神经科学与心理学2021pnas.org
人们把互联网的知识误认为自己的知识
People mistake the internet’s knowledge for their own
进阶 · 约 35 分钟读完前置知识：实验心理学基础、统计推断、元认知
导读
这篇 PNAS 论文研究互联网搜索如何影响人们对自身知识的判断。通过实验，作者发现：当人们可以上网查资料时，更容易把互联网上的知识误当成自己掌握的知识，从而高估自己的知识水平；这种“知识错觉”可能模糊内部记忆与外部信息的边界。核心结论是，便捷的网络访问会让人产生更强的“我知道”的感觉，即使实际知识仍在网上。
HN 关注点：是因为它用实验解释了程序员和现代知识工作者日常依赖搜索引擎、Stack Overflow 时产生的“我懂”错觉，并引发关于记忆外包与深度学习的讨论。
HN 评论精选
这几条评论里：补充 2 亲历 2
亲历程序员自述高度依赖互联网，真正属于自己的只是判断力与信息筛选能力。
作为一名程序员，我觉得没有互联网什么都做不了。哪怕只是某个特定网站（也许其实是两个：SO 和 GH）宕机，也会毁了我的一天，也会毁了很多人的一天。我的编码方式与能否访问在线资源紧密相连。我经常搜索诸如如何拼接字符串、某语言 for 循环语法之类的东西；也会用互联网查更高层的问题，比如某系统的内存管理如何工作，或 ECS 架构如何运作。我还花大量时间寻找适合放进自己系统的组件，所以 GitHub 挂了我会很烦。基本上没有互联网我就废了。编码工具本身、它们的所有用法示例，以及关于一切如何运作的实际知识都在上面。也许真正属于我自己的输入，只有判断哪些东西重要、哪些来源可靠、哪些人是权威。
lordnacho查看英文原文
As a coder I don't think I can do anything without the internet. Even if one specific site (maybe there are two actually, SO and GH) were down, it would wreck my day and many of yours. My style of coding is very intimately connected to having access to online resources. I regularly search for things like how to concatenate strings or the syntax of a for loop in some language. I also use the internet for higher level things like how memory management works on some system, or how something like an ECS architecture works. I also spend a lot of time looking for the right components to put into my own systems, so if GitHub were down it would bother me. Basically I'd be useless without the internet. The coding tools themselves, all the examples of how to use them, and all the actual knowledge about how everything works is on there. Perhaps the only thing that's actually my own input is the judgement about what things are important, which sources are reliable, and which people are authorities.
补充外部查询会打断思考，适度记忆能提升流畅度、沉浸感和深度学习。
就我个人而言，我开始发现很多情况下，去外部资源查东西对心流的打断比记住它更深，哪怕查询速度一样快甚至更快。（也有可能外部查询只是看起来更快，因为几个机械步骤压缩了我对时间的感知，就像键盘导航看似比鼠标快，实际未必。）例如，当我频繁使用某个 API、经常查同样的操作时，花二十分钟找出并复习重要操作——就像准备一场不许查外部资料的考试——会在熟练度和沉浸感上带来回报。另一个例子是读历史书前，我会先复习相关人名和同时代日期；这样当我想到“等等，这时 X 是多久以前？Y 发生了吗？”就能凭记忆回答。这比不得不脱离书的语境去查日期带来更丰富的阅读体验。我知道从教学角度看记忆常被负面看待，学生也觉得它疏离、令人沮丧；但我认为当你有足够经验理解它的价值，以至于达到它的努力本身不再那么负面时，审慎地运用记忆能大大让工作更轻松、学习体验更丰富。
dkarl查看英文原文
Personally, I'm starting to identify a lot of cases where looking things up in an external resource is a deeper mental disruption than remembering it, even if the lookup is just as fast or even faster. (Or, it's possible that the external lookup only seems faster because the several mechanical steps compress my perception of time, like keyboard navigation can seem faster than the mouse, even when it isn't.) For example, I find that when I'm working with an API enough that I am frequently looking up the same operations, investing twenty minutes in identifying and reviewing the important operations, as if I were preparing for an exam where I wouldn't be allowed to consult external resources, pays off in fluency and immersion. Another example is that when I am reading a history book, before I start, I review relevant names and contemporary dates. Then when I find myself thinking, "Wait, at this time, how long ago was X? Has Y happened yet?" I can answer from memory. This gives me a richer reading experience than if I needed to remove myself from the context of the book to look up those dates. I know memorization is seen negatively from a pedagogical standpoint, and schoolchildren find it alienating and discouraging, but I think when you have enough experience to understand the value of it, so that the work to achieve it isn't such a negative experience in itself, judicious application of it has a lot of power to make your work easier and your learning experiences richer.
