What did First Draft learn from fact-checking the US 2020 election?

The US election last week was, predictably, fraught with misleading claims and images surfacing on social media. The main topics were dubious ballot stuffing and voter fraud, which kept fact-checkers pretty busy.

Fact-checking organisation First Draft (FD) decided to take a 'newswire' approach, leveraging its global team across the US, Europe and Asia-Pacific. It created a private Twitter account in September which posted all the important updates from its Slack community.

Journalism.co.uk spoke to Jasper Jackson, special projects editor of FD, who led the London-based operations, about fact-checking arguably the biggest election in the world. Jackson explains some of the main experiences and takeaways in the run-up to, and over the course of, the day, as well as what work still needs to be done for future elections. The interview has been lightly edited for brevity and clarity.

FD launched a private Twitter account with updates from its Slack community. Why was this your approach for monitoring the US election?

We have had a very effective Slack community which was built predominantly around the coronavirus pandemic. It was successful as a way to share our insights into topics and get our members to collaborate around topics. We will use that model going forward as it is great for running projects over time.

Having said that, we knew the day before the US election and on the day itself, it was going to be absolutely insane for journalists. As media organisations often think about being where their audiences are, we realised that for us, journalists who are working on the information disorder space are on Twitter. So having a direct channel was the right place for us. It was viewed as a newswire for misinformation updates.

What was the logic in keeping it private rather than public?

People who work in this space have had to get to grips with issues around amplification of rumours or information which still has significant questions around it.

Simply posting something which says 'maybe this is true, maybe it is not' is actually quite irresponsible if you have not had time to fully verify it. But we still wanted to flag up potential cases of misinformation to journalists quickly.

There are still issues around how to do that in a private channel, but the point is that everyone we accepted were people we knew understood these issues around amplification. We wanted to say 'We are seeing these emerging reports of ballot stuffing, we do not know if it is true but it looks questionable', and that is not something you can do publicly and responsibly.

Because someone might misinterpret the point and amplify it?

Absolutely, and because what we are trying to do is to help journalists do the best journalism, not serving the broad public.

The Twitter account was launched in September giving two months of runway to the event. How did that help?

It did not help us identify what was bubbling up, but it did give us a dry run to know what works in that format because it is different using Tweets as new notifications.

Prior to the election itself, it was a mix of stories that we can help people with, but also sharing our resources around what was going on. As we headed into the election, it became the newswire and the constant flow of information.

What was election day, 3 November, like for you?

We benefited from having a global team in the US, Europe and Asia-Pacific (APAC) making the whole day a 24/7 news operation. I was overseeing the London-based team.

On the day, our APAC team started up by looking at any areas which were likely to have emerging disinformation. The obvious ones were issues around polling stations not being open and attempts to dissuade from voting.

In this election, the misinformation has been around mail-in ballots. APAC was looking for those leads and did a document at 7am for us in the European team. We followed up the APAC leads, plus doing another sweep of all the social media channels we were monitoring.

We added notes and details about our own findings and then we updated the FD Twitter account until our US team came online. At which point we did another hand-over document and call and they went into the day with the threads we identified. We were lucky to have that global operation to transfer through the day. We then kept going on a rolling basis for the week.

You predicted the event would be insane. Did it live up to that description?

It is very difficult to do comparisons because we are living in unprecedented times in the political and information space. But it was a hugely charged election and there is a parallel with the coronavirus pandemic.

In my opinion, one of the reasons that the pandemic attracted so much misinformation is simply that everyone was talking about it. The same is to some extent true with the election. Add to that the same contentious activity going around the country in thousands of locations with huge numbers of eyewitness media and images that can be interpreted in many ways. It was carnage.

There was a deluge of leads from different spaces, and no-one can verify it all, but there were key themes. What we expected did come up: observers, mail-in ballots and the boring stuff like what you can display outside polling stations. But also, because it was so partisan, there were people who wanted to make a point by amplifying this misinformation regularly.

The theme of this election has been around the legitimacy of the votes, which as you say, you anticipated beforehand. How did knowing that help?

