AI’s impact on cognitive ability: MIT study reveals more troubling data

Researchers at the MIT Media Lab found that relying on AI to identify fake news decreased participants' unassisted ability to detect misinformation. In a four-week study involving 67 participants evaluating news headlines and images, those who used an AI chatbot were 21% more accurate in identifying fake news with assistance. However, by the end of the study, participants' unassisted ability to identify fake news declined by 15 percentage points compared to their initial performance. Despite this decline, a quarter of participants reported feeling their detection abilities had improved, indicating a misplaced trust in AI capabilities. Anku Rani, a co-lead author of the study, explained to MIT News that users often overlook the limitations of large language models (LLMs), which are statistical models predicting sequences, and their potential impact on users. This finding aligns with previous research suggesting negative impacts of AI reliance on cognitive abilities, including a separate May study where just 10 minutes of AI use impaired participants' cognitive performance.