Personal warnings can reduce hate speech on Twitter, researchers say

Clear warnings to the right account can help reduce resentment on Twitter. That’s the end of new research evaluating whether direct warnings can reduce hate speech on the platform.

Researchers at the New York University’s Center for Social Media and Politics have found that personal precautions that warn Twitter users of the consequences of their behavior reduce the number of tweets containing obscene language within a week. Although more research is needed, the experiment shows that there is a “progressive platform for platforms that seek to reduce the use of foul language by users,” according to Mustafa Mikdat Yildirim, lead author of the paper.

In the experiment, investigators identified accounts that were at risk of being suspended for violating Twitter’s anti-hate speech laws. He looked at people who used at least one word in the “hate dictionaries” last week, who also followed a story that was recently stopped after speaking such a language.

Since then, the researchers created experimental accounts with humans as “hate speech providers,” and used those accounts to warn people. They tried several different things, but they all had the same message: that using hate speech puts them at risk of being suspended, and that it has already happened to someone who is following them.

“The user of the @date you are following has been suspended, and I suspect it was due to hate speech,” he said in a statement. “If you continue to use hate speech, you may be suspended for a while.” In another transaction, the account that issued the warning identified itself as an investigator, and also informed the person that he or she was in danger of being suspended. “We tried to be as reliable and satisfying as we could,” Yildirim told Engadget.

The researchers found that the warnings were helpful, at least in the short term. “Our results show that a single warning tweet sent by an account without 100 followers can reduce the number of obscene tweets and language by 10%,” the authors wrote. Interestingly, they found that messages that were “politely spoken” dropped dramatically, dropping to 20 percent. “We tried to enhance the dignity of our message by introducing our warning by saying ‘oh, we respect your freedom of speech, but on the other hand, remember that your hateful words may hurt others,’” Yildirim says.

In the paper, Yildirim and his co-authors wrote that their trial accounts had about 100 followers each, and that they did not agree with the accredited body. But if the same kind of warning comes from Twitter alone, or an NGO or another organization, then the warnings may be more effective. “One of the things we learned from the experiment was that the real way the game was played was to let people know that there was an account, or some other group, that was monitoring and monitoring their activities,” says Yildirim. “The fact that their use of hate speech is perceived by another person may be a very important factor in reducing the number of hate speech.”