With voter fraud and mail-in ballots, the first person who posts an image is not likely to be famous with a huge following. But, it will very quickly be amplified by a network of people who will be on your radar who regularly spread misinformation around these topics. You are prepared because you have seen the people who amplify messages around voting, for example, so they are likely to amplify again messages which fit their own narratives - even if they are of dubious provenance, which was often the case.

Second, this election highlighted the need to go and talk to people who are running the elections and understand the electoral laws.

The classic example was footage of people who were copying ballots and filling them in themselves. There is a legitimate reason for that, as some of the ballots do not go into machines if they are not read properly so they need to be re-filled and that is done under close observation.

You can only be 100 per cent sure if you know that. Very few journalists are clued up on every state and jurisdiction, so you must go and talk to the people who do know. Local media were really invaluable in this, (FD is a core partner of POLITICO's Electionland initiative), but as a journalist, you need to go and talk to officials and understand the law and the process in detail.

What we saw is that a lot of the examples of misinformation were not deepfakes or photoshopped images, it was people showing images or video and misrepresenting what they showed.

Did you have that information at the ready, or did you need to react in the moment?

Sometimes we did have it but there are also many permutations of what can crop up. We are journalists, not electoral law experts. What we had to do was be quick and know the layout of the land and which officials to contact, identify the issues quickly and get in touch to verify what was going on.

You mentioned the usual suspect sources that spread misinformation. What did you look for here?

The way social media monitoring works is that you have huge lists of people you have identified based on their activity in spreading misinformation.

You then have Boolean search terms which have traditionally picked up misinformation, and you build new ones throughout the election. Our US team built state-specific ones and spent time on those beats themselves. They had specific searches set up which pulled in queries all the time. That is coupled with lists of pages which have traditionally shared misinformation.

It is not the case of identifying individuals per se, but it is having the right net and know which shallows to go fishing in.

There were also cases of TikTok being a source of misinformation. Much is still unknown about this platform. What did you learn?

We have one person who is really good at it: Laura Garcia. You can search for specific terms and hashtags there but we were not focused on TikTok. You do tend to pick up on it on other social media and that gives you a profile of misinformation likely to be spread. But it is still new and difficult because we do not have the tools for searching TikTok.

Because there is not a Tweetdeck for TikTok, yet?

And there is no Crowdtangle for TikTok, either. We do not have those tools yet and there was a lot of misinformation on TikTok. It is an emerging problem because it has a unique search activity and algorithm.

There is potential for exposure to misinformation in ways we are not used to. We are now used to interactions on Facebook and Twitter and how that feeds into algorithmic sharing, so we know some of the basic principles. TikTok is much more of a closed box and we need to put more effort into understanding it. We will have to, it will not solve the tools problem, but it is an area where we all must pay more attention to it.

What is your focus now that the results are in and there is the potential of a disputed result?

During the election, it is easier to focus on what you could predict. While there will be familiar talking points, we will have to go back to what we do generally. That is keeping an eye out for the themes and narratives that emerge. The networks and mechanisms that encourage misinformation, the natural human tendency to spread misinformation because have an agenda, has not gone away. The infrastructure has not gone away. I do not expect it to disappear as an issue any time soon.

What have you learned about the way misinformation spreads in this election?

It is clearly possible for social media networks to take action against people spreading misinformation once they decide what that is, how to target it and how to direct resources towards it.

The social media platforms did come up with criteria that satisfy them. We need more research to know whether that is effective, and we do not know if it is, but after years of saying this was an insurmountable problem, quite rapidly the social media companies did come up with techniques and tactics to tackle it. We also do not know if it will just drive people to other platforms and that is something to monitor for the future.

Join us at our next digital journalism conference Newsrewired from 1 December 2020 for four days of industry expert panel discussions and workshops. Visit newsrewired.com for event agenda and tickets.

Free daily newsletter

If you like our news and feature articles, you can sign up to receive our free daily (Mon-Fri) email newsletter (mobile friendly).