01 / THE INVITATION · 不必先同意

# 保持好奇， 允许修正。 / Stay curious. Stay revisable.

世界不缺观点。稀缺的是： 一个让你愿意重新想一遍的地方。 / The world has enough opinions. What’s rarer is a place that makes you want to think again.

Astral Codex Ten 把科学、心理、经济和 AI 放在同一张书桌上。读它，不是为了收集正确答案，而是为了看见：一个答案，究竟是怎么想出来的。 / Astral Codex Ten puts science, psychology, economics and AI on the same desk. Read it not to collect the right answers, but to see how an answer is built.

[挑一篇，试试看 / Find your first essay↗](#read)[先动手试一下 ↓ / Try a thought experiment ↓](#experiment)

无需数学背景。无需认同作者。带着好奇就够了。 / No math background. No oath of agreement. Curiosity is enough.

FIELD NOTES / 随身问题

01 BEFORE BELIEF

## 我为什么相信？ / Why do I believe this?

是证据，还是熟悉感？ / Evidence—or familiarity?

a good place to begin.

02 AFTER EVIDENCE

## 什么会让我改主意？ / What would change my mind?

先写下来，再看结果。 / Write it down before the result.

?

把观点握松一点。 把证据看仔细一点。 / Hold beliefs lightly. Look at evidence closely.

A SHORT INTRODUCTION

## 先认识一下。 / First, a little context.

ACX 是什么？ Astral Codex Ten 是 Scott Alexander 的博客，是 Slate Star Codex（SSC）的延续。Scott 是医生，也是一位把严肃推理写得很好看的作者。[[1]](https://www.astralcodexten.com/about) / What is ACX? Astral Codex Ten is Scott Alexander’s blog, the successor to Slate Star Codex (SSC). Scott is a doctor and a writer who makes serious reasoning a pleasure to read. [[1]](https://www.astralcodexten.com/about)

这里说的「理性」，不是没有感情。 它是一种练习：用概率、证据和认知科学，分清自己知道什么、猜测什么，以及可能错在哪里。ACX 是进入这片思想世界的一扇门，不是它的全部。 / Rationality here does not mean having no feelings. It means practicing with probability, evidence and cognitive science to separate what you know from what you guess—and where you might be wrong. ACX is one doorway into that world, not the whole of it.

02 / WHY BOTHER?

## 不只更会争论。 而是更值得思考。 / Not better arguments. Better things to think about.

读完以后，某些老问题 会突然长出新的入口。 / Old questions start to reveal unexpected new entrances.

01 / DISCERNMENT

### 少被一个好故事说服。 / Be harder to fool with a good story.

「研究证明」不等于尘埃落定。样本是谁？效果多大？相反的结果呢？你会开始看整片证据，而不是只看最适合转发的那一条。 / “A study shows” is not the end of a question. Who was studied? How big was the effect? What did other studies find? Look at the landscape, not just the most shareable result.

用在：健康新闻、产品宣传、投资故事。 / Useful for: health headlines, product claims, investment stories.

02 / EXPLANATION

### 看见坏结果背后的好理由。 / See how sensible choices go wrong together.

人人都讨厌内卷，为什么它还在加速？有时不需要一个坏人：只要规则奖励某种行为，局部合理就能拼成整体荒唐。 / Why does a race nobody enjoys keep accelerating? Sometimes no villain is needed: incentives can turn individually sensible choices into a collectively miserable outcome.

用在：职场内卷、平台竞争、组织设计。 / Useful for: workplace races, platform competition, institutional design.

03 / WONDER

### 重新找回「原来还能这么想」。 / Rediscover “I’d never thought of it that way.”

一篇文章能从精神医学走到经济学，再拐进哲学。不是为了炫耀知识，而是让彼此遥远的现象，突然照亮对方。 / An essay can travel from psychiatry to economics and turn into philosophy. The attraction is not collecting facts; it is watching distant ideas illuminate each other.

用在：你还不想急着下结论的任何时候。 / Useful whenever you would rather explore than rush to a verdict.

03 / A LITTLE MIND-CHANGING

## 新证据来了。 你该改多少主意？ / New evidence. How much should you update?

桌上有 A、B 两种袋子。A 袋有 80% 红球，B 袋有 20% 红球。按滑块设定的概率选 A，否则选 B，再从袋中随机抽一球。现在，你看见了一颗红球。 / There are two kinds of bag. A contains 80% red balls; B contains 20%. Select A with the probability set by the slider, otherwise select B. Draw a random ball from that bag. You see a red ball.

它支持「选中的是 A」，却不保证就是 A。关键还在于：抽球之前，选中 A 的概率有多大？ / That supports “the bag is A,” but does not guarantee it. The missing question: how likely was A before you saw the ball?

本站设计的贝叶斯玩具模型，非 ACX 原文。数值为假设，不是现实预测。 / An original Bayesian toy model, not an ACX excerpt. These are assumed values, not a real-world forecast.

THE UPDATE DESKINTERACTIVE看见红球前，选中 A 的概率 / Chance of A before seeing red

20%PRIOR / 先验

1%99%

看见红球后，A 的概率 / Chance of A after seeing red50.0%↗

同一颗红球，不同的起点，带来不同的结论。证据不是开关，而是调节旋钮。 / The same red ball leads to different conclusions from different starting points. Evidence is a dial, not a switch.

