Seven minutes with a chatbot beat a fact sheet at reducing conspiracy beliefs in two experiments
decoder ·
A study by researchers at Carnegie Mellon, MIT, and Cornell tested whether short conversations with an AI chatbot could reduce conspiracy beliefs shortly after real-world crises. In two experiments conducted within days of the July 2024 assassination attempt on Donald Trump and the September 2025 murder of Charlie Kirk, roughly seven-minute dialogues with Google Gemini outperformed static fact sheets and control chats. The chatbot was given a curated fact base because the events occurred after its training cutoff, and it adapted its persuasion tactics based on how much was known. For the Trump event, it used epistemic humility and Socratic questioning, while for the Kirk case it relied more on factual arguments and societal harms. Participants who had the debunking conversation showed reduced belief in their own conspiracy theory and were later less likely to endorse cover-up narratives about subsequent events, suggesting an inoculation effect. The authors caution that this is a case study and that similar techniques could be misused to convince people of false conspiracy narratives.
decoder