UC Berkeley研究员Brian Christian团队发布了一项同行评审研究，揭示了使用AI工具对人类认知能力的短期负面影响。1 通过对1,222名在线参与者进行的随机对照试验，研究发现获得AI辅助的参与者在工具被移除后，准确性和持久力迅速下降。1
研究分三个阶段进行了测试。1 第一实验中，354名参与者在完成12道分数问题后被移除AI工具，结果显示AI用户随即停止准确答题。1 第二实验涉及667名参与者，得出相同结论——AI用户在移除工具后表现下降或选择放弃。1 第三实验让201名参与者参加SAT阅读理解测试，再次证实了这一趋势。1
研究论文题为《AI辅助减少坚持力并损害独立表现》，指向一个重要问题：过度依赖AI会侵蚀所谓的"生产性奋斗"，可能对人类的认知能力和专业技能造成损害。1 Christian表示："你几乎找不到一个更清晰的故事"，并指出："我们需要朝着一个方向努力，在这个方向上人类能力尽可能多地得到增强而非取代。"1 研究团队来自Carnegie Mellon、MIT、Oxford和UCLA等多个机构。1
A peer-reviewed study led by UC Berkeley researcher Brian Christian reveals that using AI tools for as little as 10 minutes significantly weakens people's ability to focus and persist on challenging problems.1 The randomized controlled trial involving 1,222 online participants demonstrated a consistent pattern across multiple experiments: participants who received AI assistance experienced an immediate decline in accuracy and determination after the AI tool was removed, while those who never used AI maintained stronger performance.1
The research team conducted three separate experiments to test this effect. In the first, 354 participants tackled scoring problems before having their AI access withdrawn, after which AI users promptly stopped answering accurately.1 A second experiment with 667 participants produced identical results—those who had used AI either performed worse or abandoned the task once the tool was unavailable.1 A third experiment involving 201 participants taking SAT reading comprehension tests confirmed the same trend.1 The findings suggest that overreliance on AI corrodes what researchers call "productive struggle," potentially damaging human cognitive abilities and professional skills.1
Christian stated that the results were striking, noting "You almost couldn't ask for a clearer story."1 Looking forward, he emphasized the importance of shaping AI's role responsibly: "We need to work toward a future in which human abilities are, as much as possible, augmented rather than supplanted."1 The research team includes scholars from Carnegie Mellon, MIT, Oxford, and UCLA in addition to UC Berkeley.1
评论
还没有评论，欢迎留下第一条。