A new study shows that a brief chat with an AI chatbot, such as ChatGPT, can significantly reduce belief in conspiracy theories. The research, conducted by a team including experts from the Massachusetts Institute of Technology, indicates that even a conversation lasting less than ten minutes can challenge deeply held conspiracy beliefs, including those related to the COVID-19 pandemic.
The researchers designed ChatGPT to be particularly persuasive, using advancements in GPT-4, a powerful language model. This AI system is capable of generating tailored responses and factually debunking misinformation presented by participants.
In the study, over 2,000 individuals who identified as conspiracy believers interacted with ChatGPT. The results showed a reduction in their conspiratorial beliefs by about 20 percent on average. Notably, approximately 25% of these participants completely renounced their previously held conspiracy theories after the conversation.
The research, published in the journal Science, involved two experiments. Participants first detailed a conspiracy theory they believed in, including evidence supporting it, and then engaged in an 8.4-minute dialogue with the chatbot. A control group discussed an unrelated topic with the AI.
Lead author Thomas Costello, an assistant professor of psychology at American University, noted, “Many conspiracy believers were willing to reconsider their views when presented with convincing counter-evidence.” He highlighted that the AI’s responses were not only detailed but also engaging, helping to build rapport with the participants.
The study reveals that providing persuasive, factual information to conspiracy theorists has been challenging in the past, primarily due to the diversity of conspiracy theories and the well-informed nature of believers. Traditional methods of debunking often struggled with these issues, as they required guessing the specific beliefs of individuals.
Co-author Gordon Pennycook, an associate professor of psychology at Cornell University, pointed out, “The AI’s ability to address specific arguments with strong counter-evidence offers a unique opportunity to see how people respond to factual rebuttals.” The AI’s rapid adaptability and precise targeting of individual beliefs offer a significant advantage over traditional debunking methods.
This research suggests that AI, by connecting and addressing various topics swiftly, can effectively counteract conspiracy theories in ways that may be beyond human capability.
OpenAI Set to Launch New AI Model 'Strawberry' for ChatGPT in a Fortnight