From fact checking to live fact checking: what to learn from the US presidential election
The rise of Donald Trump in American politics and the Brexit campaign in the UK has put fact-checking in the spotlight. Politicians have distorted the truth on occasions for a long time but misinformation has been particularly ubiquitous in the last years with the emergence of the internet.
The launch of PolitiFact in 2003 heralded the creation of many fact-checking organisations. In June 2016, Duke Reporter’s Lab tallied 105 active sites around the world, a 60% increase from a year ago.
The American presidential elections in particular sparked an increase in fact-checking experiments, from live transcripts to browser extensions, with a focus on social media. International media have been watching closely as next year will be an important time for international politics. In Europe, two major countries — France and Germany—will elect new leaders.
Although trust in the media in the US as an institution has been at a record low, fact-checking especially done live during the debates has proved successful. For Alexios Mantzarlis, director and editor of the International Fact-Checking Network at the Poynter Institute, “it seemed to strike a chord with readers so the media adapted to this”. Live fact-checking initiatives drove record traffic to several organisations like PolitiFact, The Washington Post, and NPR. PolitiFact’s website registered more than 2 million page views on 27 September, the day after the first debate.
“NPR’s live transcript stood out as one of the best fact-checking experiments during the US campaign cycle”, according to Mantzarlis. It delivered the media company unprecedented traffic both during and after the debate. Its transcript of the first debate drew more than 6 million users during the debate and the day after. 22 percent of visitors to the page during the debate stayed all the way to the end. “Annotation had been a promising tool for fact-checkers but had not really been used to its full potential until then”, says Mantzarlis. PolitiFact had tried to use it during the State of the Union in 2015 but the fact-checkers’ own annotations seemed to be lost in the sea of annotations from the general public.
NPR’s visual team started developing its special live-transcript only two weeks before the first debate. The visual team made an efficient webpage that was mobile and user-friendly. A transcription service fed their transcript into a Google doc, where a team of journalists added their annotations in suggestion mode. About 20 journalist collaborated on making the transcript, some focusing on the accuracy of the all text and others focused on fact-checking their specific beat. Amita Kelly, the digital politics editor and producer who was leading the team, edited and published the fact-checks. The team was well prepared. They had been working on their live fact-checking workflow since the Republican and Democratic conventions in July. The big difference though was that during the conventions, they knew what Clinton and Trump would say in their speeches, which is not the case during presidential debates, Kelly explains.
As moderators were not doing much fact-checking during the debates, fact-checkers could only hope that the audience would have a second screen to be informed of corrections. While many organisations used a liveblog or a single webpage like NPR for their live fact-checks, they also heavily relied on social media to distribute their corrected information.
The main issue for fact-checkers on social media is to reach members of the public. “FactCheck.org or PolitiFact can fact-check all they want during a live debate but if the people who need to hear those corrections aren’t following them on Twitter, then it’s a problem,” says Jane Elizabeth, senior manager of the accountability journalism program at the American Press Institute. The organisation did a study in 2015 on the spread of political misinformation on Twitter and found that nearly three in every falsehood on Twitter is uncorrected. Twitter has also proved as complicated for fact-checkers with only 140 characters to use. Some fact-checking tweets can be misleading. Mantzarlis says:
“If your tweet isn’t clear about a fact, you run the risk of spreading this information further”
Many organisations have tried to share their fact-checks on platforms other than Twitter. The Washington Post’s Fact Checker journalist Michelle Ye Hee Lee uses Snapchat for what she calls “snap-checks”. She shares fact-checks, each structured around a quote, analysis and rating, in a chronological story on the Washington Post’s Snapchat account. Her story is a mix of photos and videos with captions, in case the viewer doesn’t have sound on.
Every Friday for her #factcheckfriday initiative or the morning after a debate, she takes about 20 minutes to create her story, which features illustration of the candidate portraits as well as the Washington Post’s fact-checking visual to indicate the level of truth of a claim: Pinocchios. Her “snap-checks” receive the most engagement by far among the Washington Post’s Snapchat stories, she says. But, one of the limitations of Snapchat as a platform for fact-checking is this inability to add all the required context and nuance so Lee says she chooses the claims that are straightforward.
Univision’s data unit also tried to engage with its audience by creating bots to fact-check claims. They first tried a pilot project with an SMS bot in partnership with Purple, the company that sends elections news texts. But, when Purple decided to shut down its service in July, the SMS bot came to end. So Univision thought of making a Facebook messenger bot.
About 200 users have subscribed to it. It sends them a politician’s statement, which they can fact-check themselves, reply with their conclusions, and the bot tells them if they are right or wrong. They expected users to interact with at least 4 fact-checks but users go beyond that, and answer an average of 6, writes Ronny Rojas, an investigative journalist who leads the data unit.
Univision invested a lot into fact-checking for this presidential campaign. For the last debate, Univision also invited 11 journalists from Latin America and Spain to collaborate on the challenges of fact-checking. Among their projects was an interactive visualisation to show readers the proportion of lies told during the 90 minutes debate.
