:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/59493121/144717349.jpg.0.jpg)
Facebook is responsible for writing and enforcing content rules that all users have to adhere to — basically a code of conduct for what is and isn’t allowed on the service.
Writing these rules can be tricky. Some stuff is obviously inappropriate, like terrorist content or child pornography, for example. But other stuff is tougher to categorize and enforce across a global user base. What’s considered hate speech by one group of people is considered free speech by another.
It’s why deciding what’s allowed and what isn’t allowed makes Facebook CEO Mark Zuckerberg “fundamentally uncomfortable.”
Uncomfortable or not, it’s Facebook’s job. And on Tuesday, the company made an interesting move: It published the exact set of rules that Facebook employees and contractors use to decide what is allowed and what isn’t.
The idea, according to Facebook executives, is to give people a better understanding of why stuff is taken down so that there is less confusion (or anger or frustration) when some people disagree with whatever decision the company makes.
“This document mirrors the guidelines that are given to reviewers internally,” said Mary DeBree, Facebook’s head of content policy. “It is as much as possible that we can put out externally.”
Facebook is also rolling out a new appeals process so that anyone can appeal the removal of their post or photo.
Facebook’s global head of policy Monika Bickert explained that these new guidelines may look different — they are way longer and more detailed, for example — but that Facebook isn’t enforcing anything differently than it has in the past.
Here’s one example of how the new policy will look different for users. Facebook’s old Community Standards described a “direct threat” in four sentences:
We carefully review reports of threatening language to identify serious threats of harm to public and personal safety. We remove credible threats of physical harm to individuals. We also remove specific threats of theft, vandalism, or other financial harm.
We may consider things like a person’s public visibility or the likelihood of real world violence in determining whether a threat is credible.
The new set of standards outlining what Facebook considers a “threat” runs an entire page and a half.
Facebook’s role in policing content has become a big story over the past 18 months. After the 2016 U.S presidential election, in which Russian trolls used the service to spread so-called fake news and divisive content, it became clear that Facebook’s policies and moderation were letting too many things slip through the cracks.
So Facebook decided to beef up its content-review operation, pledging to have 20,000 employees working on safety- and security-related projects by the end of 2018. Bickert says Facebook already has 7,500 content reviewers worldwide, a mix of both full-time employees and contractors.
The company is also making plans to fight this stuff with technology. When asked about these topics earlier this month during his Congressional testimony, CEO Mark Zuckerberg routinely talked about the company’s use of artificial intelligence as a way that it hopes to better police user content.
That may be the case down the line, but Facebook still uses human moderators for the vast majority of its content decisions. AI works for removing known child pornography or terrorist beheading videos, for example, but isn’t used to determine what might be considered hate speech.
“There are some limited cases where the technology itself can remove the content without a person looking at it,” Bickert said. “By and large, most types of content policy violations — hate speech, bullying harassment, threats of harm — most of that has to be reviewed by people at this point because it is just so contextual.”
The new standards will roll out Tuesday to all Facebook users.
This article originally appeared on Recode.net.
