Researchers at the MIT Media Lab have published findings that suggest a troubling trade-off in the use of AI tools for news verification. The study tracked 67 participants over a month as they judged the authenticity of news headlines and images, sometimes assisted by an AI chatbot.
While participants using the chatbot were 21% more accurate at detecting fake news in the moment, their unassisted ability to spot misinformation declined by 15 percentage points by the fourth week. Their confidence in their own judgment actually increased, even as their performance worsened.
The study arrives amid growing reliance on AI for news consumption. Recent Pew Research Center reports indicate that one in five U.S. teenagers now gets news from chatbots, and one in five adults under 50 uses AI for news at least sometimes.
This research adds to a growing body of evidence that AI tools may be eroding cognitive skills rather than augmenting them. The findings raise questions about the long-term effects on media literacy, particularly among younger demographics who are most likely to adopt these tools.
Critics might argue that the study's small sample size limits its generalizability, and that the controlled lab environment may not reflect real-world behavior where users often have access to multiple information sources beyond a single chatbot.