Q&A: How CrossCheck used social media monitoring to uncover fake news

On the heels of the French election, we speak to First Draft News about the success of the CrossCheck initiative against fake news. 

Fake news has become a mainstream concern across countries. In February, we announced our initiative with First Draft News to support leading French publishers during the country’s presidential election season.

First Draft News’ CrossCheck initiative worked to gather expertise from the media and technology industries to debunk hoaxes, rumours, and other misinformation around the French election. Prominent newsrooms have taken part, including Le Monde, Agence France Presse, Liberation, BuzzFeed News and more.

NewsWhip made Spike available to journalists in these newsrooms to help them efficiently and easily spot the fake stories and sources of disinformation gaining traction on social media in France, monitor the coverage around the election for fake reports, and more.

Since we launched our partnership in February, we have seen impressive engagement with our Spike platform from these partner newsrooms. We spoke to Claire Wardle, PhD, First Draft News’ Research Director to get her thoughts on the initiative.

Can you explain how CrossCheck is helping French publishers to fight fake news and misinformation online? Who’s involved, and how does it work?

Claire: CrossCheck is a collaborative verification project which has bought together 37 newsrooms, and a number of technology companies to fight misinformation connected with the French election. We are using NewsWhip’s Spike platform and CrowdTangle to find content that is starting to circulate and has the potential to go viral. We are also asking audiences to submit questions, claims and hoaxes they would like more information about, and are running searches on Google Trends about the main candidates and issues.

Newsrooms decide to work on particular stories based on what is trending or emerging from the audience. They factcheck the claim or verify the video or image, and make that process visible to other newsrooms on Slack and via Check (a collaborative verification platform). Once the first newsroom is happy with the verification, they alert the other newsrooms who then take a look at the work that has gone into fact-checking the story. That is the ‘cross-check’ part! If they are also satisfied with the verification, they add their logo emoji in the Slack Channel to indicate they are happy. The story gets written up and AFP as the overall editorial lead for the project does a final check.

The story is published on the CrossCheck site in French and then translated into English, and the newsrooms partners then write up the story on their own sites. The stories are also pushed out on the social media channels of CrossCheck and those owned by the newsroom partners. CrossCheck has been designed to help support and power the work that newsrooms are already doing. These newsrooms already have their own audiences and it is about capitalizing on their audiences to reach as many people as possible. This is not about building a new brand.

Why are the particular challenges faced by French newsrooms around fake news during the election, and what are some of the steps they are taking to spot hoaxes and fake stories?

There aren’t as many ‘fake news’ stories in the way that we saw with the US election. It might be a language issue, or the fact that the French market is smaller so people saw less value in creating those types of sites. But there have definitely been manipulated photos, and even a very sophisticated hoax, where someone created a clone of the Belgium newspaper Le Soir to push out a false story that Saudi Arabia was financing Macron’s campaign.

First Draft News and many of its partners seem to stress the value of partnership among media companies as an effective way of debunking these stories. Why is that?

Previously newsrooms didn’t really have to worry about what was false. Photoshopped images or rumours ended up on the cutting room floor. They were’t mentioned in editorial meetings. Now I would argue there is a need for journalists to help audiences navigate the information ecosystem. People want to know what it true and what isn’t. But it makes no sense for 50 journalists in 50 different newsrooms to debunk the same rumour or to verify the same video that emerges on YouTube. There are better ways for journalists to be spending their time. So we believe the answer is collaborative verification.

If you could give one piece of advice to journalists and editors looking to improve their workflow to ensure that they are best equipped to deal with fake news spreading to their audiences, what would it be?

I would argue that it’s important to invest in tools that help you surface the rumours that are circulating online, and to ensure there are members of your team who are able to map how and when misinformation is spreading. We supported the creation of the ‘Field Guide to Fake News’ written by Jonathan Gray and Liliana Bounegru which provide ‘recipes’ for journalists to investigate the misinformation ecosystem themselves.

Once you know what is spreading it helps you choose which rumours to debunk. Focusing on stories that are not spreading is irresponsible as it can amplify rumours unnecessarily. But ignoring them means they can take hold and it’s much harder to debunk once they’ve become engrained.

What is next for First Draft News and CrossCheck?

The first step is to properly research CrossCheck to see what worked and what didn’t. We’ve always called CrossCheck a laboratory. We wanted to push something out into the field so we could test it. So once the election is over, we will start doing in-depth research. We will work with a social psychologist to test how we worded headlines, and used the visual icons. How did audiences respond to that? We need to talk to audiences, look at CrossCheck metrics to see who shared what and why. We need to know who we reached during the election campaign. We need to know whether collaborative model which looks great on a powerpoint slide worked in practice.

Once we know what worked and what didn’t, we can start looking at whether it’s possible to create a CrossCheck blueprint so that other countries can use the same model to cover their own elections, or even use it to cover misinformation that circulates at any given time.

Thanks Claire! If you’re encountering disreputable news in your reporting, let us know on Twitter. For a look into how you can spot fake news, take a demo of NewsWhip Spike.