The Twitter logo is displayed at the entrance of Twitter headquarters in San Francisco. (Kimihiro Hoshino/Agence France-Presse via GettyImages)

In May, the New Yorker called 2016 “The Year of the Political Troll.” Hate speech, intimidation, mockery, and racism are increasing globally online. In fact, there’s so much incivility, animosity, and disrespect on social media that some 65 percent of American social media users express resignation and frustration about online political conversations.

That’s not just affecting individual social media users. It’s hurting democracy. Here’s why.

Trolling discourages social media’s democratic potential

Only a few years ago, many observers and scholars were touting social media’s tremendous potential to enrich democracy. That potential included the possibility that political actors and ordinary citizens from a variety of groups might communicate directly, thereby signaling the beginning of a more participatory kind of politics and a gradual repairing of a relationship notoriously characterized by mistrust.

But in our new paper in the Journal of Communication, we show that trolling and harassment have made politicians extremely cautious about engaging with citizens on these platforms. Direct interaction brings the risk of falling victim to trolling and harassment almost immediately. Why engage with the public if doing so brings only personal abuse?

Instead, politicians use these tools for their marketing and mobilization potential. On Twitter, politicians can promote themselves, bypass gatekeepers, personalize their messages, and craft their own images without — if they prefer — responding to a single question from the public.

That reduces or even eliminates social media’s potential for open, interactive political deliberation that could better inform our democracies.

Here’s how we tested this and what we found

We empirically tested this with data from the tweets of German, Greek, Spanish, and British candidates running in the 2014 elections to the European Parliament. We analyzed the tweets they posted and the reactions those elicited.

Using automated text analysis techniques and supervised classification, we were able to measure which of the candidates’ messages aimed at engaging with voters (as opposed to broadcasting campaign messages) and whether the responses were polite and civil.

Around 5 percent of all tweets mentioning candidates were impolite — but that percentage was even higher for candidates who were trying to interact with the public.

As the figure below shows, messages posted by candidates who try to connect with voters brought vitriol. Protected by their apparent anonymity on Twitter, some people would respond to interactions by harassing or attacking the candidate. This result holds when we compare across different candidates and over time.


Tweets by candidates to the European Parliament who try to engage directly with voters tend to receive more (as in, greater quantity) impolite responses
Data: European Election Study 2014, Social Media Study.
Figure: Yannis Theocharis, Pablo Barberá, Zoltán Fazekas, Sebastian Popa.

Methodology: We use automated text analysis and supervised learning methods to measure whether each individual tweet by a candidate engages the public or broadcasts to them, and to detect impoliteness and incivility in citizens’ replies. To facilitate the visualization, the figure displays tweets aggregated into 20 different bins by country, according to their predicted probability of being considered engaging.

What do our findings mean?

1. Politicians aren’t ignoring citizens on social media. They’re avoiding them.

Some observers have claimed that politicians use social media to provide a spectacle of interactivity — and have decided to use those tools not for genuinely engaging citizens and supporters, but for “push” communications that can help them win, such as asking people to vote or contribute to their campaign.

In fact, many politicians do try to launch discussions with the public on Twitter. But when they do, they’re attacked in return — and risk becoming the target of trolling campaigns. Such campaigns can require significant damage control and can irreparably harm candidates’ public images.

A recent study commissioned by the Telegraph showed that, from the tweets addressed to British politicians, only a minority represent constructive criticism or legitimate complaints. Most politicians who post anything quickly become subject to constant personal abuse.

Even highly calculated tweets can backfire, as you can see in the responses when the Hillary Clinton campaign asked people to express their feelings about student debt in three emoji. Getting snark and criticism in return, candidates become discouraged and reluctant to engage. They adjust their social media communications, reducing their attempts at interactive discussion and instead “pushing” out messages.

Those who do risk direct engagement interactively should know what to expect and have a specialized team that can deal with this.

2. Citizens and social media platforms bear some responsibility for discouraging politicians’ civic engagement online

Even when candidates are willing to communicate directly online, they have good reasons to choose not to risk it. A significant minority of citizens tend to behave badly in these largely anonymous online platforms, in part because of the constraints imposed (or at least not prevented) by the platforms themselves.

Social media platforms bear some responsibility as well. Twitter isn’t effectively curbing online incivility. That means trolls thrive, a fact recognized by Twitter’s former chief executive.

3. Trolling is toxic for democracy

Finally, the toxic effect of this behavior for democracy must not be underestimated. Yes, trolling encourages candidates to use social media only for marketing and not for interaction.

But trolling is still more corrosive for online political discussions more generally. Incivility, online harassment and abuse increase others’ anger, anxiety and aversion. Trolling may also downgrade the overall quality of online discussions, demobilize citizens who are trolled and reduce satisfaction with the platform’s potential for discourse.

More work must be done on how to effectively reduce such harassment, to benefit both citizens individually and democracy in general.

Yannis Theocharis is senior research fellow at the Mannheim Centre for European Social Research, University of Mannheim.

Pablo Barberá is assistant professor at the School of International Relations, University of Southern California.

Zoltán Fazekas is a postdoctoral researcher at the University of Oslo.

Sebastian Popa is a postdoctoral researcher at the Mannheim Centre for European Social Research, University of Mannheim.