Chatbot Conversations Effectively Reduce Conspiratorial Beliefs, Study Finds

A Massachusetts Institute of Technology study found that a 10-minute conversation with ChatGPT decreased conspiracy beliefs by 20%. The AI chatbot, using GPT-4, engaged conspiracy theorists with persuasive evidence, successfully convincing one-fourth to disavow their beliefs. Personalized dialogue tailored to individual conspiracies proved crucial for the rebuttals.

A conversation with a chatbot lasting under 10 minutes was found to mellow down conspiracy theorists, according to a new study from the Massachusetts Institute of Technology. Researchers designed the AI-model ChatGPT to engage participants in tailored dialogues, factually rebutting their conspiratorial claims.

Leveraging advancements in GPT-4, the large language model powering ChatGPT, the researchers found that among over 2,000 self-identified conspiracy believers, conversations with ChatGPT lowered beliefs by about 20%. About a fourth of the participants disavowed their previous conspiracy theories after chatting with the AI-powered bot.

Lead author Thomas Costello noted that many conspiracy believers updated their views when presented with compelling counter evidence. The study, published in the journal Science, highlights the potential of large language models to directly refute evidence supporting conspiratorial beliefs and tailor counter arguments, a unique advantage unavailable to humans.

(With inputs from agencies.)