Facebook took moderation action against almost 1.5bn accounts and posts which violated its community standards in the first three months of 2018, the company has revealed.
In its first quarterly Community Standards Enforcement Report, Facebook said the overwhelming majority of moderation action was against spam posts and fake accounts: it took action on 837m pieces of spam, and shut down a further 583m fake accounts on the site in the three months. But Facebook also moderated 2.5m pieces of hate speech, 1.9m pieces of terrorist propaganda, 3.4m pieces of graphic violence and 21m pieces of content featuring adult nudity and sexual activity.
“This is the start of the journey and not the end of the journey and we’re trying to be as open as we can,” said Richard Allan, Facebook’s vice-president of public policy for Europe, the Middle East and Africa.
The amount of content moderated by Facebook is influenced by both the company’s ability to find and act on infringing material, and the sheer quantity of items posted by users. For instance, Alex Schultz, the company’s vice-president of data analytics, said the amount of content moderated for graphic violence almost tripled quarter-on-quarter.
One hypothesis for the increase, Schultz said, is that “in [the most recent quarter], some bad stuff happened in Syria. Often when there’s real bad stuff in the world, lots of that stuff makes it on to Facebook.” He emphasized that much of the moderation in those cases was “simply marking something as disturbing”.
Several categories of violating content outlined in Facebook’s moderation guidelines – including child sexual exploitation imagery, revenge porn, credible violence, suicidal posts, bullying, harassment, privacy breaches and copyright infringement – are not included in the report.
On child exploitation imagery, Schultz said that the company still needed to make decisions about how to categorise different grades of content, for example cartoon child exploitation images.
“We’re much more focused in this space on protecting the kids than figuring out exactly what categorisation we’re going to release in the external report,” he said.
Facebook also managed to increase the amount of content taken down with new AI-based tools which it used to find and moderate content without needing individual users to flag it as suspicious. Those tools worked particularly well for content such as fake accounts and spam: the company said it managed to use the tools to find 98.5% of the fake accounts it shut down, and “nearly 100%” of the spam.
Automatic flagging worked well for finding instances of nudity, since, Schultz said, it was easy for image recognition technology to know what to look for. Harder, because of the need to take contextual clues into account, was moderation for hate speech. In that category, Facebook said, “we found and flagged around 38% of the content we subsequently took action on, before users reported it to us”.
Facebook has made moves to improve transparency in recent months. In April, the company released a public version of its guidelines for what is and is not allowed on the site – a year after the Guardian revealed Facebook’s secret rules for content moderation.
The company also announced measures that require political advertisers to undergo an authentication process and reveal their affiliation alongside their advertisements.
Facebook’s moderation figures come a week after the release of the Santa Clara Principles, an attempt to write a guidebook for how large platforms should moderate content. The principles state that social networks should publish the number of posts they remove, provide detailed information for users whose content is deleted explaining why, and offer the chance to appeal against the decision.
“This is a great first step,” said Jillian York from the Electronic Frontier Foundation. “However, we don’t have a sense of how many incorrect takedowns happen – how many appeals that result in content being restored. We’d also like to see better messaging to users when an action has been taken on their account, so they know the specific violation.”
Facebook isn’t the only platform taking steps towards transparency. Last month YouTube revealed it removed 8.3m videos for breaching its community guidelines between October and December.
“I believe this is a direct response to the pressure they have been under for several years from different stakeholders [including civil society groups and academics],” said York.
Article By Alex Hern and Olivia Solon