SAN FRANCISCO — For more than two years, Facebook chief executive Mark Zuckerberg has touted membership in private, “meaningful” communities as central to the future of the social network.
But that strategy of pushing closed interactions within groups is increasingly butting up against the company’s capacity to monitor those that break the social network’s rules against hate speech, harassment and other ills, civil rights groups and other advocates say. They warn that the company’s move into more private communications is on a collision course with Facebook’s stated goal of cleaning up its platform ahead of the 2020 U.S. presidential election.
Facebook’s failure to monitor problematic groups surfaced again this week after an investigation by ProPublica revealed that a secret, members-only group of current and former Border Patrol agents joked callously about the deaths of migrants and used a vulgar illustration of Rep. Alexandria Ocasio-Cortez (D-N.Y.) being forced to engage in a sexual act by President Trump. The company said Friday that it also removed several problematic posts from another Border Patrol group.
The pattern of not policing private groups — which Facebook says are subject to the same rules as other parts of the social network — has become a bigger concern as the company pushes those channels as a key piece of how it wants people to use its site.
“Large private groups remain unmoderated black boxes where users can freely threaten vulnerable populations,” said Jonathan Greenblatt, chief executive of the Anti-Defamation League. “Without any AI or human moderators, it’s easy to orchestrate harassment campaigns — at minimum, this environment contributes to the normalization of bigotry and discrimination. As Facebook moves to more and more private communication, we’re concerned about this delinquency.”
Zuckerberg has been pushing a plan to transform Facebook into a group-oriented platform that would transition it from a “town square” of public news feeds into more of a “living room” designed around personal interactions with family and friends. In the past two years, he has published two lengthy manifestos — the first in 2017 on a transition toward meaningful groups and the second this year on privacy. He recently redesigned the Facebook desktop app to suggest ideas for new groups people could form through the site, as well as making it easier to create them. Facebook’s algorithms already actively suggest potential new groups Facebook users can join, steering people into groups steeped in misinformation.
About 1.4 billion people log into a Facebook group monthly, the company said. About 400 million people now consider themselves a part of groups they say are “meaningful,” four times the number two years ago, according to company estimates.
Unlike public posts from Facebook pages, which have been examined extensively by outside experts, researchers have more trouble monitoring closed groups because an administrator must give permission to become a member. Facebook’s rules also allow administrators to designate a group as secret, which means it does not show up when people search for it and members must be invited in. That’s true of the private Border Patrol group, called “I’m 10-15,” referring to the agency’s code for “aliens in custody.”
It is not Facebook’s first run-in with private groups that violate its policies. Facebook banned the conspiracy theorist media star Alex Jones last year, but private groups with thousands of members continued to promote his work. Groups dedicated to the pro-Trump conspiracy theory QAnon thrive on Facebook, as do communities that oppose vaccination. In 2017, the military launched an investigation of a secret Facebook group composed of Marines who had shared naked photos of fellow female service members. Private and public groups also played a significant role in helping white supremacists organize their march in Charlottesville in 2017.
Facebook declined to answer questions about the 10-15 group, citing its cooperation in an ongoing federal investigation. The company also said it removed posts from another group identified first by CNN as “The Real CBP Nation” for violating its policies including bullying and sexual exploitation of adults, but is allowing the group to stay up.
“We want everyone using Facebook to feel safe. Our Community Standards apply across Facebook, including in secret Groups. We’re cooperating with federal authorities in their investigation,” spokeswoman Jen Ridings said in a statement. She declined to answer questions about whether the group’s information was still up or the nature of Facebook’s cooperation with law enforcement.
She added that the company uses a combination of technology and human review to remove rule-breaking content, including in secret groups.
“There is still more we can do, and we continue to improve our technology to detect violating content,” Ridings added.
The controversy over the Border Patrol group is raising alarm among some members of Congress, where the company is already under scrutiny. On Tuesday, House Oversight Committee Chairman Elijah E. Cummings (D-Md.) said his committee would conduct its own investigation of the 10-15 group, and noted that the posts appeared to violate Facebook’s guidelines. Ridings declined to say whether Facebook would cooperate with the House investigation.
U.S. Customs and Border Protection investigated the 10-15 Facebook group in 2016 and took some disciplinary action, said an official, who spoke on the condition of anonymity to discuss a personnel investigation. The official declined to say what action was taken. In general, the official said that it does not monitor employees’ speech and therefore did not see monitoring the group as in the scope of their responsibilities.
In response to the crisis over Russian interference on its platform, Facebook has dedicated major resources to detecting misinformation, hate speech and other violations of its standards. The company employs 30,000 workers in safety and security, including 15,000 content moderators, and its engineers have improved the company’s software that detects problematic content. The company is also in the process of undergoing a major audit from civil rights groups and has solicited help from disinformation researchers and fact-checking organizations.
Despite the resources Facebook has invested in cleaning up its service, private and secret groups continue to thwart the company’s detection capabilities. While artificial intelligence plays a big role in detecting certain types of content, such as terrorism and nudity, many posts featuring hate speech and political disinformation still reach Facebook’s content moderators primarily through user reports, according to Washington Post interviews with more than a dozen Facebook content moderators. Facebook says that of the hate speech it does catch, 65 percent of it is detected by software, compared with more than 95 percent for terrorism and nudity.
Misinformation experts have repeatedly pointed out that people in like-minded groups are less likely to flag one another’s content as problematic, reducing the reporting from such groups.
As part of its push for privacy, the social network also plans to encrypt features users see as private, such as messaging, meaning the data would be scrambled so that outsiders, including law enforcement and even Facebook’s own engineers, couldn’t read it.
When advocates and lawmakers complained that this move could make it harder to clean up misinformation and hate, Zuckerberg acknowledged the trade-offs. But he said doing so was worth it in the long term because users show a preference for private interactions.
“We still have a responsibility to make sure that we do a lot to keep people safe,” Zuckerberg told The Washington Post in an interview in April. “But I think when faced with a challenge around a trade-off between encryption and safety, I think people would want us to err a little bit more on the side of encryption. That’s certainly where I personally come out on that.”
Reis Thebault and Nick Miroff contributed to this report.