Using AI to escape conspiracy rabbit holes

A new study published in Science offers a promising insight into how artificial intelligence might foster a more constructive civic discourse. The research shows that personalized dialogues with AI can reduce belief in conspiracy theories, which are generally viewed as resistant to fact-based counterarguments. By tailoring responses to specific conspiratorial claims, AI systems can provide evidence-based refutations that resonate with individuals, even those deeply entrenched in their beliefs.

Study participants who engaged in a three-round conversation with large language AI models reduced their belief in conspiracy theories by 20 percent on average, and this effect persisted for at least two months. Further, the study reveals that AI interventions don't just debunk individual theories, they also weaken broader conspiratorial mindsets.

The study challenges the belief that conspiracists are impervious to evidence. With the right tools, even deeply held conspiratorial views can be shifted. As the authors put it, “conspiratorial rabbit holes may indeed have an exit.” And that exit can put us on a path to a more informed and constructive public discourse.