Customized Twitter alert reduces hate speech
Customized Twitter alert reduces hate speech

A carefully designed set of alerts for the right accounts can help reduce hate on Twitter, according to a new study to see if targeted alerts can reduce hate speech on the platform.

Researchers at the New York University Center for Social Media and Policy found that personal warnings alerting Twitter users to the consequences of their actions reduced the number of tweets containing hate speech after a week.

Mustafa Yildirim, lead author of the article, said that while more research is needed, experience has shown that there is a way forward for platforms seeking to reduce user use of hate speech.

In the experiment, researchers identified accounts that could be banned for violating Twitter's hate speech rules. They searched for people who used at least one word in the "hatekers dictionary" in the past week and for people who followed at least one account that was recently banned for using that language.

The researchers created test accounts with warning signs of hate speech and used these accounts to send alerts to these people.

I have tested a lot of variations. But they all have pretty much the same message: Using hate speech runs the risk of getting commented, and so do the people you follow. The account you're following has been suspended. I think it's because of hate speech. If you continue to use hate speech, you could be banned.

The account that issued the warning stated that he was a professional researcher. It also tells the person that he is threatened with suspension. “We try to be as convincing and believable as possible,” Yildirim said.

Literary warnings on Twitter are more effective

Researchers have found that these warnings are effective, at least in the short term. “Our results show that a warning tweet sent from an account of less than 100 followers can reduce the proportion of hate tweets by up to 10%,” the author wrote.

They also found that politely designed messages led to an additional reduction of up to 20%. “We tried to raise morale in our messages by starting mainly with our warnings,” Yildirim said.

In the paper, Yildirim and his colleagues noted that each demo account has about 100 subscribers. It is also not associated with trusted entities. However, if the same type of warning is coming from Twitter itself, or from an NGO or other organization, the warning may be more useful.

Previous Post Next Post