In the USA, Researchers have now for the first time investigated on a large scale how organised counter-speech affects hate messages on the Internet. The conclusion: organized counter-speech can be an effective tool against hate on the Internet. This reports, among other things, “netzpolitik.org”.
Effectiveness little researched so far
The study is based on German-language Tweets from the networks of the right-wing troll army “Reconquista Germanica” and the counter-draft “Reconquista Internet“initiated by TV presenter Jan Böhmermann. Based on these Tweets, the researchers programmed an algorithm that can detect hate speech and counter-speech as well as their correlation.
How organized is hate on the net? The documentary follows a Team that has been undercover as trolls and haters on the net, and reports on controlled Shitstorms, bullying attacks and election manipulations.
This is a novelty, especially since records about the organized counter-speech have been missing so far. The researchers therefore focused on Germany, since here, alongside several organized right-wing Troll groups, “Reconquista Internet” was the first time that an organized counter-speech group had become active. Using more than nine million Tweets, the researchers investigated the question of whether counter-speech is useful on the Internet.
“For the first time, the study provides empirical evidence of what we have experienced in practice with ‘Reconquista Internet’: anyone who wants to successfully combat organised hatred, racist incitement or the clever discourse-shifting campaigns of right-wing extremist networks on the Internet must know how these covert manipulation networks work, analyse them and take action against them, " Böhmermann says to “netzpolitik.org”.
Counter speech is harder to recognize
Using a collected data set of the troll army “Reconquista Germanica” as well as a data set of the counter-speech group “Reconquista Internet”, the study authors trained the algorithm of their recognition system. Computer scientist Keyan Ghazi-Zahedi, who contributed to the study, says to netzpolitik.org”: “our AI system has learned what are typical statements of hate Accounts or counter-speech Accounts.”
To test the capabilities of their artificial intelligence (AI), the researchers recruited 55 people who had to rate a random selection from the collected Tweets. Neutral content formed the middle of the scale, counter speech and hate speech the extremes. According to the scientists, the assessments of the test subjects during the hate speech almost perfectly matched those of the programmed algorithm. In the counter-speech, the agreement was also strong. “This shows that counter-speech is generally harder to identify by humans and AI. We believe this could be because counter-speech is more diverse,” says Ghazi-Zahedi.
Hate and counter-speech over time
The scientists also examined how hate speech and counter-speech had changed over time. They collected around 200,000 conversations from the years between 2013 and 2018 that had emerged among the Tweets from large-scale accounts that had become the target of hate speech. The results show that the ratio of counter-speech increased significantly in April and May 2018-just in the time when” Reconquista Internet " was founded. In July 2018, the proportion of counter-speech was even higher than that of hate speech.
In addition, the researchers found that both hate speech and counter - speech were less pronounced on average in July 2018. This suggests that organized counter-speech can help to balance the hateful discourse on the Internet.
Organized counter-speech leads to more civilized discourse
The study continues to investigate the influence of the frequency of hate speech or counter-speech on the later course of Twitter conversations. According to Ghazi-Zahedi, it was found that hatred decreased with the appearance of counter-speech. There is a connection, the results would indicate that counter-speech would more often be followed by neutral speech than hate speech. Counter-speech is apparently more effective in bringing about a more civilized discourse when it is organized.
The researchers hope that their work will contribute to a better understanding of the dynamics behind hate speech and counter - speech. Data sets are now available that can also show which strategies against hate on the Internet promise success – and at what time they are best applied. “You could show, for example, when it makes sense to argue individually and with facts, or when you better counter it by ten and seek confrontation,” says Ghazi-Zahedi.
“Maybe could learn even the civil society in the highly polarized United States from the success of the Reconquista Internet,” accounted for “netzpolitik.org”.