Facebook-Created Fake Users Proved Algorithm Radicalized People in 2019

Leaked documents shared NY Times, CNN, NBC and others also show Facebook was unprepared for the spread of lies and conspiracy about the 2020 Election

facebook logo
Getty Images

In 2019, researchers at Facebook started creating fake accounts as part of an experiment to test how the social media app’s algorithm promoted disinformation and polarization. The result: Facebook ended up with incontrovertible proof that it contributed significantly to radicalization, particularly of conservatives.

Worse, almost as soon as it was happening, Facebook employees were aware that the algorithm was increasing the scope and reach of right wing lies and conspiracy theories about the 2020 election in the weeks after Joe Biden’s victory over Donald Trump. But worried that Facebook wasn’t doing remotely enough to stop it.

The news comes from several damning articles published Friday night by a consortium of media outlets who received leaked internal Facebook documents from attorneys representing Whistleblower Frances Haugen. This includes NBC, CNN and the New York TImes.

One example is “Carol Smith,” a “conservative mom” from South Carolina who joined Facebook in 2019. In reality, Carol was a faked account created by one of the researchers. And according to NBC, “in just two days,” “Carol” started receiving recommendations to join QAnon-affiliated groups.

The researcher running the “Carol” account didn’t respond to those prompts, NBC said, but nevertheless Facebook continued to force extremist right wing content to the account. The “Carol” account’s feed soon became “a barrage of extreme, conspiratorial, and graphic content,” researchers said.

The information was presented to Facebook leaders in a report called “Carol’s Journey to QAnon.”

Meanwhile, documents show that Facebook employees were attempting to bring attention to the spread of propaganda related to “Stop the Steal,” the movement based around lies and conspiracy theories that the 2020 election was stolen by Democrats.

According to the Times, just 6 days after the 2020 election one of Facebook’s data scientists had discovered “that 10 percent of all U.S. views of political material… were of posts that alleged the vote was fraudulent.”

And according to CNN, the Times and NBC, similar warnings were raised by multiple other employees.

A subsequent analysis of the spread of such propaganda found “the policies and procedures Facebook had in place were simply not up to the task of slowing, much less halting, the ‘meteoric’ growth of Stop the Steal,” according to CNN. 

Part of the problem, the leaked documents explain, is that Facebook insisted on treating every instance of conspiratorial or extremist content individually, instead of as part of a larger movement. “Almost all of the fastest growing FB Groups were Stop the Steal during their peak growth. Because we were looking at each entity individually, rather than as a cohesive movement, we were only able to take down individual Groups and Pages once they exceeded a violation threshold,” internal documents said.

According to all three outlets, Facebook did eventually attempt to do something about these problems, but many employees worried the effort was too little, and too late. CNN expressly compared this to how Facebook failed to address the problem of “coordinated inauthentic behavior,” primarily from Russian sources, during the 2016 election.

However, the internal documents do confirm that after the Jan. 6 insurrection by a mob of Trump supporters attempting to overthrow the government, Facebook took direct action to stop the growth of affiliated groups.