AI Helped People Spot Fake News—Then Made Them Worse at It: MIT

An MIT study released this week found that AI assistants initially improved users' ability to detect misinformation, but this effect was temporary and led to a long-term decline in independent fact-checking skills. Participants using AI tools were better at identifying fake news during the study period, with an average accuracy of 78% compared to 69% for those not using AI. However, after a period of reliance on AI, the same participants showed a reduced ability to spot falsehoods without assistance, scoring 65% accuracy compared to a baseline of 72% for a control group. The research, conducted by the MIT Initiative on the Digital Economy, involved 500 participants over a six-week period, exposing them to a mix of real and fabricated news articles. The study suggests that while AI can be a useful tool for immediate fact-checking, over-reliance may diminish critical thinking and independent verification skills. Researchers noted that the AI's explanations for why a piece of news was false were highly effective in the short term, but the underlying learning process for users seemed to be bypassed. This highlights a potential trade-off between convenience and the development of robust media literacy. The findings were published in the journal "Nature Human Behaviour" on May 15, 2024.