NORDIS and its Swedish partners, Källkritikbyrån and the Fojo Media Institute, look back at an election marked by misinformation, from claims of election interference and voter fraud to AI-generated content.
The Swedish elections are over, but the aftermath looks set to be dramatic as complex negotiations loom over the formation of a new Swedish government, following allegations of electoral fraud against several politicians.
Rumours of electoral fraud also played a role during the campaign. But looking back, can the election be described as one marked by misinformation?
NORDIS spoke to its Swedish partners at the fact-checking organisation Källkritikbyrån and the Fojo Media Institute at Linnaeus University about the role of misinformation in Sweden’s elections.
Q: What have been your main focuses during the election campaign?
Åsa Larsson, editor-in-chief Källkritikbyrån: As factcheckers we constantly check claims in social media, but it’s always good to have organisational muscles in intense news situations. For this reason, we were really glad when the Fojo Media Institute invited us to collaborate on the Valkollen project. It gave us strength in numbers and the resources we needed for this election. Our goal during this election period was to stay on top of viral claims made in connection to the politics, the parties and the voting procedures. I’m content with the result, we addressed a lot of fake and misleading claims.
Andreas Önnerfors, researcher at Fojo Media Institute: A key component of fact-checking during elections was that Fojo could enable powerful monitoring across platforms. We also spent a lot of thoughts into foresight, I e developing potential scenarios before, during and after elections and likely areas of manipulation.
Q: The elections were dubbed Sweden’s first AI election. How did this manifest itself?
Åsa Larsson: AI is now in everyone’s hand. That means that it’s used by the regular perpetrators, such as clickbait manufacturers, foreign powers trying to influence the Swedish voters and people who spread twisted claims for ideological reasons. But it’s also used by anyone who wants to comment politics or make a funny video.
For example, someone made a fake news story about the post trucks of the rather disliked post company in Sweden. They were said to have gotten into an accident and dropped a whole lot of ballots in a river. “I saw that somewhere on Facebook”, people said in comment sections. The image was made with AI, but the claim could have just as easily been spread before AI entered the stage, with regular image manipulation.
“Is it April 1st?? What the hell is going on?” Photo: Screendump from Facebook
Q: What type of misinformation characterised the election campaign?
Åsa Larsson: Unfounded rumours of election fraud characterised the election campaign. These are now a fertile ground for conspiracy theories and fueled by the situation in Dalarna in the middle of Sweden, where there is a scandal regarding the Left Party and possible voter fraud.
A lot of minor errors during the voting was spread as full proof of intentional voter fraud to change the outcome of the election. A new system the voter booths and ballots apparently sometimes resulted in missing ballots in every booth, which was vehemently spread as proof of wrongdoing, often by the smaller parties who need to distribute their ballots themselves to all of the voter sites.
In the last election around 700 million ballots were printed and only 20 million were used, which means that there will be a lot of litter left behind. Sometimes people found ballots in the dumpsters and took it as a sign of election fraud.
There were also accusations of electoral interference, several of those stories being spread as truths without evidence. For example the Sweden Democrats launched a campaign site close to the election day and it got taken down, flagged by the web host for suspected inauthentic behavior. The host company said that it was reviewed and reinstated, but during the down period prominent Sweden Democrats claimed that the site had been taken down “due to a massive influence campaign from leftist activists”.
Andreas Önnerfors: One particularly chilling aspect we detected in the run-up to the elections were two conspiracy theories, one about ‘globalist elites’ manipulating the elections and democracy and the other about ‘imported voting cattle’ voting for the political left. The first of these two tried to discourage people from voting, demobilizing the electorate with the idea that democracy is a zero-sum game, rigged by a malign elite.
The other pointed out toxic enemy images, the immigrant as the threatening other in the body politics in a secret alliance with the internal enemy – left parties. It stirred up quite substantial racism and after the election results started to live its own life, both amplified by domestic politicians and alternative media and furthermore massively boosted by foreign amplification.
Q: To your estimate, how much damage was caused by foreign interference?
Åsa Larsson: Foreign actors were active and it’s always hard to tell to which extent they affect the general population, but as far as we can estimate, not a lot. Operation Overload is meant to inundate journalists and factcheckers but the only time I see reporting on examples from them is when a report surfaces, like it did during the election.
They obviously fan flames with fake accounts and bot activity, but I can’t think of an example where they have been the source, like a post with some claim that then became viral and influenced voters.
Andreas Önnerfors: Obviously, with the amount of data we gathered for the project we weren’t able to deeply investigate typical patterns of coordinated inauthentic behavior yet. We have seen strange movements of online content across continents, but need to dig deeper. But all in all, there are no immediate signs of coordinated campaigns and their expected content according to our forecast scenarios.
The positive interpretation is that foreign actors know that Swedish information resilience is high and thus don’t even try, a negative interpretation could be that domestic assets independently amplifying talking posts of geopolitical adversaries already are well established in the information landscape and are not in need of further support or amplification.