Findings suggest that we’re less likely to stand our ground with AI agents, giving it an edge in debates.
23 October 2026
By Emma Young
We are increasingly seeking not just facts from AI, but opinions, advice, and even companionship. In fact, note the authors of a recent paper in the Journal of Personality and Social Psychology, many people receive guidance from AI on everything from which books or movies they might want to watch to which hotels to stay in, or which recipes or products to try.
"But can conversational AI wield enough persuasive influence to meaningfully substitute for human advice and opinions?" wondered Geoff Tomaino at the University of Florida and SungJin Jung at the Chinese University of Hong Kong. In fact, their study suggests that AI can wield more.
In the first of a series of online studies, 393 participants used a 21-point scale to rate the strength of their agreement (or disagreement) with the statement, 'It is better to exercise in the mornings than in the evenings.' Then, they were told they were going to be matched with an AI-powered chatbot or another human participant to discuss this topic for two minutes. After the chat, they used the same scale to again rate their endorsement.
In fact, all of the participants were paired with Chat GPT-4, which was instructed to try to change the participant's mind. And when the researchers calculated the level of attitude change (in terms of the number of points the person shifted on the scale after the chat), they found that those who were told they'd be talking with AI experienced a significantly bigger shift. The AI group had an average 3.33-point change, compared with 1.94 points for those who thought they'd been talking to a person.
The researchers analysed all the conversations and found that the participants were less argumentative when they thought they were talking with AI rather than with a human. They think this explained why the 'AI-framed' partner was more effective at altering their opinions.
In a second study, they found the same pattern of results for a different statement ('It is better to read ebooks than physical books'). Then, in further experiments, they gathered evidence to suggest that people think AI is less persuadable than humans, "making attempts to sway its opinion seem futile."
This finding could explain why participants were less likely to stand their ground in discussions with what they were told was AI. But when it comes to the question of why they were more easily persuaded by the AI's arguments, there are a few possible explanations.
It may be that a less argumentative conversational tone left them more open to AI arguments. Perhaps, though, they also felt that AI may have a more informed, objective, dispassionate view on a topic, potentially making its opinions more valuable than a stranger's. Whether people are more persuadable by AI than a human expert, or a person they trust — such as a close friend — would need further study to explore.
Whatever the exact mechanisms, though, as the researchers conclude, the findings "suggest that LLMs can serve as potent influencers of opinion." You may think that this is a good or a bad thing, they add — at least, until an AI changes your mind.
Read the paper in full:
Tomaino, G., & Jung, S. (2026). From people to prompts: Understanding susceptibility to persuasion in conversations with artificial intelligence. Journal of Personality and Social Psychology. Advance online publication. https://doi.org/10.1037/pspi0000531
Want the latest in psychological research, straight to your inbox?
Sign up for Research Digest's free weekly newsletter.
SOURCE:
No comments:
Post a Comment