It's very difficult to change an individual's opinion. Nowadays, this is all the more true as points of view are polarized and seem to be firmly anchored in everyone's psyche. Is it even possible to convince someone else on any subject? It seems so... by being an artificial intelligence.
The Swiss Federal Institute of Technology in Lausanne conducted an experiment with dozens of people invited to debate. Some were to be put in written contact with another human candidate or with a robot using ChatGPT 4.0. The algorithm, moreover, had been fed personal information for each interlocutor. They had to give their point of view on one of the sixty questions in two paragraphs, receive the other opinion and then judge whether the exchange had changed their mind.
This was 8 times more common among those who discussed with an algorithm than with another human. This is because the algorithm can use the thousands of data points it has accumulated in psychology and marketing, and adapt these with knowledge of the person it is talking to, to find more convincing formulations. They were better able to avoid demonizing the other person's point of view, and come up with solid counter-arguments. Interestingly, participants who often identified the robot were far more accepting of having been convinced by an AI than by another human. So much for ego in the face of a machine with millions of data points.
Running time: 10:54
To listen to the interview
Learn more about this
resource
See more resources from this institution