These people said the preferential treatment has undercut Facebook’s own efforts to curb misinformation, in particular the technologies put in place to downgrade problematic actors. Toward the end of last year, around the time Facebook-owned Instagram was rolling out labels obscuring fact-challenged posts and directing users to accurate information, the company removed a strike against Donald Trump Jr. for a fact-check on the photo-sharing service that would have made him a so-called repeat offender, fearing the backlash that would have ensued from the accompanying penalties, according to two former employees familiar with the matter.
These penalties can be severe, including reduced traffic and possible demotion in search. One former employee said it was among numerous strikes removed over the past year for the president’s family members.
A spokesman for Donald Trump Jr. did not respond to multiple emails seeking comment. Facebook spokeswoman Andrea Vallone did not dispute the detail, saying the company is “responsible for how we apply enforcement, and as a matter of diligence, we will not apply a penalty in rare cases when the rating was not appropriate or warranted under the program’s established guidelines.”
The kid-glove treatment contradicts claims of anti-conservative bias leveled by Trump and his children, as well as by Republican leaders in Congress. It also renews questions about whether Facebook is prepared to act against the systematic spread of falsehoods that could intensify as vote tallies are reported this week.
Facebook, in a bid to avoid previous election missteps, has issued a slew of new policies, including limits on political ads and rules against premature assertions of victory. But the current and former employees say the company’s four-year-old fact-checking program, introduced in response to the flood of fake news that marred the 2016 election, has failed to constrain the most prolific purveyors of false and misleading content.
The program relies on independent fact-checkers rather than involving the company in judgments about the veracity of content, and it makes available a range of ratings for dubious material, from false to partly false to missing context — which are labeled on the offending posts. Exceptions and political considerations, however, shape the consequences that these ratings trigger, according to people who have been involved in the program’s execution in the run-up to the election, consistently steering the company toward less robust enforcement.
Fact-checking at Facebook has faltered, said Mike Ananny, an associate professor at the University of Southern California who completed a 2018 review of the program, “because its business model requires a scale and speed and level of engagement mismatched to controlling misinformation.”
One of the people familiar with internal deliberations said some efforts to improve fact-checking and content review have been stymied by concerns about a disproportionate impact on conservative users. Members of Facebook’s public policy team recently floated a proposal ensuring that a new system to escalate harmful posts do so evenly along ideological lines, the person said, so that 50 percent of the escalated material would be conservative and 50 percent would be liberal, even if the material was not equivalent in potential risk.
“No such policy exists,” Facebook’s Vallone said.
But the person who was involved in the discussions said the idea showed how efforts to combat misinformation are viewed internally as a political liability. “Too often we’ve made politically expedient exceptions at the expense of our own rules, which we generally believe to be fair," the person said.
Delayed and uneven enforcement of the company’s rules is evident in particular on prominent right-leaning Facebook pages involved in sharing news about the election. More than a dozen such pages identified by The Washington Post shared content debunked by Facebook’s own third-party fact-checkers twice within 90 days over the last six months, meeting the definition of repeat offenders described by multiple people familiar with the company’s process and backed up by internal communications. But many of these pages were still attracting significant engagement and still purchasing ads, despite rules for repeat offenders that prescribe significant penalties, including reduced distribution of content and the revocation of advertising privileges.
Some had clearly violated Facebook’s two-strike rule for “false information,” the strictest rating available. Several had three or more fact-checks against them within 90 days, though some were lesser ratings of partly false or missing context.
The largest outside group supporting Trump’s reelection, for example, has repeatedly posted material judged as false by Facebook’s third-party fact-checkers. The false claims circulated by the group, America First Action, involve hot-button domestic policy issues core to the presidential campaign. One video, accusing former vice president Joe Biden of seeking to defund the police when he has in fact resisted that call, was labeled as false. A different video leveling the same claim, posted three days later, earned an identical label.
Even though it received two false ratings within 90 days, in addition to repeated fact-checks applied to posts about Biden’s energy agenda and tax plan, America First Action is still able to advertise, according to Facebook’s public archive. There is no evidence that its distribution has been reduced, according to engagement data from the social media analysis tool CrowdTangle and Facebook fact-checking partners who spoke on the condition of anonymity because the company is their client.
“I’m baffled by the policy,” said the head of one such organization, singling out America First Action for getting away with repeat offenses. “We repeatedly flag offenders that nevertheless seem to prosper and continue to do ads.”
The super PAC did not respond to a request for comment. Vallone declined to comment on the status of America First Action’s page or of any other. She also declined to make anyone from the fact-checking or news integrity team available for an interview. She maintained that “many” of the pages inquired about by The Post “have been penalized for repeatedly sharing misinformation in the past three months." She did not specify the penalties or to which pages they had been applied.
