‘A constructively critical relationship’: Facebook’s fact-checking program shows improvement

In the last month, Facebook started sharing reports with AFP on five countries that include data like the number of domains that have been taken down and the number of notifications sent to users as a result of AFP’s work. It’s part of the platform’s pledge to share quarterly reports that include customized stats that reflect the work and impact of each fact-checker.

Still, there’s more Facebook can do, including explaining whether advertising was stopped because of fake news. While the platform is showing willingness to listen, it’s not easy for it to open the hood.

“Don’t underestimate the difference between the online journalism culture and that culture of a data-business,” said Chetwynd.

Others are finding similar sticking points. Full Fact, a U.K. fact-checking charity, which started working with Facebook last month when the platform rolled out its fact-checking program to the U.K., is also leaning on Facebook for more data.

“We want Facebook to be sharing data transparently and more widely. It’s clear Facebook can share more information,” said Will Moy, director at Full Fact. “We’ll be telling them that is what we expect a responsible internet company to do.”

So far, Full Fact has fact-checked just 10 stories on Facebook, including debunking a picture that was shared over 25,000 times of a horse living in a flat in Preston, Lancashire, which turned out to be a picture of a model horse in a window in Illinois, U.S. More pernicious claims include an false image stating that illegal immigrants and refugees can claim a much larger yearly benefit than British pensioners, which was shared over 2,000 times on Facebook.

Every three months, Full Fact will publish a report on its processes, how effective the program has been and how the program needs to improve. Until then, Full Fact is figuring out the right processes in how it debunks stories, like how much time and resource to dedicate to it — currently an editorial team of six reviews content — and guidelines on which stories to review. Three topic areas where it will intervene are during the aftermath of terrorist attacks, in the lead-up to elections and content around health.

How much leverage fact-checking agencies and charities will have on getting Facebook to open up is questionable. In January, two fact-checking agencies in the U.S., Snopes and Associated Press which were being paid by Facebook, announced they were ending their part in the program for now but hope to work with it again. In the U.K., The Cairncross Review, which explores the future economic model of news, released recommendations that independent regulators should supervise platforms’ efforts to improve users’ news experience, including identifying unreliable sources.

Facebook faced criticisms, mainly for not showing its workings but also exploiting a PR opportunity, allegedly putting journalists in the firing line and pointing fact-checking agencies toward certain topics, which the platform denies.

During the yellow-vest protests in France last November, which saw a lot of misinformation shared, Chetwynd saw that once it debunked stories as false, they stopped being shared. Facebook can duke the algorithm to reduce future impressions of something that is “fake” by an average of 80 percent.

Facebook still works with 39 fact-checking partners globally across 20 languages, among other initiatives to help fight fake news, like machine learning.

“Fact-checking was never going to be one-size-fits-all,” said Chetwynd. Instead, there’s lot of different misinformation, and it needs to be tackled in different ways.

“It’s not just about resource; that is a useful metric, but how well you diagnose the problem and interventions put in place,” said Moy. “Facebook will benefit from sharing more of their data. It will need to justify what it can and cannot share.”