Researchers have come across a surprising capability of artificial intelligence. It is the debunking conspiracy theories. A recent study found that a version of ChatGPT powered by GPT-4 Turbo is capable in persuading individuals to reconsider beliefs in conspiracy theories. Two such examples were that COVID-19 was an intentional population control measure or that the 9/11 attacks were orchestrated by the U.S. government.
The study was conducted by a team led by psychologist Gordon Pennycook from Cornell University and mainly focused on how the human mind interacts with evidence when presented thoughtfully. It involved more than 2,000 volunteers who interacted with the AI by inputting conspiracy-related beliefs. The chatbot assessed the belief and asked participants to provide supporting evidence and it was followed by a rating of their confidence in the belief. The AI managed to reduce the belief of participants in conspiracy theories by an average of 20%. A quarter of the participants saw their confidence in the conspiracy theories drop below 50%.
Pennycook was taken aback by the results. He was initially skeptical of the potential effectiveness of AI. He initially didn’t think that it was going to work. He emphasized on how the research challenged the idea that once people are entrenched in conspiracy thinking.
The key advantage of AI was its ability to generate a wealth of counter-evidence and address logical flaws in real-time.
Psychologist Elizabeth Loftus from the University of California praised the study and said that the AI has helped in reducing overconfidence by showing how much the participants didn’t know. Many volunteers reported feeling that the AI was the first to truly understand and challenge their beliefs with valid counterpoints.
