In collaboration with Jon Leopold and Dee Walker

‍

Artificial Intelligence (AI) has entered family life so quickly that two questions already feel past due: Should children use AI? How and when?

A 2026 Common Sense Media² survey found that 70 percent of U.S. teens use AI for schoolwork, while PEW³ has documented similar widespread chatbot⁴ use.

AI enables cognitive offloading¹: the use of external aids to reduce mental effort and conserve resources for more meaningful activities (Grinschgl and Neubauer, 2022; Risko and Gilbert, 2016).

Cognitive offloading is not inherently harmful: calculators and search engines have long reduced mental effort without necessarily diminishing thought.

However, generative AI is different because it can summarize evidence, build arguments and draft polished prose. This allows users to reach an answer while bypassing much of the reasoning that would normally produce it. This process can support adaptive coping, helping individuals regulate stress and sustain mental health.

On the other hand, the same tools may create cognitive overload: an erosion of introspection, over-reliance on algorithmic feedback, and anxiety induced by hyper-monitoring and optimization (Grinschgl et al., 2021; Skulmowski, 2023). The question is not whether AI is good or bad for mental health but how it is reshaping the very architecture of coping (Gerlich, 2025).

Recent research suggests concern. A 2025 Microsoft–Carnegie Mellon study of 319 knowledge workers⁵ found “higher confidence in AI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking”. Yet AI did not simply eliminate critical thinking; it often shifted it from generating material to verifying and integrating responses. The OECD⁶ reached a similar conclusion in its 2026 Digital Education Outlook: general-purpose AI may improve immediate performance without producing equivalent learning gains, while more deliberate uses—such as challenging arguments or providing feedback—can support learning and critical thinking.

AARP research challenges the idea that AI expertise belongs mainly to the young. AI use among Americans over age 50 rose from 9 percent in 2023 to 30 percent in 2025, with even higher use among those still employed. Older adults are also selective: they show interest in uses such as translation and information simplification, while cautious about privacy of personal data.

The generational divide may be less useful than it appears. A teenager may know the newest AI tools before a grandparent. A grandparent who spent decades in medicine, law, education, business, science, or public service may be much quicker to recognize when an answer is technically fluent but weakly supported.

This suggests a different model of inter-generational AI literacy: not one generation teaching another how to use a machine, but people with different forms of expertise examining the machine together. A useful family discussion might therefore begin not with What can AI do? but with harder questions:

• When does convenience become dependency?
• When does assistance become substitution?
• Which mistakes are harmless enough to delegate, and which decisions require accountable human judgment?

These are not questions technology can settle for us but are questions about standards of evidence, responsibility, and judgment. As AI becomes more capable, deciding where those standards of use belong may prove more consequential than mastering any particular AI tool.

Notes

1. Cognitive offloading is the practice of using external tools, physical actions, or digital resources to reduce mental effort and manage information such as using: a shopping list or notes on paper, a calculator for math problems, calendar alerts and phone reminders, GPS for navigation, and AI tools to draft emails or summarize text.
2. Common Sense Media (CSM) is an American nonprofit organization that reviews and provides ratings for media and technology with the goal of providing information.
3. Pew Research Center is a nonpartisan fact tank that informs the public about the issues, attitudes and trends shaping America and the world. It does not take policy positions.
4. A chatbot is a software program designed to simulate conversation with human users through text or voice and may be powered by artificial intelligence.
5. A knowledge worker is a professional whose main capital is their expertise, critical thinking, and specialized knowledge.
6. Organization for Economic Co-operation and Development is an international group of 38 member countries that works to promote economic growth, boost employment, and improve living standards worldwide.

Sources

1. Common Sense Media, Teens in the AI Era: Schoolwork and Skills That Matter
2. Pew Research Center, How Teens Use and View AI
3. Microsoft Research / Carnegie Mellon University, The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers
4. OECD, OECD Digital Education Outlook 2026: Exploring Effective Uses of Generative AI in Education
5. AARP Research — Data Story, Navigating the World of AI: Awareness, Attitudes, and How People Expect to Use It

‍

中文版 / Chinese Version

人工智能、认知卸载¹与一场新的代际对话

作者：Emily Zhang
协作：Jon Leopold、Dee Walker

‍

人工智能（Artificial Intelligence，AI）进入家庭生活的速度非常之快，现如今，有两个问题早就应该摆上台面：儿童是否应该使用人工智能？如果应该，又应如何使用、从什么时候开始使用？

Common Sense Media²于2026年开展的一项调查显示，70%的美国青少年会使用人工智能完成学校作业；皮尤研究中心（Pew Research Center）³的研究也发现，青少年群体已广泛使用聊天机器人⁴。

人工智能使“认知卸载”（cognitive offloading）变得更加容易。所谓认知卸载，是指借助外部工具减少自身的认知投入，从而节省心力，将其用于更有意义的活动（Grinschgl and Neubauer, 2022; Risko and Gilbert, 2016）。

