Tech

How Facebook Troll Farms Reached 140 Million Americans

The key revelations include:

  • In October 2019 around 15,000 Facebook pages were operated with a mostly US audience from Kosovo and North Macedonia, who were known as “problematic actors” during the 2016 election.
  • Collectively, these troll farm pages – which are treated as a single page in the report for comparison purposes – reached 140 million US users monthly and 360 million weekly users worldwide. For comparison: The site of the supermarket chain Walmart reached the second largest US audience with 100 million.
  • The troll farm sites teamed up to form, among other things, the largest Christian-American site on Facebook, 20 times larger than the next largest – with a monthly reach of 75 million US users, 95 percent of whom had never followed one of the sites . They also put the largest African American page on Facebook, three times the size of the next largest, reaching 30 million US monthly users, 85 percent of whom have never followed any of the pages. With indigenous users on Facebook, 400,000 users are reached every month (the second largest page of its kind, 90 percent did not follow the page before). In terms of sites for women, the troll farms were the fifth largest with 60 million US users per month. Here 90 percent of the users did not follow the page before.
  • Troll farms primarily affect the USA, but also the United Kingdom, Australia, India and Central and South American countries.
  • Facebook has conducted several studies that confirm that content that is more likely to receive a lot of engagement from users (likes, comments, and shares) is more of what is known to be problematic. Still, the company has continued to rank content in users’ newsfeeds based on the highest engagement.
  • Facebook prohibits sites from posting content that has merely been copied from other areas of the platform, but does not enforce this policy against known problematic actors. This makes it easy for foreign actors who do not speak the national language to post completely stolen content and still reach a large audience. At one point in time, up to 40 percent of page views were from US sites that contained primarily non-original content or material of little originality.
  • Troll farms have also found their way into Facebook’s Instant Articles and Ad Breaks partnership programs, which are supposed to help real news organizations and publishers monetize their articles and videos. Due to a lack of basic quality controls, up to 60 percent of instant article reads were used for content that had been plagiarized by other sites. This made it easy for troll farms to interfere in the debate unnoticed and even receive payments from Facebook for it.

The report looks specifically at troll farms in Kosovo and North Macedonia run by people who do not necessarily understand American politics. However, due to the way in which Facebook’s “reward systems” are designed for the news feed, they could have a significant impact on political discourse. In the report, Allen gives three reasons why these sites can reach such a large audience. First, Facebook does not penalize pages for publishing completely copied content. If something has gone viral before, it will likely go viral again when posted a second time.

Second, Facebook pushes “appealing” content on pages for the algorithm to people who do not follow it. If the friends of the users comment on or share posts on one of these pages, these users will also see the content in their news feeds.

Third, the Facebook ranking system pushes content with a lot of engagement in the users’ news feeds. In most cases, troll farm operators appear to have financial rather than political motives; they post what gets the most approval, regardless of the actual content. However, since misinformation, clickbait, and politically divisive content tend to receive a high level of participation (as confirmed by Facebook’s own internal analyzes), troll farms tend to post more of them over time, the report said.