In a recent study, researchers found that AI, specifically ChatGPT, can effectively challenge and debunk conspiracy theories, surprising even the scientists conducting the experiment. By engaging with over 2,000 volunteers, the AI was able to lower belief in false theories—such as those about Covid-19 and 9/11—by 20% on average. The study revealed that, contrary to popular belief, people are responsive to evidence when presented effectively.
The experiment utilized GPT-4 Turbo to interact with participants about their conspiracy theories, providing real-time counter-evidence and addressing logical flaws. This method proved more successful than traditional human attempts, highlighting the AI's ability to match and counteract the extensive, albeit often misleading, evidence that conspiracy theorists gather.
Psychologist Gordon Pennycook of Cornell University noted that the study was unexpectedly uplifting, challenging the notion that people deeply entrenched in conspiracy theories are impervious to reason. Elizabeth Loftus from UC Irvine remarked that the AI's ability to present information and reduce overconfidence contributed to its success.
The findings, published in Science, suggest that AI could play a significant role in combating misinformation by presenting balanced and logical counterpoints, thereby addressing the skepticism of conspiracy theorists.