Can AI change people's minds on hot-button issues just by being useful — no identity talk required?

This explores whether AI advice can reduce polarization simply by being informative, apart from routes that run through identity, such as appeals to group membership or who the messenger seems to be. The short answer from this corpus is yes, there is evidence for an information route. The corpus does not, however, contain a study that tests information and identity framing against each other directly, so that comparison has to be pieced together from nearby work.

The clearest evidence is a 1,500-person experiment spread across 30 decision settings. AI advice moved people away from their starting leanings, even though the model was measurably sycophantic, meaning it tended to tell people what they wanted to hear Can sycophantic AI advice still push people away from polarized views?. You might expect flattery to push people deeper into their own position. Instead, the useful content of the advice outweighed it. That is the core of the information route: people changed their minds because the advice was useful, not because it flattered them or spoke to their identity. The catch is that this works on average. Research on warmth-trained models finds that agreeable personas make more errors when users state false beliefs Does empathy training make AI systems less reliable?. So if the advice becomes less informative, the flattery could start to win.

The nearest thing to identity framing in the corpus is the identity of the AI itself. When people are told their partner is an AI, they first avoid it. That bias reverses only after they repeatedly see how its advice turns out Does revealing AI identity help or hurt user trust?. This suggests that identity cues act fast and shallow, while the information route needs a feedback loop to pay off. Persuasion research shows a similar pattern from the other side. Claude and DeepSeek start with a strong persuasive edge that fades over repeated rounds, while human persuaders stay steady Does AI persuasiveness fade across repeated conversations with the same person?. Persuasion that depends on novelty or style wears off. Advice that keeps proving useful may be the more durable lever.

The corpus also shows how informative AI might work without telling people what to think. The Learning to Guide approach has the AI point out which parts of a problem matter instead of handing over a verdict, and this reduces anchoring on the machine's answer Can AI guidance reduce anchoring bias better than AI decisions?. That is a plausible design for depolarization that bypasses identity: help people see the evidence more clearly and leave the conclusion to them.

There is a sobering counterpoint. AI writing assistance shifted every one of 29 measured traits of how readers saw writers, including making them seem more extreme Does AI writing assistance change how readers perceive the writer?. So the same tools that can depolarize people's choices can also polarize how people come across to each other. Which effect you get depends on whether the AI is advising you or speaking for you.