[Paper Review] On the Influence of Twitter Trolls during the 2016 US Presidential Election

Nikos Salamanos, Michael J. Jensen+3 more|arXiv (Cornell University)|Oct 1, 2019

Misinformation and Its Impacts17 references19 citations

TL;DR

This study analyzes 152.5 million tweets from 9.9 million users, including 822 state-sponsored 'troll' accounts, to assess their influence during the 2016 U.S. presidential election. Using a large-scale interaction graph of 9.3 million nodes and 169.9 million edges, the authors find that while trolls participated in viral cascades, they played a minimal role—only four accounts were truly influential—and authentic users were the primary drivers of information diffusion.

ABSTRACT

It is a widely accepted fact that state–sponsored Twitter ac- counts operated during the 2016 US presidential election spreading millions of tweets with misinformation and inflam- matory political content. Whether these social media cam- paigns of the so–called “troll” accounts were able to manipulate public opinion is still in question. Here we aim to quantify the influence of troll accounts and the impact they had on Twitter by analyzing 152.5 million tweets from 9.9 million users, includ- ing 822 troll accounts. The data collected during the US elec- tion campaign, contain original troll tweets before they were deleted by Twitter. From these data, we constructed a very large interaction graph; a directed graph of 9.3 million nodes and 169.9 million edges. Recently, Twitter released datasets on the misinformation campaigns of 8,275 state–sponsored accounts linked to Russia, Iran and Venezuela as part of the investigation on the foreign interference in the 2016 US election. These data serve as ground–truth identifier of troll users in our dataset. Us- ing graph analysis techniques we qualify the diffusion cascades of web and media context that have been shared by the troll accounts. We present strong evidence that authentic users were the source of the viral cascades. Although the trolls were par- ticipating in the viral cascades, they did not have a leading role in them and only four troll accounts were truly influential.

Motivation & Objective

• To quantify the influence of state-sponsored Twitter troll accounts during the 2016 U.S. presidential election.
• To determine whether these troll accounts significantly manipulated public opinion or shaped viral information cascades.
• To use verified datasets of foreign-linked accounts to identify and analyze troll behavior in real-time social media dynamics.
• To assess the role of authentic users versus trolls in the diffusion of political content on Twitter.

Proposed method

• Collected and analyzed 152.5 million tweets from 9.9 million users, including 822 verified troll accounts linked to Russia, Iran, and Venezuela.
• Constructed a large-scale directed interaction graph with 9.3 million nodes (users) and 169.9 million edges (interactions) to model information diffusion.
• Utilized graph analysis techniques to trace and qualify the diffusion cascades of web and media content shared by troll accounts.
• Leveraged Twitter’s public dataset identifying 8,275 state-sponsored accounts as ground-truth labels for troll user classification.
• Tracked the propagation of content through the network to distinguish between influence originating from trolls versus authentic users.

Experimental results

Research questions

• RQ1 To what extent did state-sponsored troll accounts drive the virality of political content on Twitter during the 2016 U.S. election?
• RQ2 Were troll accounts the primary sources of information cascades, or did they merely amplify content initiated by authentic users?
• RQ3 How many troll accounts were actually influential in spreading content across the network?
• RQ4 What was the relative contribution of authentic users versus trolls in the diffusion of inflammatory or misleading political content?

Key findings

• Troll accounts participated in viral cascades but did not lead them; authentic users were the primary source of information diffusion.
• Only four out of 822 troll accounts were found to be truly influential in spreading content across the network.
• The majority of high-impact cascades originated from non-troll, authentic users rather than state-sponsored accounts.
• The interaction graph revealed that while trolls were active, their reach and influence were limited compared to organic user-driven virality.
• The data show that misinformation and inflammatory content spread primarily through authentic user engagement, not coordinated troll campaigns.
• The study provides strong evidence that state-sponsored troll accounts had minimal structural influence in the broader information ecosystem on Twitter.