认知卸载本身并不是一个负面的概念。长期以来，计算器和搜索引擎都在帮助人们减少部分认知负担，但这并不意味着人的思考能力必然会因此下降。

然而，生成式人工智能有所不同。它不仅可以检索信息，还能够概括证据、组织论证，并直接生成语言成熟、结构完整的文本。这意味着，用户可能在很大程度上绕过原本产生答案所需要经历的推理过程，直接获得结果。与此同时，这一过程也可能成为一种适应性应对方式（adaptive coping），帮助人们调节压力、维持心理健康。

另一方面，同样的工具也可能造成认知过载（cognitive overload），表现为自我反思能力逐渐削弱、对算法反馈产生过度依赖，以及因持续监测和不断追求“优化”而产生焦虑（Grinschgl et al., 2021; Skulmowski, 2023）。因此，真正的问题并不是人工智能究竟“有利于”还是“有害于”心理健康，而是它正在如何重新塑造人类自身的应对机制及其基本结构（Gerlich, 2025）。

近期研究已经提出了一些值得关注的问题。

微软与卡内基梅隆大学于2025年针对319名知识工作者⁵开展的一项研究发现：“对人工智能越有信心，与越少投入批判性思维存在相关性；而个体对自身能力越有信心，则会在批判性思维方面更胜一筹。”

然而，人工智能并非仅仅让批判性思维消失。相反，它往往改变了批判性思维发生的环节：人们在生成内容上投入的思考有所减少，而在核实信息、整合回答以及判断结果是否合适等方面投入更多认知活动。

经济合作与发展组织（OECD）⁶在《2026年数字教育展望》（Digital Education Outlook 2026）中也得出了类似结论：通用型人工智能可能提升使用者在具体任务中的即时表现，却未必带来与之相当的学习成效；相比之下，如果更有意识地使用人工智能——例如让其质疑已有论点，或针对学习成果提供反馈——则有可能促进学习和批判性思维。

美国退休人员协会（AARP）的研究则反驳了“人工智能主要是年轻人的专长”这一观念。美国50岁以上人群的人工智能使用率从2023年的9%上升至2025年的30%；在仍处于就业状态的50岁以上人群中，使用比例更高。

与此同时，年长者对人工智能的态度并非简单的排斥，而是具有明显的选择性。例如，他们对翻译、信息简化等用途表现出较高兴趣，同时对涉及个人数据隐私的人工智能应用保持更加谨慎的态度。

因此，所谓“代际鸿沟”或许并没有想象中那么重要。

一名青少年可能会比祖父母更早了解最新的人工智能工具；但一位在医学、法律、教育、商业、科学或公共服务领域积累了数十年经验的祖父母，却可能更快识别出：某个答案虽然在专业表达上十分流畅，却缺乏充分、可靠的证据支持。

这提示我们，或许应该建立一种不同的代际人工智能素养模式：它并不是由某一代人教另一代人“如何使用机器”，而是让拥有不同知识、经验和专业判断的人共同审视人工智能的输出及其使用方式。

因此，一场真正有价值的家庭讨论，或许不应该从“人工智能能做什么？”开始，而应该聚焦一些更难回答的问题：

• 便利何时会变成依赖？
• 辅助何时会变成替代？
• 哪些错误的代价较低，可以交由人工智能处理？哪些决定则必须由能够承担责任的人作出判断？

这些问题，并不是技术本身能够替我们回答的。

它们真正关乎的是我们如何理解证据标准、责任与判断。随着人工智能能力不断增强，决定在什么情况下、以什么标准使用人工智能，或许会比掌握任何一种具体的人工智能工具更加重要。

注释

1. 认知卸载（cognitive offloading）是指借助外部工具、身体动作或数字资源来减少认知投入、辅助信息管理的做法。例如：用购物清单或纸质笔记帮助记忆，使用计算器解决数学问题，设置日历提醒和手机提醒，借助 GPS 导航，以及使用人工智能工具起草电子邮件或总结文本。
2. Common Sense Media（CSM）是一家美国非营利组织，主要对媒体内容和科技产品进行评测与评级，旨在为公众提供相关信息，帮助其作出更明智的选择。
3. 皮尤研究中心（Pew Research Center）是一家无党派事实研究机构，致力于向公众提供有关美国及全球重要议题、社会态度和发展趋势的信息。该机构不对具体政策立场表态。
4. 聊天机器人（chatbot）是一种通过文字或语音与人类用户进行对话的软件程序，旨在模拟人与人之间的交流。这类程序可以由人工智能技术驱动。
5. 知识工作者（knowledge worker）是指主要依靠自身的专业能力、批判性思维和专门知识开展工作的专业人员。
6. 经济合作与发展组织（Organisation for Economic Co-operation and Development，OECD）是一个由38个成员国组成的国际组织，致力于促进经济增长、扩大就业并改善全球生活水平。