To evaluate Twitter’s success in fighting hate speech on the platform, it helps to have an accurate assessment of the situation. A study by independent company Sprinklr has now found that the reach of hate speech on Twitter is even lower than the platform’s own model quantified.
As Elon Musk promised, Twitter 2.0 has become more open and transparent. Among other things, the company shares its ratings of hate speech to give a real picture of what happens during communication on the platform. While some journalists and other loud voices said hate speech skyrocketed after Twitter was bought by Musk, the facts show otherwise.
Back in December 2022, the platform released information about its recent efforts to reduce hate speech coverage. Twitter Safety clarified that counting the number of tweets containing particular slurs is not an accurate way to measure hate speech. The context of each statement matters, and obviously not all slurs are used with hatred. After doing the research, Twitter saw that there was a steady downward trend on the platform in the number of experiences expressing true hate speech (read more about this here).
In order to continue to monitor its progress in this area, Twitter announced that it recently partnered with Sprinklr for an independent evaluation of hate speech on Twitter. While the full data will be published later, the platform shared the overall results without giving specific numbers.
Sprinklr’s AI-based model showed that the reach of hate speech on Twitter is even lower than the platform’s own model. The difference was in the context of the conversations that included swear words and how both companies defined tweet toxicity. Twitter clarified that Sprinklr defines hate speech more narrowly by evaluating slurs in the nuanced context of their use. On another side, the platform has looked more broadly at the potential toxicity of using insults.
To quantify hate speech, Twitter and Sprinklr start with the 300 most common English-language slurs. They count not only how often they are used in tweets, but also how often they are seen by other users—that is, how many impressions they get. Most of the time, the use of swear words on Twitter is not hate speech, research shows. At the same time, the platform is actively working to reduce the reach of tweets containing swear words. Sprinklr’s analysis found that hate speech gets 67% fewer impressions per tweet than non-toxic slurs.
Twitter assured that it continues to combat hate speech, using other languages, new terms, and more precise methodologies, while increasing transparency. Therefore, over time, the platform will become an even safer place for users.
Verum News 