Can you use AI, or artificial intelligence, to predict who is messing up in the TV program Wie is de Mol? This is what six ICT students from Fontys in Eindhoven are trying to discover. Every Saturday they are glued to the TV and afterwards they start analyzing. The super-fast thinking of the computer should ultimately lead them to the Mole. “We think we are in the right direction,” says former student Niek van Dam.
For those who are not familiar with the program: In Wie is de Mol, a group of famous Dutch people try to earn as much money as possible for the pot by carrying out assignments. One participant tries to thwart this. It is up to the other candidates and the viewer to unmask that saboteur.
“With AI you can unmask who is underneath that makeup.”
The ICT students all watch the program and after the first episode, teacher Frans Mouws came up with the idea of doing more with it. “Frans noticed that the Mole was in the picture with make-up during one of the assignments,” Niek explains. “With AI you can unmask who is underneath that makeup and that is how the project started.”
The students first developed their own software and all the information they need is now retrieved by AI. “That means collecting and processing a lot of data and hoping that a good outcome will be achieved,” says student Jeroen Eillebrecht.
“We can see which candidate is being talked about a lot and is seen as suspicious.”
They soon discovered that there are more ways to unmask the Mole. His group is now busy analyzing all social media traffic surrounding the program. “For example, we can see which candidate is talked about a lot and is seen as suspicious,” says Jeroen.
They also look at the social media use of the candidates and their families. “The participants are not allowed to post anything on social media during the recordings. But we can, for example, see in messages from friends who came home last. They could be in the final and could be De Mol.”
Meanwhile, student Yap Yung Hui compared the photos of the face-painted people from the first episode with those of the candidates. Based on facial features, she initially identified Justin as the suspect Mole, but he was dismissed last week. So AI is not a magic ball. “It remains speculation,” says Niek. “The program makers could also make the faces thicker with make-up, which makes it more difficult.”
“Thanks to AI, everything happens ten times faster than we could.”
So it is still human work, but that also makes it fun for the students. “We have to do the hard work, but thanks to AI everything happens ten times faster than we could.”
The students have now made further progress with their research and now think they are on the right track. “We now suspect Kees van der Spek, Anna Gimbrère and Jeroen Spitzenberger,” says Niek. The first two emerged from the photo analysis and Jeroen from the social media analysis.