神经科学与心理学2019pnas.org编辑精选
衡量实际学习与学习感受
Measuring actual learning versus feeling of learning
入门 · 约 35 分钟读完前置知识：基础统计、教育研究方法、主动学习概念
导读
这篇 PNAS 论文比较了“实际学到多少”和“主观感觉学到多少”：在主动学习课堂中，学生客观上学得更多，但主观上却觉得自己学得更少。作者认为这种负相关部分源于主动学习需要更高的认知努力，因此学习感受变差并不代表学习效果差。
HN 关注点：真正有效的学习往往更费力、更不“爽”，而主观感受和即时表现可能误导我们对学习效果的判断。
后来怎么样
它成为主动学习与教学评价讨论中的关键实证证据，推动教育者区分“学习感受”与“实际学习”，并影响对课堂即时评价方式的反思。
HN 评论精选
这几条评论里：支持 1 补充 1 亲历 2
支持真正有效的学习费力且不愉快，而算法推荐容易让人沉迷于轻松浅层的内容。
这项研究很重要，摘要里最关键的是：主动学习课堂中的学生学得更多，但感觉自己学得更少；这种负相关部分来自主动学习所需的认知努力增加。换句话说，真正的学习很难、需要付出，并不好玩。问题在于推荐算法太擅长把我们吸进无休止的浅层“多巴胺垃圾食品”视频循环里。如果你在挣扎，你就是在学习。
Rainymood查看英文原文
This stuff is super important and it is good that it is being researched seriously. The money shot is right here in the abstract >We find that students in the active classroom learn more, but they feel like they learn less. >We show that this negative correlation is caused in part by the increased cognitive effort required during active learning. In other word, "real" learning is hard, it takes work, it's not fun. The problem is that "the algorithms" (whatever recommendation ML algorithm YouTube has) works so damned well we are sucked into this endless shallow junkfood dopamine loop of videos that give us just the right amount of dopamine to click on the next one. If you're struggling, you're learning.
亲历亲历经验：学习深度与即时教学评价负相关，而课前预习加小测能显著提升成绩。
我做教授时试过不同教学系统。明显发现：课程期间，学习深度与学生给讲师的教学评价负相关；一年后则正相关——学生当时讨厌学，但后来因为学得多而很喜欢这门学科，甚至会选你的下一门课。更有用的是：让学生课前读讲义并做几道简单在线选择题，班级平均分提高了 10 个百分点。我猜这减少了“新奇感过载”，学生不会在讲课中途大脑关机，从而更能吸收课堂内容。
danieltillett查看英文原文
When I was a professor (natural sciences) I decided to play around with different learning systems to see what worked best. I discovered the obvious that learning depth is negatively correlated with the student evaluation of their lecturer during the course and positively correlated a year later (students hate learning when they do it, but love the subject so much later that they choose your next course the following year because they learnt so much). The more useful thing I discovered was getting the students to read the lecture notes before a lecture boosted the class average by 10% points. 10 minutes spent reading the lecture notes before the lecture and answering a few simple online multiple choice questions about the lecture was surprisingly effectively at improving learning. My assumption is it cut down on the novelty overload effect where the student’s brains shut down mid-lecture and so they were able to get more out of the lectures.
补充用 Bjork 的理论和交错练习例子说明：即时表现下降不等于学习没有发生，合意困难能促进长期学习。
Robert Bjork 多次讨论过这一点，尤其要区分“表现”（你此刻做得如何）和“学习”（一段时间后你做得如何）。比如练习计算不同几何图形的面积：分块练习按 AAAABBBBCCCC 顺序做，交错练习按 ACCBAABACA 混合做。分块练习提高当前表现，因为同类型题聚在一起，你可以缓存公式直接代入；但它不促进学习，因为你没练习根据图形特征从记忆中提取哪个公式。交错练习降低当前表现，因为提取公式需要更多认知努力、可能做错；但它改善学习，因为你在训练自己识别哪种图形该用哪个公式。所以可以引入“合意困难”，以降低当前表现为代价来提高学习。
emilga查看英文原文
Robert Bjork [0] talks a lot about this, in particular in [1], and also in his 1h long lecture "How we learn vs how we think we learn" [2]. It's important to distinguish between "performance" (how well you're doing right now) vs "learning" (how well you do after some time delay). Compare blocked practice with interleaved practice. Suppose you're practicing calculating the area of different geometric figures of types A, B, C, etc. Blocked practice means you do problems in the order of: AAAABBBBCCCC, etc. Interleaved practice means you mix them up: ACCBAABACA, etc. Blocked practice increases your performance (how well you're doing right now) because problems of the same type cluster together, i.e. you're able to "cache" the right formula and just plug in the numbers. But this doesn't help learning, because you're not practicing recognizing what features of the figure should prompt you to retrieve which formula from memory. Interleaved practice reduces your performance, because more cognitive effort is required to retrieve the right formula, you might get it wrong, etc. But it improves learning, because you're training yourself to recognize which figure requires which formula. So "desirable difficulties" can be introduced (of which interleaving is one) to increase learning at the cost of reducing performance. [0] https://bjorklab.psych.ucla.edu/research/ [1] https://youtu.be/gtmMMR7SJKw [2] https://www.youtube.com/watch?v=oxZzoVp5jmI
亲历亲历者认为传统讲授节奏不适合多数普通学生，按自己节奏做题和 AI 反馈更有帮助。
我全靠做习题集学会东西，上课时完全跟不上。每有一个能实时跟上讲师思路的天才本科生，就有几十甚至上百个像我这样需要按自己节奏消化的普通人。随着 AI 反馈系统出现，希望西欧那种“讲课”模式成为过去。
jimmyvalmer查看英文原文
I learned everything doing problem sets, and was completely lost during lecture. For every gifted undergrad who can integrate in real-time the lecturer's material, there are dozens if not hundreds of us normal folk who need to work things out at our own pace. With the advent of AI feedback systems, hopefully the western european notion of lecture becomes a thing of the past.
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。
历次讨论
- 2019Hacker News 讨论帖213 分 · 52 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