This year, roughly half of the world’s population will be heading to the polls. And although the rise of AI brings many advantages, fake news is lurking more insidiously than ever around the corner of our democratic society. The World Economic Forum has even identified disinformation as one of the biggest risk factors that can destabilize a society.

Deepfake videos, for example, can be a playful gimmick, but in the context of election communication, they are primarily a worrying trend. Earlier this year, Immersive Lab researcher Jeroen Cluckers was invited by consultancy agency ICF gave a talk on AI and disinformation and provided the communications professionals in attendance with insight into the process of creating deepfakes.

Watch the full talk between Jeroen and his guests here.

Fake it until you make it

Jeroen Cluckers: “AI content is improving almost weekly. A while ago, you could still recognize an AI image by hands with a striking number of fingers, but that is already a thing of the past. We have definitely passed the uncanny valley point: a recent study showed that we are no longer able to distinguish real faces from AI faces. There are currently no suitable applications to support us in this, although George Bara—my co-speaker at ICF Next—is working on such a fake news detection tool. If you zoom in on an image with a trained graphic eye, you can recognize a generated image by the pattern it is constructed from, a kind of ‘noise’. Although that may not be the case for much longer.”

The good, the bad and the ugly

In the frenzy surrounding the US elections, the use of AI in political propaganda caused quite a stir on several occasions over the past year.

The Republicans pulled a stunt with a dystopian AI-generated video intended to illustrate what another four years under Biden would mean for the nation. Ron DeSantis, meanwhile, incurred Trump’s wrath with manipulated social media images of the presidential candidate and Anthony Fauci. And in Belgium, too, reactions to the AI resurrection of the late Jean-Luc Dehaene were divided, to say the least. All deepfakes and AI manipulations aside, can generative AI also play a constructive, less questionable role in political campaigns? We asked our AI guru Jeroen Cluckers:

“AI content could be very valuable in concretizing party programs, which often remain largely abstract to voters. Imagine if all political parties were to enter their agenda items and vision statements into ChatGPT, with the task of making that material digestible in text and images, and what that would mean for society… Then you could see, for example, what Antwerp would look like according to the plans of party X or Y. I think such an application would be very relevant in helping people make the most informed choice possible in the voting booth.”