“Fake news!” How AI is countering a major threat to democracy

Viktoras Daukšas of Debunk.eu explains how AI is being used to spot and counter Russian disinformation campaigns

In August 2019, Russian media reports began to emerge that orphans were being brought to summer camps in Lithuania and taught to kill. The catch? The story wasn’t true – it was ‘fake news’. Although summer camps do exist in Lithuania (as they do the world over), the young attendees weren’t being taught warfare techniques.

Misinformation campaigns have long been used by hostile governments against enemies and competitors, yet in the past few years the problem of ‘fake news’ has grown dramatically in scale. Facilitated by social media, it’s easier than ever to spread lies and confusion online. But concerned citizens are starting to take a stand – and they’re using AI to help.

Viktoras Daukšas, head of Lithuanian organisation Debunk.eu, likens the problem to a contagion: “Disinformation, as with any contagious disease, can touch everyone personally – countermeasures are key to prevent the spread of the virus”. Daukšas, who will be speaking at Cloud Expo Europe London this March, believes AI is an important corrective in the treatment of this ‘illness’.

Debunk the bunkum

The notion of online ‘fake news’ has only really come to prominence in the West in the past 4-5 years, especially with allegations that Russia spread countless false stories during the US 2016 presidential campaign, the Brexit referendum and other democratic exercises around the world. But the phenomenon has a longer history. In Baltic States (Lithuania, Latvia, Estonia), former members of the Soviet Union, citizens have long been exposed to falsehoods principally emanating from their large neighbour.

“Lithuanian academics stated to analyse disinformation from the year 2000,” reports Daukšas, and a handful of organisations have started to take the mantle on since then. Debunk.eu was founded in 2017 with the central belief that AI could make it a lot quicker to assess the validity of a story published online.

Prior to Debunk.eu, much of the fact checking that went into working out if a story was false required journalists, academics or concerned citizens to laboriously follow up suspicious sounding stories themselves. This is very time consuming – reporters would lose hours every day calling up other news outlets or trying to locate sources and verify quotes.

And this is where Debunk.eu comes in. Daukšas describes the initiative as “a team of highly qualified disinformation experts and IT development professionals”. Their stated aim was to “start collaboratively countering the ever-increasing issue of disinformation in Lithuania”.

Tackling the problem

The team behind Debunk.eu began by building a platform which uses AI to “tackle the disinformation problem head-on”. Daukšas explains that the platform can:

• Monitor 1,500 domains and 30,000 articles every day
• Use AI to spot disinformation within two minutes of publication
• Provide journalists, think thanks, fact-checkers and academia insights about potentially harmful content online and automate at least 50 percent of manually performed tasks
• Unite major media outlets in Lithuania on one platform

So how does it work then? Debunk.eu uses AI to weigh up several factors to decide if a story could be fake news – from the content itself to the time it was published or the outlet that posted it.

For example, one factor the platform is trained to spot is articles containing words that are associated with fake news stories. Such content often contains highly emotive language and themes – topics such as poverty, health scares, rape and murder are common. This then alerts human editors to the most likely ‘fake news’ stories coming out, so they can debunk them first and halt the spread.

“The Debunk.eu platform automates a lot of features,” says Daukšas. But crucially humans are never out of the picture. The “main decisions are left for experts. The system has two human verification levels – that helps to train system better and avoid mistakes.” Once Debunk.eu has detected a potentially fake story, it can then alert government agencies and media outlets warning them not to republish falsehoods.

Stop the press!

Given the incredible speed at which disinformation can spread, time is of the essence when it comes to spotting and countering it. “In the case that a large, potentially harmful campaign is detected, the system automatically notifies journalists to debunk it before the story goes viral and the consequences become irreversible,” says Daukšas. This is vital because even if a story is later shown to be false, many people who viewed the original content may not see it was later discredited and continue believing the untruth.

And the platform is effective. Daukšas proudly points out that “since the inception of Debunk.eu, journalists have debunked over 200 disinformation campaigns detected by the platform which were read more than 13 million times.”

The initiative has also received plenty of media and political attention for its work and was a finalist in a NATO competition for tech that detects malicious content online. This success has also set the platform to grow internationally. With backing from Google and the EU, Daukšas says the organisation will be “expanding its operations to Latvia and Estonia” and this will potentially “serve as a sandbox for drafting successful strategy and worldwide expansion in the future”.

Reinforcing the Fourth Estate

Daukšas does not see AI platforms like Debunk.eu as a ‘silver bullet’ to disinformation campaigns. Ultimately, citizens need to be “taught to distinguish between reliable and untrustworthy sources of information” themselves. But, AI enhanced platforms like Debunk.eu can play a valuable tool in supporting the media to fight false information and lies.

“Debunkers play the role of online police – they identify and expose the bad actors, thus informing the audience which ‘internet neighbourhoods’ to avoid” argues Daukšas. “Debunking allows readers to see the scale, instances and sources of intentionally harmful techniques targeted at them.”

When AI and education are combined, society may just have a robust counterweight to one of the biggest challenges to modern democracies.