🌿English Version

As AI tools become woven into everyday life, researchers are beginning to warn that our growing dependence on large language models（LLM 大语言模型） may come with hidden cognitive costs. At MIT, research scientist Nataliya Kosmyna noticed that many students were forgetting course content more easily than before, while internship cover letters looked suspiciously polished and similar. These observations led her to investigate whether AI was subtly reshaping how people think.

A growing body of research now suggests that “cognitive offloading（意译为认知外包）” to AI—outsourcing tasks like writing, summarizing, or even deciding—may weaken key mental abilities. Studies show that young people, in particular, could be vulnerable to declines in skills such as critical thinking(批判性思考) and memory when they rely too heavily on AI-generated answers.

Kosmyna’s own experiment offers an early glimpse(一瞥) of this effect. She asked 54 students to write short essays under three conditions: using ChatGPT, using Google search, or using no tools at all. Brainwave measurements revealed striking differences. Students who wrote without AI showed widespread neural activity, while those using ChatGPT displayed up to 55% less activation in regions linked to creativity and information processing. Many could not recall what they had written afterward, and several felt they had “no ownership” of their work. 

Other studies echo this pattern. Researchers at the University of Pennsylvania describe a phenomenon they call “cognitive surrender(认知投降)”—a tendency for people to accept AI output with minimal scrutiny(审查), even when it contradicts their own intuition. Similar effects appear in high‑stakes environments: medical professionals who used AI to screen(筛查) for colon cancer(结肠癌) became worse at spotting tumors(肿瘤) on their own after only three months of assistance.   

The long‑term consequences remain uncertain, but early signs are troubling. In a follow‑up test four months later, students who had previously relied on ChatGPT showed weaker neural connectivity when asked to write without AI support, suggesting they had not deeply engaged with the earlier tasks(暗示他们在之前的任务中并没有认真投入). Some researchers worry that chronic (长期的，慢性的)underuse (未充分利用) of cognitive effort could contribute to cognitive decline later in life.

Yet the article also highlights a more hopeful path. A small group of participants—fewer than 10%—used AI not as a replacement for thinking, but as a tool for gathering information that they then analyzed themselves. These individuals produced more accurate work and showed stronger brain activation.  

Experts argue that the healthiest approach may be a form of “hybrid intelligence(混合智力)”, where humans think first and use AI to challenge or refine their ideas rather than to generate them. Techniques such as asking AI to act as a “nemesis(死对头)” that critiques your reasoning, or prompting it to ask questions instead of giving answers, can help maintain cognitive engagement.

The message is clear: AI can support our thinking, but only if we resist the temptation to let it think for us. For long‑term brain health, creativity, and independence of mind, we must continue to challenge ourselves—because deep thinking remains our most powerful human advantage.

🌿 中文版，不是逐句直译，但内容结构与英文一致

随着 AI 工具越来越深入日常生活，研究者开始警告：我们把越来越多的思考任务交给大型语言模型，可能正在付出认知上的代价。

在 MIT，研究科学家 Nataliya Kosmyna 注意到，学生们比几年前更容易忘记课堂内容，而求职者的求职信也变得异常相似、过度精致。她怀疑，AI 正在悄悄改变人们的思维方式。

越来越多的研究显示，把写作、总结、判断等任务“外包”给 AI，会削弱关键的认知能力。年轻人尤其容易受到影响，例如批判性思维和记忆力。

Kosmyna 的实验提供了一个早期信号。她让 54 名学生在三种条件下写短文：使用 ChatGPT、使用 Google 搜索（关闭 AI 摘要）、完全不用工具。脑电数据显示：不用 AI 的学生大脑活动最活跃；使用搜索引擎的学生仍保持一定活跃度；而使用 ChatGPT 的学生，大脑相关区域的活动减少了 高达 55%。许多人甚至无法回忆自己写过什么，并觉得作品“不属于自己”。

