In a research paper published September 13 in Science, MIT authors Thomas H. Costello [in photo above] and David G. Rand, along with Gordon Pennycook, investigated whether dialogs with a generative artificial intelligence (GenAI) interface could convince people to abandon their conspiratorial beliefs. The team used personalized, in-depth conversations with 2,190 Americans in three sessions with a large-language machine-trained chatbot. The treatment reduced participants’ belief in their chosen conspiracy theory by 20% on average, challenging research suggesting that such beliefs are impervious to change. The paper, Durably Reducing Conspiracy Beliefs Through Dialogues with AI, question the conventional wisdom and asked whether it may be possible to talk people out of the conspiratorial “rabbit hole” with sufficiently compelling evidence.
Watch: NBCNightlyNews interview with Costello about the project,
Read: The New York Times coverage of the research.
Several other media accounts followed, and there were comments on social media about how such technology could be used to “deprogram” citizen of their beliefs.
Read the full paper in Science