Article Summary:
Michael Protzman, a former demolition contractor, attracted QAnon conspiracy believers who gathered in Dallas in 2021, expecting to witness John F. Kennedy and John F. Kennedy Jr. resurrect and crown Donald Trump. When this prophecy failed, many followers remained loyal, demonstrating the difficulty of changing deeply held conspiracy beliefs.
Researchers from MIT, Cornell, and Sloan have discovered a promising approach using artificial intelligence to reduce conspiracy theory adherence. In studies involving over 2,000 participants, a GPT-4 Turbo AI conversation partner successfully reduced conspiracy beliefs by 20% across various topics, including COVID-19 and the 2020 presidential election.
The AI was trained to personalize arguments and refute specific evidence supporting conspiracy theories. Participants not only showed reduced belief but also demonstrated increased willingness to disengage from conspiracy-promoting social media and challenge other believers. Remarkably, these effects remained stable even 10 days and two months later.
The researchers suggest that tailored evidence can potentially help people exit “conspiratorial rabbit holes.” The AI’s effectiveness may stem from being perceived as a neutral, non-threatening source, unlike human interactions that might trigger defensive identity-protective responses.
This finding is significant because previous attempts to change extreme beliefs have been unsuccessful. Psychological research has shown that when beliefs become integral to personal identity, people tend to become more entrenched, even when confronted with contradictory evidence.
The AI approach offers a potentially groundbreaking method for addressing the persistent challenge of changing deeply held conspiracy beliefs by providing personalized, non-confrontational reasoning.