Troll farms peddling misinformation on Facebook reached 140 million Americans monthly ahead of the 2020 presidential election, report says

Advertisement
Troll farms peddling misinformation on Facebook reached 140 million Americans monthly ahead of the 2020 presidential election, report says
Content moderators work at a Facebook office in Austin, Texas. Ilana Panich-Linsman/Getty Images
  • Troll farms were building massive audiences and peddling propaganda ahead of the 2020 presidential election, a report says.
  • Facebook's design and algorithms helped spread the troll farm content, MIT Technology Review said.
  • The social media platform has struggled to squash disinformation campaigns since the 2016 presidential election.
Advertisement

Troll farms creating propaganda and misinformation on Facebook reached about 140 million Americans a month in the run-up to the 2020 presidential election, according to an internal company report obtained by MIT Technology Review.

The report came out in 2019 as part of an almost two-year effort to understand troll farm activity on the platform and was recently given to MIT Technology by a former employee not involved in researching it.

The social media company failed to clamp down on misinformation and troll farm content after the 2016 election, the report found.

Complimentary Tech Event
Transform talent with learning that works
Capability development is critical for businesses who want to push the envelope of innovation.Discover how business leaders are strategizing around building talent capabilities and empowering employee transformation.Know More

The company "pursued a whack-a-mole strategy that involved monitoring and quashing the activity of bad actors when they engaged in political discourse, and adding some guardrails that prevented 'the worst of the worst,' MIT Technology Review said.

By 2020, troll farms, or groups that work together to create and disseminate false information online, "were still building massive audiences by running networks of Facebook pages," the MIT report said.

Advertisement

15,000 pages viewed by a US-majority audience were run out of Kosovo and Macedonia, who engaged in disinformation campaigns during the 2016 election. Collectively, the troll farm pages reached 140 million US users monthly and 360 million global users weekly in late 2019 as Americans prepared to vote in one of the most frought US presidential elections in history.

"This is not normal. This is not healthy," wrote Jeff Allen, a former senior-level data scientist at Facebook who authored the report. "We have empowered inauthentic actors to accumulate huge followings for largely unknown purposes."

The troll farm pages churned out content for the largest Christian American page on the site reaching 75 million US users a month, the largest African-American page at 30 million users a month, the second-largest Native American page at 400,000 users a month, and the fifth-largest women's page on the site at 60 million users a month.

A majority of the users who viewed this content had never followed these pages.

Facebook did not immediately respond to Insider's request for comment about the report. Joe Osborne, a Facebook spokesperson, told MIT Technology Review the company has "stood up teams, developed new policies, and collaborated with industry peers to address these networks. We've taken aggressive enforcement actions against these kinds of foreign and domestic inauthentic groups and have shared the results publicly on a quarterly basis."

Advertisement

Since 2016, bad actors have successfully spread American election conspiracies and COVID-19 misinformation multiple times. Politicians and regulators have criticized Facebook and other platforms' inability to limit foreign interference. Researchers and technology rights advocates have provided their own resources to battle disinformation online, but most companies have opted to use their in-house misinformation algorithms.

Facebook itself has tried to provide more transparency in their content moderation practices, but their current approach - using algorithms to flag possible bad content and having human reviewers look at this content case-by-case - has been criticized as more of a band-aid solution than a permanent one.

{{}}