亲历资深开发者从记忆事实转向记忆流程/决策树，主动外包可查细节。
我开始开发软件时还没有互联网，那确实很不一样。自从可搜索信息无处不在，我对记住什么、如何记忆的选择已显著改变。我不再需要在桌上堆 K&R、处理器手册、Inside Macintosh 或 Java 书。现在桌上唯一的技术书恰好是《Anti-patterns》。过去几年我一直在“清理”纸质书，尤其是技术书，已经扔掉整整一架子。我仍保留一整架 Knuth TAoCP、数学、类型论、图形学、图论、控制论、物理，以及一些怀旧书，比如我最初的 Motorola 68000 处理器指南、Rodney Zaks 的 Z80 书、《BASIC 冒险游戏编程》、Commodore 64 Innerspace Compendium 等。当小狗咬坏我的《Linux in a Nutshell》时，我本想买新版，但决定通过公司的 O'Reilly Safari 在线试用一个月，结果两年多都没再补纸质版。我会把记忆风格的主要变化概括为：从专注于可查的细节和事实，转向用记忆承载决策树、流程和方法论——这些要么查不到，要么经过显著的个人定制。我有意识地把更简单、对我更没价值的记忆方面“外包”出去。其中一些是自动处理的……
bondolo查看英文原文
I started developing software in the days before the Internet. It was indeed very different. My choices about what to remember and how I remember things have changed significantly since the ubiquitous availability of searchable information. I no longer have tomes of K&R, processor manuals, Inside Macintosh or Java books on my desk. The only technical book which happens to be on my desk at the moment is "Anti-patterns". I have been "weeding" my paper books, especially technical books over the last couple of years and gotten rid of an entire bookshelf of books. I still have an entire bookshelf with Knuth TAoCP, math books, type theory, graphics, graph theory, control theory, physics and some nostalgic books like my original Motorola 68000 processor guide, Rodney Zaks Z80 book, "Adventure game programming in BASIC", Commodore 64 Innerspace Compendium, etc. When a puppy chewed up my copy of "Linux in a Nutshell" I was tempted to buy the new edition but decided to try it online for a month with O'Reilly Safari through work and have not bothered to replace the paper copy for more than 2 years. I would categorize the main change in style of memory as a shift from focus on remembering details and facts, which I can look up, to using memory for decision trees, processes and methodologies which either can't be looked up or are significantly personally customized. I consciously "outsource" the simpler and to me, less valuable, aspects of memory. Some of this is handled automatically by the IDE I use and the rest is done with mostly Google searches.
补充深度思考和跨领域连接需要信息在记忆中，记忆虽有成本但收益巨大。
我常听到一种思路：既然能查，为什么还要深入学习？答案是：除非信息能轻易从记忆中调取，否则你无法对它进行遐想或深入思考。遐想、大脑的“默认模式”会组织并帮助理解信息。如果我在学校里用 Anki 之类的工具认真学习和记忆各科信息，我会有更多信息来连接不同想法。除非信息被记住，大脑无法充分发挥这种能力。话虽如此，也要知道记忆有时间成本。单个数据点花的时间很少，但在 Anki 里记 2—3 万个数据点是认真的时间投入。如果它对你重要，就去记。收益很大，也有助于抵御晚年的认知衰退。
moosey查看英文原文
I often hear a thought process to the effect of: > Why learn something in depth when it can just be looked up? And the answer is that you can't daydream about something or think about it deeply unless that information is easily pulled up from your memory. Daydreaming, out the "default mode" of the brain organizes and helps is too understand information. If I had taken the care to study and memorize information regarding, well, everything in school, via a tool like Anki, I would have a lot more information that I could use to connect disparate ideas together. The brain can't do this to the full extent possible unless that information is memorized. That being said, it's important to know that memorization has a cost in time. The time for a single data point is low, but 20-30k data points in Anki is a serious time commitment. If it's important to you, memorize. The benefits are huge, and it will help too ward off mental declines later in life.
相关资源
从这篇论文所有 HN 讨论帖的全部评论里整理出来的链接。
- 视频与演讲
历次讨论
- 2021Hacker News 讨论帖304 分 · 189 评
同领域还值得读
神经科学家能理解微处理器吗？（2017）
Could a Neuroscientist Understand a Microprocessor? (2017)
提出一个思想实验：把神经科学常用的分析方法（损伤实验、电生理、功能定位等）应用到一个我们完全了解其原理的工程系统——6502 微处理器（运行游戏）上，检验这些方法能否真正“理解”…
神经科学与心理学2017 · 411 分 · 135 评进阶 · 约 60 分钟编辑精选大脑衰老呈现非线性转变，提示中年“关键窗口”
Brain aging shows nonlinear transitions, suggesting a midlife "critical window"
这篇 PNAS 论文研究大脑衰老的非线性变化，发现大脑老化并非匀速进行，而是在中年（约 40-60 岁）存在一个关键的"临界窗口期"，此时干预可能有效。
神经科学与心理学2025 · 295 分 · 159 评进阶 · 约 75 分钟编辑精选社交网络中大规模情绪传染的证据（2014）
Evidence of massive-scale emotional contagion through social networks (2014)
这篇 2014 年发表于 PNAS 的论文研究了 Facebook 上约 70 万用户的信息流中，减少正面或负面内容是否会传染性地影响用户自己发布内容的情绪。
神经科学与心理学2020 · 285 分 · 72 评入门 · 约 45 分钟编辑精选裸盖菇素使人类大脑去同步化
Psilocybin desynchronizes the human brain
这篇 Nature 研究用个体化的 fMRI 精密成像追踪受试者服用裸盖菇素（psilocybin）前后的脑功能连接变化，发现药物会造成大范围的脑网络「去同步化」，尤其是默认模式…
神经科学与心理学2024 · 273 分 · 259 评进阶 · 约 60 分钟编辑精选对智力的赞扬会削弱儿童的动机与表现（1998）
Praise for Intelligence Can Undermine Children's Motivation (1998)
这篇 1998 年的经典论文（Dweck 等人）通过实验研究表扬方式对儿童动机的影响：夸孩子“聪明”（表扬智力）会让他们回避挑战、害怕失败，而表扬努力则更能培养坚持性。
神经科学与心理学2016 · 253 分 · 155 评入门 · 约 45 分钟编辑精选