Study: just 10 minutes of AI use makes people give up more easily once help is removed; 3 RCTs, 1,222 participants (report)

Verified (fact)

Researchers led by first author Grace Liu of Carnegie Mellon University, with UC Berkeley's Center for Human-Compatible AI (CHAI), MIT, the University of Oxford and UCLA, ran three randomized controlled trials with 1,222 participants in total. In the first experiment (354 participants, 307 analyzed), a group solving fraction problems with ChatGPT lost AI access after the 12th problem. Afterward, accuracy was 57% for the AI group versus 73% for the control group. Skip rates were 20% and 11%, respectively. The second experiment, with 667 participants, added a pre-test and gave the control group the same kind of assistance screen. Berkeley News covered the study on Oct. 9.

Needs fact-check (claim)

The researchers argue that relying on AI for only about 10 minutes reduces both persistence in finishing difficult tasks alone and accuracy.

Why it matters (interpretation)

The study is notable for confirming, through randomized controlled trials, that AI reliance affects learner persistence even after short use. It provides evidence that AI tutors or educational chatbots that give answers immediately may harm long-term learning.

Sources

• Press AI 10분 사용, 인내심 떨어뜨린다…버클리 연구 결과 ZDNet Korea
• Community AI를 단 10분만 사용해도 어려운 일을 끈기 있게 해내는 능력이 약해짐 GeekNews