Facebook and its policies have been in the public eye for quite some time now, and recently, whistleblower Frances Haugen called out the company for knowing how Instagram was affecting teen users and still not doing much to prevent negative impact. The Wall Street Journal recently published an article claiming Facebook employees that work to keep offensive content off the social media platform are concerned there is no reliable screen for it. Now, Facebook’s VP of Integrity has responded to this claim, reports The Verge.
Hate speech prevalence reportedly dropped by 50%, claims Facebook
Facebook vice president of integrity Guy Rosen wrote in a blog post that the prevalence of hate speech on the social media platform has dropped by almost 50% over the past three years. He additionally claimed that the statement the technology that Facebook uses to fight hate speech is inadequate is false. According to his blog post, the accusation that Facebook is deliberately misrepresenting its progress on the matter is also false. Additionally, he said that Facebook doesn’t want hate on its platform, nor its users or advertisers, and he added that the company has been transparent about its work to remove offensive content. He adds that the integrity work takes multiple years and that although the prevention of hate speech tech will never be perfect, Facebook’s teams are working to continually develop their systems, identify issues, and build solutions. This post comes after an article by the Wall Street Journal in which Facebook employees that work on keeping offensive content off Facebook have stated that there is no reliable screening for it. The report on the WSJ states that according to internal documents from two years ago, Facebook reduced the time human reviewers were focusing on hate speech complaints.
Additionally, according to the report, the social media giant has made other adjustments to reduce the number of such complaints, so this would create the appearance Facebook’s AI systems have been more successful in enforcing the rules than what was the reality.
According to the post, Facebook employees found that AI systems were removing posts that generated between 3 and 5% of the views on hate speech the social platform had, and less than 1% of all content that was violating the rules against violence and incitement. This report basically means Facebook’s AI, according to the employees that worked on the matter over at the social media company, was not removing as many guidelines-violating posts as needed.
However, Facebook’s VP of integrity Rosen stated that focusing on content removals alone was the “wrong way” to look at the fight against hate speech. He stated in the blog post that removing the content is just one method Facebook uses to fight it, and he added that they need to be certain something is hate speech before it gets removed.
One other method of fighting against violence-inciting content or hate speech on the platform is focusing on the prevalence of hate speech people actually see on Facebook. Facebook has tools that reduce how much hate speech content people see, and Rosen stated that for every 10,000 views of content on Facebook, there are only 5 views of hate speech content.
According to him, the prevalence of hate speech content on Facebook is now about 0.05% of all the viewed content.
On the other hand, documents obtained by the Wall Street Journal show some significant pieces of content that had evaded Facebook’s detection, such as car crashes with graphic injuries, and violent threats to minority groups.
These reports come after ex-Facebook employee Frances Haugen testified before Congress that Facebook was aware of the negative impact it and Instagram are causing to teens and was not doing enough to solve the issue.
