The 2016 presidential election will never die—or, at the very least, we appear doomed to discuss it forever. Earlier this month, NYU’s Center for Social Media and Politics published a study in Nature Communications that complicates one purported element of Donald Trump’s ascension: the influence of Russian Twitter trolls. The researchers looked at roughly 1.2 billion tweets from the lead-up to the 2016 election. They sought to quantify just how many ordinary U.S. Twitter users were exposed to Russian accounts, and to better understand how that exposure did or did not change users’ political attitudes and voting behavior.
One of their findings quickly made headlines, although not in a way they intended. The researchers determined that Russian troll accounts on Twitter demonstrated little ability to change voter behavior. The majority of U.S. Twitter users surveyed were simply not exposed to posts from the Internet Research Agency, Russia’s troll farm. And many of the ones who were self-identified as highly partisan Republicans—people who seemed likely to vote for Trump anyway.
So far, so good. In hindsight, the findings seem logical: Simply seeing a few random tweets from Russians pretending to be angry, partisan Americans isn’t the sort of thing that causes somebody to drop everything and reconsider their politics. But plenty of diehards are still relitigating the 2016 outcome. The study has been weaponized and its findings distorted or downplayed, depending on one’s political views. As with so many well-meaning efforts to understand the effect of digital platforms on our politics, the nuance of the work has been once again flattened and corrupted by the incentives of those very same platforms.
For instance, Glenn Greenwald tweeted the study as proof that “Russiagate” was “one of the most deranged and unhinged conspiracy theories in modern times.” Breitbart definitively declared, “Democrat Narrative Falls Apart: Study Finds Russian Trolls Had Little Influence on 2016 Voters.’” NYU’s center and its authors attempted to correct the record with a Twitter thread, to little effect (their thread has been retweeted fewer than 60 times; Greenwald’s received nearly 5,000 retweets and 1.7 million views).
To its credit, the research is full of caveats. Its authors note that the study covers only Twitter’s social-network domain and leaves out much bigger platforms, such as Facebook. Similarly, they argue that the study does not address other prongs of Russia’s documented efforts to meddle in the election, including its email-hacking campaigns targeting the Democratic National Committee and people connected to Hillary Clinton, which were leaked and covered by national media. Josh Tucker, one of the report’s authors, told me repeatedly that the study was just a small piece of a complicated puzzle and did not suggest that Russian efforts had no effect on the 2016 outcome. “The entire paper is predicated on the fact that Russians tried to interfere in the 2016 election, which I take as a serious national-security issue,” he said.
The research is part of a trend. In recent years, there’s been pushback on the way #resistance tweeters and even mainstream news outlets have used Russian “bots” or trolls as an easy scapegoat to help explain both the provenance of successful right-wing narratives and some of the popular support for MAGA Republicans. Journalists such as Michael Lewis and the French reporter Anthony Mansuy have also gone back to reexamine the Cambridge Analytica data scandal, questioning the notion that the company’s psychographic profiling and targeting meaningfully swayed the election results. It all serves to illustrate that the result of the 2016 election is far more complicated than any single factor can explain.
“I have mixed feelings about this,” Tucker told me when I asked him about how he thought his study might fit into this broader reappraisal. “We had a geopolitical rival trying to interfere in an election, and that was real and serious. This was not something that should’ve been swept under the rug.” But, he added, “campaigns spend billions to try and do this, so why are we sure some tweets moved the needle?”
And Tucker gestured toward an unintended consequence of the endless conversation about Russia’s interference: “I worry that we spent four years thinking about the fragility of American elections and how easy it is to change the outcome, and that makes the soil more fertile for claims of the illegitimacy of elected candidates.”
Russia’s attempts to meddle were not confined to Twitter. In 2017, Facebook estimated that 126 million users may have viewed Russian-sponsored posts, as opposed to 32 million who, according to the Nature study, were exposed on Twitter. Scholars such as Kathleen Hall Jamieson have done extensive work suggesting that Russia’s email hack-and-leak operation, aided by media amplification, was likely a contributing factor in the electoral outcome. It’s foolish to suggest that a few troll tweets swung the 2016 election, but it’s also unwise to dismiss a multipronged attempt to disrupt the American democratic system because of this one study.
“What [the research shows] is this is an important piece of the larger 2016 puzzle,” Kate Starbird, a co-founder of the University of Washington’s Center for an Informed Public, told me. She explained that the study only confirms what media-studies scholars have long known: that one piece of targeted information or propaganda rarely changes one’s opinion. (That idea is known as the “hypodermic needle effect.”) More likely, people are influenced by ideas that travel to mass media, and then by convincing personalities who repackage and disseminate that information. The NYU study doesn’t factor in this indirect exposure, she argues, which would include things like troll tweets embedded into news articles that are then shared across the internet. “Just because we can’t measure impact doesn’t mean there isn’t impact,” Starbird said.
It’s easy to blame overly credulous media or various pundits and researchers for simplifying the complex election interference narrative down to “Russia did it,” but it’s also important to remember how primed many Americans were to believe in the power of social networks to manipulate user emotions and shape public opinion. In 2014, Facebook faced intense backlash and generated tons of headlines over a study in which researchers said they had manipulated the emotional state of nearly 700,000 users. Two years before that study, Facebook released research purportedly demonstrating the power of its digital I Voted stickers to increase voter turnout. A credulous public and press took company claims about platforms’ ability to influence behavior at face value. So it makes sense that, in the aftermath of Clinton’s surprising defeat, people latched onto a simple narrative. People may have freaked out, but they didn’t do so without reason.
If this all sounds mealy-mouthed and frustrating and inconclusive, it’s because studying the flow of information across dozens of open and closed ecosystems and assessing the impact is exceedingly difficult. Ten thousand analyses and reappraisals will never provide a smoking gun that we can point to as the exact reason the 2016 election turned out as it did. “It’s a complicated story,” Starbird told me. “It’s always been a complicated story. If this study or any other makes it look simple, that’s a mistake we’re making.”
We’re still struggling to incorporate that complexity into our broader understanding of how social platforms affect elections, and in that case the NYU study does not offer any heartening signs. The thoughtful context provided by digital forensics is immediately drowned out by the very information ecosystem it’s trying to demystify. Few minds are changed. More than six years after the 2016 election, we don’t really know the impact of Russian meddling except that angry, conspiratorially minded Americans continue to fight over whether it happened and to what extent it mattered. Maybe that’s proof enough that the trolls succeeded—if not in annihilating our democratic system, then at least in making so many of us trust it and one another less.