A recent study suggests AI, particularly ChatGPT-4, can hold its own in a debate, and even out-persuade humans when it knows a bit more about who it's talking to.
Researchers at the Swiss Federal Institute of Technology in Lausanne ran 600 online debates.
Participants were matched either with a human or with the AI and assigned a position on topics like school uniforms or abortion.
Before and after each debate, they rated how much they agreed with the argument.
When no personal data was shared, AI and humans were equally persuasive.
But when information like age, gender, or political leanings was available, AI came out on top, changing opinions more often than human debaters in 64% of those cases.
The study found:
AI is as persuasive as humans, and more so when it has personal data
It works best on neutral or low-stakes topics by adapting arguments to individuals
There's growing concern over misuse, especially in politics and online radicalisation
The researchers say AI’s edge comes from its ability to tailor arguments in a more analytical, structured way, and to adapt to each individual.
It’s not just making solid points; it’s making the kind of points that resonate with the person it’s talking to.
While the findings raise concerns, especially around how this could be used to influence elections or spread misinformation, there’s also potential for positive use.
AI could help challenge conspiracy thinking or support healthier decision-making.
Not AI reading you like a Buzzfeed quiz 💀