“We don’t disclose the details of these thresholds publicly for very real concerns about gaming the system, but we do send notifications to groups, pages, accounts and advertisers when they’ve received a strike and are receiving reduced distribution, and when they are a repeat offender," Vallone said. She also defended the fact-checking program, saying it makes Facebook the “only company that partners with over 80 fact-checking organizations to apply fact-checks to millions of pieces of content.”
The fear of appearing biased against Trump and other conservatives runs up to the highest levels of Facebook, and has shaped everything from the algorithm deciding what appears in the News Feed to the process of reviewing potentially harmful content. Allegations of preferential treatment in the fact-checking process leaked into public view this summer, when a Facebook engineer published information, first reported by BuzzFeed, showing that company managers were intervening on behalf of right-leaning publishers.
Meanwhile, conservatives gain some of the largest audiences of any publishers on Facebook, a trend that has continued through the election. Over the last week, the top political Facebook pages receiving the largest increases in views were right-leaning pages, including Donald Trump, Fox News, Breitbart News, along with Joe Biden’s Page and the left-leaning NowThis Politics, according to an internal report on traffic viewed by The Post.
Only by disclosing the underlying data on fact-checks and the consequences they yield, said Matt Perault, a former director of public policy at Facebook who now runs Duke University’s Center on Science and Technology Policy, can the company “show that it’s enforcing its policies consistently.”
As Facebook declines to count certain fact-checks as strikes against a page or user, other posts pushing dubious claims — already addressed by Facebook’s third-party fact-checkers — are not even getting labeled. The pattern is stark for some pages, including one operated by talk radio host Rush Limbaugh, who boasts more than 2.3 million followers.
In August, Limbaugh publicized the claim that Anthony S. Fauci, the nation’s leading infectious-disease expert, owns half the patent for a coronavirus vaccine. One of Facebook’s third-party fact-checkers debunked the claim, but no label was applied to Limbaugh’s post, which has gained more than 17,000 shares, comments and likes.
Later the same month, Limbaugh shared a link to a story on his website questioning whether Biden had delivered his convention speech live or in fact prerecorded it — a conspiracy theory that gained traction among some right-wing commentators. One of Facebook’s third-party fact-checkers debunked the claim, but no label was applied to Limbaugh’s post. Limbaugh did not respond to a request for comment.
The same pattern can be observed in Facebook’s treatment of right-wing blogger Pamela Geller, whose page racked up a false rating last month but, she claimed, has not been penalized. A post this month, amplifying a news story debunked by one of Facebook’s third-party fact-checkers, was never labeled. Geller deleted the post following an inquiry this week and said in an email she had received no notification from Facebook about repeat offender penalties.
Facebook’s Vallone declined to disclose the average time it takes a post to get labeled but said “we do surface signals to our fact-checking partners to help them prioritize what to rate."
Some users do say they have been punished for what Facebook claims are repeated falsehoods. Peggy Hubbard, a former Republican congressional candidate who posts pro-Trump memes to her more than 350,000 followers, wrote in an email that she had been “locked out of all accounts.” And in August, Facebook barred one pro-Trump super PAC, the Committee to Defend the President, from advertising following what a company spokesman, Andy Stone, called “repeated sharing of content determined by third-party fact-checkers to be false.”
At least four times in July and August, the Gateway Pundit, a right-wing news site recently cited in a congressional hearing as a victim of anti-conservative bias, posted stories rated as false or misleading by Facebook’s independent fact-checkers.
The site publicized the claim that Fauci would “make millions” from a coronavirus vaccine (he will not); that a Democratic fundraising platform was making payments to Black Lives Matter protesters (the platform, ActBlue, denied this); that the common cold was being treated as a covid-19 positive result (it was not); and that Sen. Kamala D. Harris (D-Calif.), the vice-presidential nominee, was haunted by a “dark secret” that her ancestors owned enslaved people (the ancestral fact is common to many African Americans).
Some of these stories were later modified, avoiding sanction by Facebook. But false posts are shared rapidly, while fact-checks are slow to be applied. So, too, false posts frequently outperform true ones. In the case of the Gateway Pundit, an examination of more than 800 posts this summer found that those labeled as false or misleading on average earned nearly 50 percent more likes, comments and shares than were garnered by the page’s overall posts in the same period.
Jim Hoft, the creator and editor of the Gateway Pundit site, did not respond to a request for comment about what penalties, if any, his Facebook page has faced. Pinned to the top of the site’s page on Facebook are instructions for its more than 616,000 followers to ensure they can see posts from the Gateway Pundit “at the top of your News Feed.”