Laptop met handen
Laptop met handen

Why you should (not) let AI decide who you vote for

Are you unsure about how to vote? Simple: just ask an AI chatbot and within seconds you'll have ready-made voting advice. Easy? Yes. Desirable? Not really. And that's not just because of the misinformation with which AI tools can influence your vote, says communication scientist Gabi Schaap. “Aren't some choices too important to outsource?”

It's very relatable. You want to vote wisely, but you don't feel like reading through every party manifesto, watching election debates and figuring out exactly what hypotheekrenteaftrek (mortgage interest relief) is and what it means when it goes up or down.

“Asking AI for voting advice saves you a lot of time and energy,” Schaap admits. “This tendency is psychologically explainable: people like to choose the option that will give them the best results with the least effort.” Moreover, AI tools are very accessible sources of information. “Someone who wants to know more about democracy or about different parties can learn a lot from them. In theory, at least," adds Schaap.

Gabi Schaap

Enhanced spell checker

Because in practice, as research shows, AI tools such as ChatGPT spew out a lot of misinformation, for example about how to register to vote in the American elections. In the run-up to the Dutch elections, the Autoriteit Persoonsgegevens (Dutch Data Protection Authority) also warned against relying on chatbots as voting aids. One reason for this is the so-called “vacuum cleaner effect”, whereby AI tools mainly mention parties on the fringes and hardly ever recommend centrist parties. “We sometimes forget, but ChatGPT and similar tools are nothing more than souped-up spell checkers that put appropriate words together. If the programme 'recommends’ certain parties, it is primarily because those parties are talked about more and the programme recognises them better, not because that party is so well aligned with your preferences and views.”

In addition, Schaap emphasises that we must remember which organisations are behind the best-known AI tools. “Google and OpenAI developed these programmes primarily to make money, not to give voting advice or improve democracy. Those goals may happen to coincide, but you can't just assume that.”

Autonomy or efficiency?

And what if a public, non-commercial AI tool becomes available? A glorified voting guide? “I understand that people just want to know who to vote for, but still, and this is mainly my own feeling, I wonder whether you should want something or someone else to determine your voting choice for you. Choices such as who you marry, which job you choose and who you vote for are, in my opinion, a significant part of being human; they seem too important to me to outsource to technology.”

Nevertheless, people seem to be increasingly accepting of rapidly developing technology, much to the surprise of Schaap and his colleagues in his field. “Just ten years ago, studies showed that people were not quick to relinquish their autonomy, but that image is beginning to shift and people are increasingly accepting technologies such as AI because of the convenience and efficiency it brings. For us as scientists, the years ahead are very interesting: will people choose autonomy or convenience? And why do they do that?”

Although Schaap recognises the advantages, he has his reservations about embracing AI too readily. And of course, it's not one or the other. You can also come up with legislation that allows people to use AI, but not outsource everything, for example by prohibiting AI from giving voting advice. “As humanity, we are also capable of cloning people, but we have agreed not to do so. Similarly, we can also agree that people should continue to make their own choices.”

Contact information

Organizational unit
Faculty of Social Sciences
Theme
Behaviour, Artificial intelligence (AI), Politics