The Internet Research Agency, a Russia-based group of Internet trolls, relied on local news more than it did fake news to disrupt the 2016 presidential election, finds a new analysis by NYU's Social Media and Political Participation lab.
The Internet Research Agency, a Russia-based group of Internet trolls, relied on local news more than it did fake news to disrupt the 2016 presidential election, finds a new analysis by New York University’s Social Media and Political Participation (SMaPP) lab.
“Russian trolls shared five times more local news content than they did junk news content, our research shows,” says Professor Joshua Tucker, co-director of the SMaPP lab. “This suggests that attempts simply to fight fake news will not be enough to stop future attempts at manipulating the information environment surrounding elections on social media.”
“We suspect that the IRA relied so heavily on local news sources because they believed that Americans trust their local media outlets more than other sources,” said Leon Yin, a research scientist at the SMaPP lab. “In fact, a Pew Research Center report from 2017 found that Americans not only trusted news from local sources more than news from national sources, but also more than from friends and family.”
The findings and larger analysis, “Your Friendly Neighborhood Troll,” are contained in a SMaPP Data Report available on the SMaPP Lab website and are summarized in a post on the Monkey Cage blog.
Just prior to the 2018 mid-term elections, Facebook announced that it had removed accounts it thought were linked to the Kremlin-backed Internet Research Agency (IRA), a troll farm that has been indicted for meddling in the 2016 presidential election by Special Counsel Robert Mueller.
Given the IRA’s role in two U.S. election cycles, the SMaPP researchers, who also included Franziska Roescher, Richard Bonneau, and Jonathan Nagler, sought to more deeply understand the agency’s online strategy.
To do so, it analyzed all known IRA-operated Twitter accounts and examined what kind of online content the trolls shared ahead of the November 8, 2016 presidential election. Overall, it found that IRA-linked accounts shared less fake, or “junk,” news than might be expected and instead relied heavily on local news sources.
Specifically, the study drew on a dataset shared online by Twitter’s Elections Integrity Initiative of more than 9 million tweets sent by approximately 3,600 IRA-linked accounts in 2016. It discarded tweets without links and accounts that posted in Russian, leaving 556 accounts that tweeted approximately 209,000 links between January 2016 and November 2016.
In order to understand how troll behavior compared to legitimate behavior on Twitter, the team collected tweets from two comparison groups over the same time period: politically engaged users and random users. The sample of random Twitter users contained 1,344 accounts that tweeted approximately 106,000 links; the sample of politically engaged users encompassed 1,952 accounts that shared roughly 437,000 URLs. The full results of the analysis, including additional methodological details, can be found in the SMaPP Data Report.
Among SMaPP’s findings were the following:
· The IRA did not seem to have embarked on a concerted misinformation campaign on Twitter in 2016. Only about 6 percent of all links the trolls shared on Twitter led to known junk news sites. Those are websites that have been classified by Merrick College researchers to contain extreme bias, conspiracy theories, rumors, or junk science. However, this was more than the average politically interested user linked to these types of sites (4 percent of all links) and much more than an average Twitter user (0.4 percent of links).
· The number of links to junk news sites spiked sharply in the weeks immediately leading up to the fall elections, suggesting that the trolls put extra effort into spreading falsehoods just before Election Day. At its peak, the IRA tweeted out more than 2,000 links to junk news sites in a single week in September 2016.
· The IRA relied heavily on local news media: 30 percent of all links led to local news websites, such as Kansas-based station KSNT or regional website cleveland.com, amounting to five times more local news content than junk news content. By comparison, politically interested users had 1.9 percent of links point to local news sources and random users had only 0.3 percent of links going to local news sources.
The researchers add that some IRA accounts also pretended to be local news outlets themselves. They used Twitter handles such as “Atlanta_Online” or “KansasDailyNews” and tweeted heavily about the region they pretended to be located in, sharing links to articles from real local news outlets. These fake local news outlet Twitter handles were responsible for the vast majority of the IRA’s local news sharing: 80 percent of all local media content came from just 27 accounts. Many of their tweets talked about “Trump,” “Clinton,” or “#politics,” meaning that these local news sources were at least in part used to spread news about the presidential race.
In addition, another set of IRA accounts masqueraded as left or right partisans, the SMaPP study found. It examined the tweeting behavior of 14 IRA accounts that pretended to be white conservative Americans -- using Twitter handles such as “USA_Gunslinger” or “SouthLoneStar” -- and eight IRA accounts posing as left-leaning partisans, usually pretending to be African-Americans, with screen names like “BlackToLive” or “TrayneshaCole.”
These partisan accounts also shared local news, but were quite polarizing in their framing: while both sides often tweeted about the police, crime, and minorities, the right-leaning accounts tended to focus on crimes committed by minorities and be supportive of law enforcement, while the left-leaning accounts rather emphasized crimes against minorities and police brutality.
Despite the invasive nature of the IRA’s efforts, the group’s efficacy is uncertain.
“While the IRA activity seems to reveal a degree of sophistication about American politics in terms of its use local news sources, the accounts look decidedly less sophisticated in terms of geographical targeting, at least in so far as the 2016 presidential election is concerned,” the researchers write. “The troll farm tweeted most of the local news content from outlets based in Kansas, California, and Missouri -- none of which were considered swing states in 2016.”
About the study’s authors:
Leon Yin is a research scientist in the NYU Social Media and Political Participation (SMaPP) lab.
Franziska Roscher is a Ph.D. candidate in politics at New York University and a graduate research associate in the NYU Social Media and Political Participation (SMaPP) lab.
Richard Bonneau is professor of biology and computer science at NYU and a co-director of the NYU Social Media and Political Participation (SMaPP) lab.
Jonathan Nagler is professor of politics at NYU and a co-director of the NYU Social Media and Political Participation (SMaPP) lab.
Joshua A. Tucker is professor of politics at NYU and a co-director of the NYU Social Media and Political Participation (SMaPP) lab as well as director of NYU’s Jordan Center for the Advanced Study of Russia.