Relying on AI chatbots can boost short-term accuracy but erode users’ independent judgment over time, according to a Massachusetts Institute of Technology study tracking 67 participants across four weeks. Tools such as ChatGPT and Claude improved participants’ immediate ability to spot fake headlines and images by 21%, yet their unassisted performance fell 15.3% by week four, suggesting critical-thinking skills atrophied with overuse. The researchers found prescriptive AI guidance delivered speed and certainty but discouraged deeper reasoning, and many users mistakenly believed they were improving even as their skills declined. While the sample skewed toward the US and UK and the study’s duration was limited, the findings raise flags for educators and the public as AI becomes a common filter for online information. The authors argue AI systems should emphasize guided questioning to strengthen, not supplant, users’ analytical abilities.
Related article:
