantial number of Twitter users with their tweets. When looking at their ability of spreading news content and making it viral, however, we find that their effect on social platforms was minor, with the significant exception of news published by the Russian state-sponsored news outlet RT (Russia Today).
Introduction
While the existence of "troll farms"—specifically Russia’s Internet Research Agency (IRA)—is well-documented by journalists and intelligence agencies, quantifying their actual impact on the digital information ecosystem has remained elusive. Previous research often conflated human trolls with automated bots, failing to distinguish the nuanced behavior of state-sponsored actors. This study addresses the modus operandi of these trolls, the specific narratives they disseminate, and their mathematical influence on the spread of information across major social platforms like Twitter, Reddit, and 4chan.
Methodology
The researchers analyzed a ground-truth dataset of 2,700 Twitter account IDs identified by the US Congress as linked to the Russian IRA. From this, they extracted 27,000 tweets from 1,000 accounts active between January 2016 and September 2017. This data was compared against a baseline of 96,000 tweets from random users with similar posting frequencies.
To quantify influence, the study employed Hawkes Processes, a stochastic model where past events increase the probability of future events. By fitting this model to the appearance of specific URLs across platforms, the authors estimated the "strength of connections" between communities and quantified how much one platform's activity "caused" activity on another.
Characterizing the Troll Modus Operandi
The analysis reveals a professionalized infrastructure. Unlike the average user who primarily tweets from mobile devices, the majority of Russian trolls utilized the Twitter Web Client. This suggests a "desk job" environment where operators manage multiple accounts from a browser. Furthermore, their activity patterns are highly revealing. While trolls often set their locations to the US to appear local, their posting times peak consistently between 14:00 and 15:00 UTC, aligning with a standard 9-to-5 workday in Moscow (UTC+3).
Strategic Account Deployment
Trolls employ a "reset" strategy, frequently deleting their entire tweet history and changing screen names to adopt new personas. This allows an account to transition from a generic news profile to a hyper-partisan identity just before a major political event. The timing of account creation is equally strategic, with significant spikes occurring shortly before major events like the 2016 Republican National Convention and the 2017 "Unite the Right" rally in Charlottesville.
Quantifying Influence: The RT Exception
When measuring the general dissemination of news URLs on Twitter, the trolls' influence was found to be negligible, at approximately 0.01%. However, a stark exception exists for content from RT (Russia Today). For RT-specific links, the trolls were four times more effective at influencing Twitter users and five times more effective on Reddit compared to their influence on other news sources. This indicates that while they struggle to make mainstream news go viral, they act as a highly effective, dedicated distribution arm for state-funded narratives.
Conclusion
Russian trolls are professional actors who manage to remain active for long periods through strategic identity shifts and localized personas. While their overall impact on the broader news cycle is minor, their ability to amplify state-sponsored media like RT is significant. Future work should focus on developing real-time detection tools that can account for the "identity-reset" behavior and cross-platform influence identified in this study.
Summary updated