MIT study finds chatbot news help can weaken judgment

AI-based chatbots can help people keep up with news, but research from Pattie Maes and colleagues at the MIT Media Lab found a risk in relying on them to judge credibility. Participants evaluated paired news headlines and images over the course of four weeks, and were initially 21% percent more accurate at telling fake news from real when aided by a chatbot.

By week four, participants had become 15% worse at identifying fake news without AI than they were before, even though roughly a quarter reported feeling better at it. The researchers described the pattern as an “AI dependency paradox,” which has also been observed in medicine.

Anku Rani, a PhD student in media arts and sciences and one of the study’s lead authors, said users can forget that LLMs are statistical models that predict the next token. Valdemar Danry, SM ’23, said AI systems that provide direct answers tend to encourage reliance, while Socratic questioning can better support independent truth discernment, albeit with more time and effort.