The Washington PostDemocracy Dies in Darkness

Rapidly expanding fact-checking movement faces growing pains

More than 200 people gathered in Rome for the fifth international fact-checking conference. (Giulio Riotta)

ROME — Political fact-checkers from more than 50 countries gathered here to take stock of a fast-growing journalistic movement that has gained clout and influence while attracting criticism and heightened skepticism in an increasingly partisan age.

Facebook has enlisted 24 fact-checking organizations in 14 countries to help weed out fake news on the social network, while policymakers and parliamentarians in Brazil, Italy and Spain, and at the European Union, have sought advice from fact-checkers on the challenge of misinformation. Google now highlights fact checks in its search results and Bing has developed a special fact-checking page that features recent fact checks.

But fact-checkers have increasingly come under attack, facing accusations of bias and partisanship that the neutral journalistic format was supposed to avoid.

“A dark cloud hangs over us,” said Alexios Mantzarlis, director of the International Fact-Checking Network (IFCN), the umbrella organization that organized the meeting, when he opened the three-day conference on June 20. “The disaffection and distrust that have plagued mainstream media outlets for many years is now spilling over to fact-checkers. In Turkey, the Philippines and especially Brazil it broke out in the form of concerted campaigns aimed to vilify fact-checking as an instrument.”

Only four years ago, some three dozen fact-checkers met for the first time in London, in a small college classroom, hoping to spark greater global cooperation. That meeting led to the creation of the IFCN, which is housed at the Poynter Institute.

The number of fact-checking projects now stands at 149 in 53 countries, according to a count in February by the Duke University Reporters’ Lab. That’s triple the number recorded four years ago, but the figure is already out of date. Two fact-checking organizations have opened in Panama in recent months, for instance.

A meeting of fact-checkers in Buenos Aires two years ago resulted in the creation of a code of principles for the movement, which laid out guidelines on a commitment to political neutrality, transparency on sourcing and funding and robust correction policies. The code gained new importance after Facebook decided it would only enlist fact-checkers who are annually assessed to be in compliance with the code. A European Commission report in April on misinformation highlighted the IFCN code as a model that should be followed. “A dense network of strong and independent fact-checkers is an essential requirement for a healthy digital ecosystem,” the report said.

Currently, there are 57 verified signatories of the IFCN code. Yet for all the increased clout, fact-checkers in many countries operate with few resources and shoestring budgets. Mantzarlis noted that of the 42 IFCN verified fact-checkers attending the conference, 26 operated with a budget of $100,000 or less in 2017.

The Rome meeting, the fifth “Global Fact Summit,” brought together 230 people, 70 percent of whom were fact-checkers and the rest academicians, experts and representatives of Facebook, Google, WhatsApp and other technology companies. Attendance was capped after almost 1,000 people applied to attend, signifying the global interest in the fact-checking movement.

“Fact-checkers are no longer the fresh-faced journalistic reform movement pushed forward by the tail winds of positive expectations,” Mantzarlis warned. “We are wrinkly arbiters of a take-no-prisoners war for the future of the Internet. And yet I think that in too many ways we still behave like in those early days when we were an experiment — when our good qualities were refreshing and our bad ones part of the learning curve.”

As Mantzarlis put it, “We can’t play the young and scrappy card anymore.”

One persistent compliant is that fact-checkers suffer from “selection bias,” in that they decide what to fact-check. The problem has become particularly acute for U.S. fact-checkers in the era of a president persistently tweeting and speaking falsehoods — and Republican domination of Congress. The result is that the percentage of fact checks of Democrats has fallen since the end of the Barack Obama presidency, creating an imbalance that some readers — and fact-checkers — find troubling.

In the past year, PolitiFact sent reporters and editors on the road to three Trump states — Alabama, West Virginia and Oklahoma — and fact-checked 54 local claims in an effort to reach skeptical audiences in those states.

The fact-checkers at Liberation newspaper in France — which is considered left-leaning —decided to confront charges of selection bias by overhauling their format and rebranding the product as The revamped organization now only answers queries posed by readers — and all of the questions, even the unanswered ones, will be made public in a searchable database.

