Twitter needs to harness automation to combat falsehoods

May 28, 2020
Category: Fact-Checking,IFCN

Twitter got some great headlines this week when it started linking President Donald Trump’s tweets to fact-checks.

“Twitter has finally started fact-checking Trump,” said Recode. “Twitter’s first fact-check on President Trump calls out ‘false claims,’ ” said Ars Technica.

But Twitter’s effort was more symbolic than substantial. Although the company has taken some good initial steps, it needs to do a lot more – at a massive scale – to show it is serious about checking the avalanche of misinformation on its platform. That will also help the company use a wider array of fact-checking journalists to show its impartiality.

The misinformation problem is huge — and extends far beyond Trump, whose countless falsehoods mostly go unchallenged by Twitter’s moderators, even when he flings unsubstantiated murder allegations at his political enemies. Thousands of other politicians and political groups tweet false claims hourly and armies of followers spread them.

This is not an issue that any tech platform can handle case-by-case. It’s an internet-sized problem that demands an internet-scale solution, using technology that can be deployed as fast as falsehoods spread.

The world’s fact-checkers can help. Most now use a standard called ClaimReview that has created a database of about 54,000 fact-checks that can be accessed by Twitter (or anyone) to identify false and misleading claims. It grows by 89 new fact-checks every day.

ClaimReview isn’t sexy. It’s just a set of standards developed by our Duke Reporters’ Lab in a partnership with Google that independent fact-checkers use to identify the claim they have checked, who made it and whether it was accurate. But ClaimReview makes it easy for Google, Facebook (disclosure: both companies have provided grants for our work on automated fact-checking and ClaimReview in the Reporters’ Lab) and, someday soon, perhaps Twitter, to quickly match a claim with a fact-check. That’s the power of automation. And ClaimReview is available and free for anyone to use.

Automation is important because Twitter is mostly using an old-school approach – humans – to combat misinformation. That’s not a good long-term strategy because it won’t address the thousands of false tweets that go unanswered every day. The solution needs to be scalable.

The problem isn’t limited to claims that are spoken or written, which is why we’re developing a sibling of ClaimReview called MediaReview that fact-checkers will use when they debunk false or misleading videos or images. MediaReview, which like ClaimReview will be open to anyone, can help Twitter slow the spread of fake videos and bogus images.

Scaling up also will help Twitter rebut attacks that it’s being partisan. With automation, ClaimReview can be applied even-handedly to politicians from all parties.

Misinformation is big and worrisome. But Twitter has the opportunity to build on its encouraging first steps and make a substantial dent in the problem by enlisting the work of the world’s independent fact-checkers. The company should move beyond symbolic efforts and do it in a big way.

Bill Adair is the Knight Professor of Journalism and Public Policy at Duke University and the director of the Duke Reporters’ Lab.