Other fact-checking initiatives included PolitiFact’s beta Chrome extension, the FactPopUp, built with the Duke Reporters’ Lab, which would let fact-checks pop up on your screen over a livestream of the debate. About 380 beta testers saw 29 pop-ups based on PolitiFact tweets with links to claims that had already been checked, hence the speed of the fact-check. Mantzarlis praised the initiative saying it was a good way of ending the spread of misinformation because the closer you get [the claim and the fact-check] together, the more likely you are to debunk something.
The Duke Reporters’ Lab also developed with Jigsaw of Alphabet (Google’s parent company), a visually attractive and shareable widget, Share the Facts, embedded on news websites such as The Washington Post, PolitiFact, and FactCheck.org.
Attempts at creating fully functioning automated fact-checking have created wide interest among fact-checkers. For Elizabeth:
If this was the year of social media as a platform for distributing the information, in 4 years it will be the elections of the automated fact-checking technology.
Some American researchers are trying to analyze some text and identify factual claims with their project, Claim Buster. The creators called the project the “Holy Grail” of fact-checking, even though it’s not a fully comprehensive fact-checking system.
However, Full Fact, a British fact-checking organisation, has said they are months away from finishing practical automated tools for journalists and fact-checkers. These sorts of tools would allow human fact-checkers to spend time on the more nuanced and complicated claims. More improvement in automated fact-checking is expected in the coming years if not months.
But until now, most organisations are still favoring their staff with their special expertise and years of coverage of specific beats over automation. The Washington Post’s attempt with its automated pilot project, Truth Teller, “was put to an end to focus, and invest in their fact-checking team,” according to Molly Gannon, The Post’s spokesperson.
Fact-checking has been an important conversation during the American presidential campaign. Audiences turned to fact-checkers for accurate information, following and sharing their content on social media and even going on specific webpages. Media organisations devoted important staff resource to make their live fact-checking successful and automation was left aside. They are not the only ones to have realized the need for more accurate information though. Amazon’s Alexa’s new feature was launched. It can fact-check basic claim from a database of 2,000 professionally curated fact-checks. Google News also launched special fact-check labels and it seems to be working. After Hillary Clinton told viewers during the debate to check if Trump had supported the Iraq war, which he claims he didn’t, FactCheck.org received 1.1 million page views on their fact-check of the issue, the organisation reported. It was the first story on a Google search.
Five pieces of advice on how to fact-check an electoral debate or an electoral campaign:
- Preparing key facts and databases are essential to make live fact-checking a smooth operation. Doing groundwork research before elections are in full swing can be smart especially as election cycles in many countries are shorter than in the US. “The most important thing is to be prepared by knowing what the candidates have said in the past and how they frame the issues,” says the Washington Post’s Fact Checker columnist Glenn Kessler, who’s fact-checked five rounds of presidential debates.
- Train journalists with the best fact-checking practices. “It’s not like somebody can sit down and be a political fact-checker overnight,” says Elizabeth. Her organisation, the American Press Institute, helped Univision to develop its fact-checking platform at the beginning of the year. Define your journalists’ roles. At NPR, journalists were only listening for the parts of the debate that were related to their area of expertise such as immigration or education for instance. There were also 3 journalists reading and fixing mistakes on the transcript that was being fed into the Google doc, and one who was listening to the audio of the debate specifically.
- Many journalists put the emphasis on the need for transparency especially as some audiences have shown distrust toward mainstream media. Let the readers know your primary sources and put hyperlinks. Elizabeth recommends uploading transcripts of speeches, interviews to show all the research.
- Use social media to their full potential but be careful not to mislead your audience to create more engagement. Make sure the nuance and context of a fact-check is clear when you create your post or tweet. Visuals are important on social media. Univision’s fact-checking team has a designer in charge of making some of the fact-checks. Hashtags are good for audiences to easily find your fact-checks. CNN used #CNNRealityCheck, which was clear, specific and could not be confused with unrelated hashtags. Some newsrooms are particularly creative; Chequeado, a fact-checking organisation in Argentina, is even trying fact-checking GIFs.
- If you try a new project, familiarise your journalists with the tool: Kelly’s biggest challenge leading the NPR live transcript was “to make sure people knew what we were asking them and how they should go about doing that”. Journalists were trained to work with the live transcript in a Google doc.. At Univision, the fact-checking team and journalists from Latin America and Spain spent two days before the last presidential debate to brainstorm ideas for their fact-checking (including visualisation tools).
Josh Stearns—Democracy Fund
Fact checking has taken center stage in this election, but newsrooms need to go beyond fact checking politicians statements and help debunk viral misinformation too. (CNN, 1 November 2016)
Lynn Walsh—Society of Professional Journalists
“For journalists, it’s just one more validation that their work is trustworthy. For the public, it provides a way to cut through the clutter that sometimes exists — like articles that don’t clearly distinguish between fact and opinion.” (TechNewsWorld, 19 October 2016)
Brian Stelter — CNN
“Journalists cannot just play these soundbites, quote these claims, and then move on to the next subject. We can’t just let it seep into the discourse like it’s normal. We have to stop and fact check and contextualize.” (The Huffington Post, 8 August 2016)