Twitter now removes half of tweets containing abusive content automatically without relying on content moderators or users reporting them.
The social media platform’s 2019 third-quarter report revealed that 50% of those tweets that were removed due to abusive content were done so through improvements in machine learning.
“We also continue to make progress on health, improving our ability to proactively identify and remove abusive content, with more than 50% of the Tweets removed for abusive content in Q3 taken down without a bystander or first-person report,” Twitter CEO Jack Dorsey said in the report.
In comparison, Twitter took down 38% of abusive content proactively in the first quarter of this year and 43% in the second quarter, according to Venture Beat.
Twitter defines abusive content as “an attempt to harass, intimidate, or silence someone else’s voice.”
The social network first rolled out a policy for hate speech in 2017, which included prohibiting and removing accounts affiliated with organizations that promote violence, as well as deleting tweets that glorify violence and permanently suspending users who continue to violate the policies.
Twitter asked its users for feedback on its hateful conduct policy in September 2018, and according to a July blog post, the company received more than 8,000 responses. Some of the feedback asked for clearer language on examples of violations, narrowing down what is considered a hate group, and having more consistent enforcement for hateful speech on Twitter’s end.
In July, Twitter updated its rules against hateful content to include any tweets that dehumanized people based on religion. Twitter already removes tweets with threats based on race, ethnicity, religious affiliation, and more, but July’s update set more precise standards for hateful content based on religion.
Like Twitter, Facebook also uses artificial intelligence machine learning to remove content or accounts that are considered harmful proactively. Facebook uses “coordinated inauthentic behaviour” as a red flag to ban accounts on its platform. The term is used to describe “when groups of pages or people work together to mislead others about who they are or what they are doing.”
