Mandatory Credit: Photo by MARTIN DIVISEK/EPA-EFE/REX/Shutterstock (9843027ck) Right-wing protesters shout behind a row of police men in Chemnitz, Germany, 01 September 2018. Organizations of civil society and right-wing groups called for several demonstrations on the weekend after two refugees from Syria and Iraq were arrested on suspicion of stabbing a 35-year-old man in what police described as a 'scuffle between members of different nationalities' at a city festival in the East German city Chemnitz. Demonstrations after stabbing of a 35-year-old man in Chemnitz, Germany - 01 Sep 2018
German anti-refugee protest in 2018 © Martin Divisek/EPA-EFE/Rex/Shutterstock

Does hate speech on social media drive people to attack refugees? A study published this year suggested that Facebook posts triggered hundreds of violent crimes against refugees in Germany.

Two PhD students at the UK’s Warwick University, Karsten Müller and Carlo Schwarz, studied their native Germany between 2015 and 2017. During surges in online anti-migrant sentiment, they estimated that areas with higher Facebook populations saw up to 50 per cent more anti-refugee incidents — mostly violent crimes, including refugees’ homes being set on fire — than the national average. They attributed this to the spread of hate posts.

Social media can push potential perpetrators over the line,” says Mr Schwarz. “Their views get more extreme [from reading hate posts] and, at some point, they might decide to assault someone.”

The pair estimate that Facebook posts from the rightwing Alternative for Germany (AfD) party drove 13 per cent of the 3,335 recorded anti-refugee crimes from January 2015 to February 2017.

To gauge potential exposure to Facebook posts, they scraped location data from thousands of German users of Facebook groups, including the AfD’s. They then matched the Facebook user location information with local weekly data on anti-refugee incidents.

A pattern emerging from their work led them to conclude Facebook posts were causing the additional violence: when people in places that usually see more anti-refugee attacks have limited access to the platform, because of internet outages or service disruptions, the violence sharply decreases. “I’m quite confident our evidence indicates a causal link,” says Mr Schwarz.

To disprove this, he says, “you would need an alternative explanation for why a Facebook outage makes the relationship between social media use and hate crimes disappear”.

Researchers have already explored correlations between online hate and offline attacks, but the pair are among the first to suggest causation. Other studies in the works include one looking at whether Russian communities with more users of VKontakte, a popular social media network in Russia, see more ethnic hate crime.

“We need to replicate this German study elsewhere,” says Ghayda Hassan, a psychology professor at the Université du Québec à Montréal who studies online radicalisation. She says it contains “the most robust set of analyses and data I have seen” but, like other experts, she has questioned whether it definitely proves causation.

Mr Müller and Mr Schwarz admit that their findings, which are yet to be peer reviewed, will require more work. But they hope the paper will help society to consider the potential effects of proliferating virtual hatred.

Technology companies have been facing greater scrutiny for hate speech on their platforms and have removed content more quickly. European Union data in January showed that Facebook, Twitter, YouTube and Microsoft remove on average 70 per cent of flagged illegal hate speech within 24 hours, compared with 28 per cent in 2016.

“The discussion is now moving into this wider, less well-defined area of hate speech, conspiracy theories and other socially damaging content, which may not be illegal,” says Carl Miller, research director for the social media centre at Demos, a think-tank.

Facebook declined to comment on the study but said it was “working hard to prevent the spread of hate speech online”.

Copyright The Financial Times Limited 2024. All rights reserved.
Reuse this content (opens in new window) CommentsJump to comments section

Follow the topics in this article