Reuters built an algorithm to battle fake news on Twitter
The pernicious effect of fake news has become a major talking point in the wake of the US election. Do social networks such as Facebook and Twitter have a responsibility to curb misinformation? How should they go about doing this in the first place?
Reuters has one idea on how to tackle this. The news organisation has developed a tool to help it spot and verify breaking news on Twitter, using algorithms to filter tweets in real-time – sorting credible reports from dubious stories.
Reuters News Tracer has been in the works for two years, but has only now been spoken about publicly, in interviews with Nieman Lab and the Columbia Journalism Review. Reg Chua, Reuters’ executive editor of data and innovation, told Nieman Lab that the tool was born out of “an existential question for the news agency”.
“A large part of our DNA is built on the notion of being first, so we wanted to figure out how to build systems that would give us an edge on tracking this stuff at speed and at scale,” Chua said. “You can throw a million humans at this stuff, but it wouldn’t solve the problem.”
The News Tracer organises tweets that it deems relevant to breaking events. On a basic level, this means tweets that feature the word “bomb”, for example, are clustered together. The more complicated task, however, is verifying the newsworthy content against a mass of opinions and potential misinformation.
To do this, Reuters’ algorithm assigns scores to tweets based on 40 factors – ranging from whether the tweet is coming from a verified account, whether the tweet includes images, and the syntax of the tweets themselves. Tweets written all in capital letters, for example, are less likely to be true, according to Chua.
If a story cluster has a certain verification level, Reuters will tweet out its own breaking news report about the event. Reporters will then follow up on the story themselves. The idea is that this system brings an element of automation into the reporting process, but is nuanced enough not to spread misleading information.
Given the speed of information enabled by smartphones and social media, tools such as this are arguably necessary to keep pace with breaking events. Perhaps more important is the ability for the system to verify reliable reports at a speed that outflanks the spread of misinformation.
Other news outlets, such as Le Monde, are experimenting with methods to squash fake news. Moves by news organisations are only one side of the equation, however. Social media platforms such as Facebook will also need to contend with how they prioritise verified stories over dubious reports.