Since September 2017, the CheckNews team of eight journalists has answered 1,600 questions from the 8,000 that have been submitted.

Pauline Moullot of CheckNews joked that it is “the slowest search engine in the world,” with one question taking six months to answer. But the format has proved such a hit that CheckNews is now working with the IFCN to take the concept worldwide and build a global database of fact checks. Already, it has been used during the recent elections in Tunisia.

Fact-checkers have also made significant advances in automating the search for statements to fact-check — with the ultimate (and perhaps unattainable) goal of providing instant fact checks as politicians make speeches.

In Argentina, the fact-checkers at Chequeado created Chequeabot, which automatically scans text from 30 media outlets around the country and identifies claims from politicians; it also matches possible claims from the group’s existing database of more than 1,000 fact checks to allow for quick reposting or tweeting of the previous fact check.

In the United Kingdom, Full Fact created a tool that monitors transcripts from the BBC and debates in Parliament and identifies checkable claims; it also matches claims from an existing database of fact checks. Brazil’s Aos Fatos, meanwhile, has developed a Facebook messenger bot — “Fatima,” shorthand for fact machine — which automatically answers reader questions about rumors and claims; it will be expanded to Twitter for the upcoming election.

Facebook's CEO testified for hours before Congress this week, but his statements weren't always accurate. (Video: Meg Kelly/The Washington Post)

In the United States, the Duke Reporters’ Lab, via its Tech & Check Cooperative, provides PolitiFact, and The Washington Post Fact Checker with a daily list of possible statements to fact- check, selected automatically through computer monitoring of CNN broadcasts, Twitter accounts and soon NBC’s “Meet the Press.” Duke has also developed FactStream, an app that automatically pulls fact checks from the three main U.S. fact-checkers.

Mevan Babakar, head of automated fact-checking at Full Fact, demonstrated to the conference an impressive tool that, via voice recognition, creates automated fact checks from data on the U.K.’s Office for National Statistics. Speaking into a microphone, Babakar said “unemployment was going down” and the tool instantly created a chart on the British unemployment rate. The hope is that the tool would allow reporters to quickly challenge politicians who may make a false claim during an interview or news conference.

The spread of misinformation across social networks has posed additional challenges. Fact-checkers have expanded the use of video to reach people not inclined to read fact checks. WDR German Public Television’s Wahlwatch feature, for instance, only produces videos designed to be shared on social media. One fact check led German Chancellor Angela Merkel to issue a correction after a fact check of one of her statements went viral.

Facebook product manager Tessa Lyons told attendees that the company has learned from its association with fact-checkers and is making improvements to reduce the amount of fake news spread through the social network. For instance, the company is using machine learning to detect duplicate fake news articles that have been debunked — and it will soon use machine learning to predict pages that are more likely to share misinformation. In four countries — France, Ireland, Mexico and India — Facebook also has expanded a test that allows fact-checkers to debunk manipulated and out-of-context photos and videos.

Facebook CEO Mark Zuckerberg faced a grilling by European lawmakers in Brussels May 22. He talked about efforts to erase fake news and to prevent data misuse. (Video: Reuters)

But in many countries, a bigger problem is the spread of hoaxes on WhatsApp, an encrypted instant messaging and phone service that has 1.5 billion users, mainly outside the United States. WhatsApp’s groups are restricted to 260 people — and the company never monitors the private-encrypted exchanges — so it’s difficult to crack down on hoaxes. People have been killed across India, including in vigilante mob lynchings, because of misinformation spread on WhatsApp.

Carl Woog, policy communications lead at WhatsApp, said the company has rolled out a way for businesses, such as news organizations, to develop a presence on the platform so that WhatsApp users can receive news and information from them. Silla Vacía in Colombia and Verificado in Mexico, which seek to debunk WhatsApp hoaxes, have already used the business app to respond efficiently to readers on the platform. During the Mexican earthquake, Verificado relied on 500 volunteers to help track down rumors and either confirm or debunk them.

(Note: Glenn Kessler is a member of the advisory board of the International Fact-Checking Network.)

(About our rating scale)

Send us facts to check by filling out this form

Keep tabs on Trump’s promises with our Trump Promise Tracker

Sign up for The Fact Checker weekly newsletter