其他研究也出现类似现象。宾夕法尼亚大学的团队提出了“认知投降（cognitive surrender）”：人们在使用生成式 AI 时，会不加质疑地接受答案，甚至让 AI 覆盖自己的直觉。在医疗场景中也出现了类似问题：使用 AI 辅助筛查结肠癌三个月后，医生在没有 AI 的情况下反而更难发现肿瘤。

长期影响尚不明确，但早期迹象令人担忧。在四个月后的追踪测试中，曾依赖 ChatGPT 的学生在没有 AI 的情况下写作时，大脑连接度更弱，暗示他们在第一次任务中并未真正投入思考。

不过，文章也指出了更积极的方向。少数参与者（不到 10%）把 AI 当作信息来源，而不是替代思考的工具。他们的表现更好，大脑活动也更强。

专家认为，最健康的方式是“混合式智能（hybrid intelligence）”：先自己思考，再用 AI 来挑战或完善想法，而不是让 AI 直接给答案。比如让 AI 扮演“反对者”来指出你的错误，或让它只提问不回答，以保持你的思考活跃。

结论很明确：AI 可以帮助我们，但前提是我们不能让它替我们思考。为了长期的脑部健康、创造力和独立判断力，我们必须继续主动动脑，因为深度思考仍然是人类最强大的能力。

📌 原文长句拆解

下面三句都来自BBC Future原文，我们引用来做一下结构拆解。看多了就发现，其实BBC的长句并不难。真正难的还是精准的动词运用。

① “The concern that researchers like Kosmyna have is that if we become too reliant on AI, it could affect the language we use and even our ability to do basic cognitive tasks.”

bbc.co.uk

意思：研究者（如 Kosmyna）担心，如果我们过度依赖 AI，它可能会影响我们的语言使用方式，甚至影响我们完成基本认知任务的能力。

结构拆解：

• The concern … is that …主干结构：某个担忧是……→ 一个典型的“名词 + be + that 从句”结构。

• that researchers like Kosmyna have限定 concern 的定语从句→ “研究者所拥有的担忧”

• if we become too reliant on AI条件从句→ “如果我们过度依赖 AI”

• it could affect the language we use主句部分 1→ “它可能影响我们使用的语言”

• and even our ability to do basic cognitive tasks并列结构→ “甚至影响我们做基本认知任务的能力”

② “One study by researchers at the University of Pennsylvania suggests that some people undergo something they term ‘cognitive surrender’ when using generative AI chatbots.”

bbc.co.uk

意思：宾大研究者的一项研究指出，有些人在使用生成式 AI 聊天机器人时，会经历一种他们称为“认知投降”的现象。

结构拆解：

• One study … suggests that …主干：某项研究表明……

• by researchers at the University of Pennsylvania介词短语修饰 study→ “由宾大研究者进行的”

• that some people undergo something …宾语从句→ “一些人会经历某种现象”

• they term ‘cognitive surrender’省略关系代词的定语从句→ “他们把它称为 cognitive surrender”

  * 注意这里用term表示动作“称为”

• when using generative AI chatbots时间状语→ “在使用生成式 AI 聊天机器人时”

③ “A recent multinational study team found that medical professionals who used an AI tool to screen for colon cancer for three months were subsequently worse at spotting the tumours without it.”

bbc.co.uk

意思：一项跨国研究发现，使用 AI 工具筛查结肠癌三个月的医生，在没有 AI 的情况下反而更难发现肿瘤。

结构拆解：

• A recent multinational study team found that …主干：某研究团队发现……

• medical professionals … were worse at spotting the tumours宾语从句主干：医护人员更不擅长发现肿瘤

• who used an AI tool to screen for colon cancer for three months限定 medical professionals 的定语从句→ “那些使用 AI 工具筛查结肠癌三个月的医护人员”

  *注意这里用screen表示医学动作“筛查”

• were subsequently worsesubsequently = 随后、之后→ 强调“使用 AI 后的负面变化”

• at spotting the tumours without it介词短语→ “在没有 AI 的情况下识别肿瘤”