Recent research from MIT Sloan School of Management and American University has revealed a surprising ally in the fight against conspiracy theories: artificial intelligence. A study led by professors David Rand, Thomas Costello, and Gordon Pennycook shows that AI-powered conversations can effectively reduce belief in conspiracy theories — and these effects last for at least two months.
Read also: 9 Counter-intuitive Trends in AI
In their paper, “Durably reducing conspiracy beliefs through dialogues with AI“, published in Science, the researchers demonstrate that large language models like GPT-4 Turbo can engage conspiracy believers in personalized, evidence-based dialogues that challenge and disprove their beliefs. This technology offers a potentially powerful tool to counter misinformation in ways that human conversations often can’t.
Testing AI on Conspiracy Beliefs
The study engaged over 2,000 conspiracy believers, asking them to explain their beliefs and evidence in their own words. GPT-4 Turbo was then used to craft tailored arguments to address these specific concerns. Unlike the often frustrating human-to-human conversations on conspiracy theories, these dialogues were designed to be personalized and informed by vast amounts of data. The AI spent an average of 8.4 minutes per conversation, addressing the individual’s particular viewpoint and providing relevant counter-evidence.
Long-Lasting Impact
The results were significant: participants’ belief in their chosen conspiracy dropped by an average of 20%, with 1 in 4 disavowing the conspiracy entirely. Even more remarkably, these effects remained two months after the conversations. The AI wasn’t just effective at reducing beliefs across specific conspiracy theories, like those related to COVID-19 or the 2020 U.S. election, but it also shifted participants’ behaviors, such as unfollowing conspiracy theorists online.
The Potential of AI in Social Research
While the study highlights the positive potential of AI to challenge misinformation, it also raises ethical considerations. As AI can be used to convince users of the truth, it could equally be deployed to reinforce falsehoods. Researchers caution that responsible use of AI is crucial to prevent harmful manipulation.
The study’s groundbreaking methodology offers a glimpse into the future of social science research, where AI can revolutionize how we understand and influence human beliefs.
This research challenges the notion that conspiracy theorists are unreachable, suggesting that the right information — delivered in the right way — can change minds.