Link: In two experiments with 2,190 American conspiracy theorists, conversations with GPT-4 Turbo reduced their belief in conspiracy theories by about 20% on average (Jennifer Ouellette/Ars Technica)

Conspiracy theories are widely believed across the U.S., with up to 50% of Americans subscribing to them. These beliefs are tough to change, often intensifying when challenged with facts.

A recent study published in Science reveals that AI chatbots may effectively reduce belief in conspiracy theories by offering personalized counterarguments. This approach has shown promise even two months after the initial conversation.

The AI chatbots custom-tailor responses, addressing each conspiracy believer's specific arguments. This method contrasts with general approaches, which often fail due to the diverse nature of individual conspiracy beliefs.

In experiments involving 2,190 participants, these tailored chatbot conversations led to a 20% reduction in conspiracy beliefs. The conversations were based on detailed inputs from participants about their specific conspiracy theories.

The persistence of the reduction in false beliefs is notable, according to researchers, although it may not completely eliminate such beliefs. This breakthrough suggests a new direction for debunking misinformation effectively.

 #

--

Yoooo, this is a quick note on a link that made me go, WTF? Find all past links here.