The spread of fake news online has come to light as a serious problem in the aftermath of the U.S. election. And while fact checking has been integral to quality newsrooms since, well, forever, automated fact checking hasn’t. Today, the sheer volume of sources from which information is spread makes verification a strain on newsroom resources. That’s why there are a myriad of projects underway aimed at automating the process.
Time is of the essence, with a raft of national elections set across Europe in 2017. Here’s a look at five automated fact-checking projects:
News feed demotions
Naturally, Facebook is cooking up new ways to automate fact checking, having drawn a lot of heat for its role in the spread of political misinformation around the American election. The social network is releasing tools aimed at snuffing out false posts and has brought in third parties to investigate and flag reported stories. The social platform has enlisted the help of fact checkers at media outlets ABC News and the Associated Press, along with FactCheck.org, Snopes and Politifact, all of which will use a tool created by Facebook to help evaluate the truthfulness of stories that have been flagged as fake. Once a story has been verified to be false, it will be demoted in the news feed.
New tools for roaming journalists
Full Fact, a British charity, has raised €50,000 ($52,000) from Google’s Digital News Initiative to invest in new methods of automating fact checking, whether it be tools for newsrooms or mobile apps journalists can use when note-taking at events. The charity has provided fact-checking resources to newsrooms for several years, but its mission is now to scale what it can do using statistical analysis and natural language processing tools already available.
“There’s still a lot of misinformation floating around parliament,” said Full Fact digital products manager Mevan Babakar. A lot of unverified statistics are often touted, and then parroted everywhere. “A few years ago, a headline in The Sun said 40 percent of knife crime was committed by 15-17-year-olds. That story went on to be debated in Parliament, and later a law was changed. When we looked into it, we found it was one police officer perpetrating it. There are countless more examples,” said Babakar. To stop misinformation like this escalating, Full Fact is using its Google funding to develop automated checking tools, which it hopes to roll out by December 2017.
One of the products being looked at by Full Fact is a mobile app, which journalists can use when covering live press events. The concept is that a journalist can record a speaker’s session, and while listening can scroll down and watch the words appear on screen. As soon as someone makes a claim or waves a statistic around, the journalist can highlight it in red, and it can be verified. “That could also give you a great follow-up question if it showed it not to be true,” added Babakar. This particular concept is just at the idea phase for now. “We’ll see how far the money stretches,” she added.
Washington Post fact checks Trump
Stateside, the Washington Post has released a Chrome plug-in aimed at fact checking posts from one Twitter account in particular: Donald Trump’s. The “RealDonaldContext” plug-in simply adds a grey box to each of Trump’s tweets, adding relevant and important context or flagging misinformation.
Starting in November, Washington Post editors began appending text like “There’s Important Context Missing” or a blunt “This is incorrect or false” after fact checking tweets. The tool was devised by Philip Bump, a reporter at the publisher. Granted, this requires more manual work on behalf of the journalists checking the facts, and yet most media experts believe the human element will be critical to most automated fact-checking tools in future.
Full Fact is also looking into the idea of adding plug-ins to Twitter to verify a politician’s tweets. This is just at the concept phase, but the idea is that whenever a false statistic or comment is tweeted, a person about to retweet it would see a pop-up saying that the information has been fact checked and found to be wrong — and ask if they still want to proceed.
This research report is available exclusively to Digiday+ members. Join now for access.
Le Monde’s hoax-busting database
French newspaper Le Monde has a big project underway to crack down on fake news. Les Décodeurs, a 13-person fact-checking division within the newspaper, is working on a “hoax busting” database, designed to root out false information, at scale. The publisher already has a search engine through which readers can fact check politicians’ statements.
But it’s now working up an automated grading system, by way of a Google Chrome extension. The project has also received funding from Google’s Digital News Initiative. People who download the extension will be able to see which stories have been verified, which are dubious and those that are outright false via green, yellow and red flags that will appear against articles.
Standardizing automated debunking
One of the trickiest aspects of fake news is that the debunking never travels as fast or as far as the original hoax. FirstDraftNews, a London-based fact-checking coalition, is actively organizing meetups with publishers and major tech platforms to see if this is some kind of automated fact check can be scaled on social media. So far it has coordinated several closed-door meetings with a range of publishers and social platforms in the U.K. “We hold off-the-record meetings so partners can talk freely. All the social networks have been open to participating in collaboration.” said co-founder Jenni Sergeant.
“We want to establish what’s possible and how we can then standardize it. It’s complex because there are nuances like satire to consider.”
Image: courtesy of Geekandsundry.com.