Could organized ‘counter-speech’ curb online hate?
Analysis of a polarized Twitter debate in Germany revealed that organized counter-messages were effective in defusing online hate speech.
Around the world, the internet is plagued by online hate speech—insulting and harmful messages about groups or individuals, and even calls for violence against them. Although no single solution is likely to rein in hate speech, a new preliminary study suggests that posting positive counter-messages, and organizing anti-hate groups whose members can support each other online, may help reduce the prevalence and reach of abusive or threatening discourse.
The growth and breadth of hate speech on social media and other online platforms has complicated efforts to confront and curb it. But an analysis of German conversations on Twitter between 2013 and 2018 suggests that counter-speech, especially from organized groups, could deflect at least some of the hate. “Being able to support people with a similar message matters,” Joshua Garland, a fellow in applied complexity at the Santa Fe Institute and lead author of the study, said in an interview. He cautioned, though, that the research hasn’t yet clarified the full extent of the organizing tactic’s potential impact.
Garland and study co-author Mirta Galesic, a professor of human social dynamics at the Santa Fe Institute, presented their findings in a pre-recorded presentation released Oct. 19 as part of the Council for the Advancement of Science Writing’s New Horizons in Science briefings at the ScienceWriters2020 virtual conference. The Santa Fe Institute is a nonprofit research center in New Mexico dedicated to the study of complex systems.
Galesic explained in her video presentation that the traditional definition of hate speech includes “any kind of insult, discrimination or intimidation of individuals or groups solely because of their perceived political beliefs, religion, ethnicity or gender.” Current definitions have broadened, she noted, and include “any kind of spreading of harmful stereotypes or prejudice about these groups.”
One possible way to curb hate speech, censoring hateful posts, has raised ethical questions about infringing on freedom of speech. An alternative, Galesic said, is fighting hate speech with counter-speech that opposes the hateful posts by posting videos, pictures, or memes that support the victim of hate speech or confront the perpetrator.
The Reconquista battle: Organized hate draws an organized response
In their study, published as a preprint on the arXiv.org website, the researchers gathered millions of tweets by following online conversations that started as responses to articles from major news outlets in Germany and labeled them as hate or counter-speech.
By examining all of the tweets, Garland and Galesic’s team observed that the proportion of hate and counter-speech—and the number of “likes” they garnered—remained similar through most of the 2013–18 study period. That trend started to change when two new groups appeared on the internet: Reconquista Germanica, a far-right group that used hate speech online in an attempt to lure voters to the Alternative for Germany party, and Reconquista Internet, launched by German satirist Jan Böhmermann to counter the hateful messages of Reconquista Germanica.
After the official founding of Reconquista Germanica in 2017, Twitter conversations became more polarized, with neutral speech shifting towards hate speech. After Reconquista Internet formed in 2018, however, the proportion of hate speech decreased, from roughly 30% to about 25% in the analyzed sample of political conversation. At the same time, the proportion of counter-speech increased from 13% to 22%. After the emergence of Reconquista Internet, positive counter-speech received more likes and longer discussion.
The emergence of the two groups also provided a way for the researchers to refine algorithms to classify online speech. A set of 500,000 randomly selected tweets from Twitter accounts associated with each group enabled them to train classifier algorithms to recognize hate and counter-speech. The classifiers then labeled new tweets presented to them as hateful, neutral, or counter-speech, giving the researchers a way to see what kind of speech dominated each conversation.
Their results suggest that organized groups whose members support each other online could sway the Twitter conversation away from polarized discourse, the researchers said. They caution, however, that their results have not yet been peer reviewed. During an interview, Garland added that the research can’t draw firm conclusions about cause and effect, given the potential for “massive outside influences” such as political speeches, rallies, and terrorist attacks. If researchers see a jump in counter-speech and decline in hate speech after a big rally, for example, “it’s hard to say, ‘Was it caused by the counter-speech movement, or was it caused by people in Germany being sick of it?’”
Nevertheless, Garland and Galesic said their results are encouraging and suggest that creating organized groups like Reconquista Internet may help to steer online speech toward positive counter-speech and away from hate speech. The researchers are now planning a deeper dive to better understand how Twitter users can reduce hate speech online. They say some potential tactics—disarming hate with humor or actively supporting victims, for example—could go a long way toward making the internet a less hostile place.