看看怎么算 / Show the calculation

(0.20 × 0.80) / (0.20 × 0.80 + 0.80 × 0.20) = 50.0%

红球在 A 中出现的概率，是在 B 中的 4 倍；赔率因此乘以 4，而不是概率乘以 4。 / Red is four times as likely under A as under B. Multiply the odds by four—not the probability.

04 / YOUR FIRST THREAD

## 从一个真正在意的问题开始。 / Start with a question you actually care about.

不用从头补课。下面三篇是 ACX 官方介绍推荐的 SSC 经典。中文标题与导读由本站编写，链接通往英文原文。 / No syllabus required. These three SSC classics are recommended on ACX’s own About page. Chinese titles and reading notes are our adaptations; links lead to the English originals.

全部入口 / All paths我想判断得更准 / Better judgment我想看懂内卷 / Understand the race我厌倦了抠字眼 / Beyond word games

3 essays / 3 篇文章

EVIDENCE / 2014

START HERE · 01

### 小心那个 只拿一项研究说事的人。 / Beware the Man of One Study

有证据，不等于证据站在你这边。挑选研究，能把同一个领域讲成两个完全相反的故事。 / Having evidence is not the same as having the weight of evidence. Selective studies can tell opposing stories about the same field.

带走一个问题： 被省略的研究说了什么？ / Take one question with you: What did the omitted studies find?[读英文原文 / Read the original↗](https://slatestarcodex.com/2014/12/12/beware-the-man-of-one-study/)

INCENTIVES / 2014

GO DEEPER · 02

### 关于摩洛克的沉思： 没有坏人的坏结局。 / Meditations on Moloch

如果所有人都不喜欢某种竞赛，为什么没人停下来？从军备竞赛到污染，看协调失败如何吞掉共同利益。 / If nobody likes a race, why can nobody stop? From arms races to pollution, explore how coordination failures consume shared interests.

带走一个问题： 谁能不为退出付出代价？ / Take one question with you: Who can afford to stop competing?[读英文原文 / Read the original↗](https://slatestarcodex.com/2014/07/30/meditations-on-moloch/)

CATEGORIES / 2014

CHANGE THE FRAME · 03

### 分类为人服务， 不是人为分类服务。 / The Categories Were Made for Man

我们是在争事实，还是在争词语的边界？用「鲸算不算鱼」打开一个问题：分类究竟要帮我们做什么？ / Are we disagreeing about facts or the boundaries of words? “Is a whale a fish?” opens a better question: what is this category for?

带走一个问题： 这个定义帮我们解决什么？ / Take one question with you: What problem does this definition solve?[读英文原文 / Read the original↗](https://slatestarcodex.com/2014/11/21/the-categories-were-made-for-man-not-man-for-the-categories/)

想看今天的讨论？读 ACX。想学更系统的方法？逛 LessWrong。两者有关联，但不是同一个网站，也不代表同一种立场。 / For ongoing essays, read ACX. For a broader library of rationalist methods, explore LessWrong. They overlap, but are not the same site or a single shared position.

[Astral Codex Ten ↗](https://www.astralcodexten.com/)[LessWrong ↗](https://www.lesswrong.com/)

05 / KEEP YOUR OWN MIND

## 借它的工具。 不必借它的身份。 / Borrow the tools. Keep your own mind.

最好的理由，不是「这群人更聪明」。而是：这些方法能不能帮你发现自己的错误。 / The best reason is not “these people are smarter.” It is whether their methods help you notice your own mistakes.

这是不是一种优越感？ / Is this just intellectual superiority?

不该是。会说「贝叶斯」不代表不会自欺。把「我是理性的人」当身份，恰恰可能让人更难接受反证。方法要靠结果检验，不靠圈内认可。 / It should not be. Knowing Bayesian vocabulary does not prevent self-deception. Identifying as “a rational person” can make counterevidence harder to hear. Test methods by results, not by in-group approval.

我需要同意作者的政治或 AI 观点吗？ / Must I agree with the author on politics or AI?

不需要。这里有值得争论的观点，也有可能过时的判断。把文章当作可拆解的论证，不是权威结论。读他的公开纠错页，也读有力的反对意见。 / No. Some claims are contested; some judgments may age badly. Treat an essay as an argument to inspect, not an authoritative verdict. Read the public corrections page and strong opposing views.

[ACX 的公开纠错页 / ACX’s public corrections ↗](https://www.astralcodexten.com/p/mistakes)

每件事都要算概率，会不会太累？ / Does everything need a probability?

不用。精力有限，判断也有成本。先把这些工具用在代价高、容易冲动、值得反复做的决策上。情感告诉你在乎什么；推理帮你看清怎样更接近它。 / No. Attention is limited, and analysis has a cost. Start with decisions that are costly, emotionally charged or recurring. Feelings tell you what matters; reasoning helps you work toward it.

不付费也能开始吗？ / Can I start without paying?

可以。上面的三篇经典可免费阅读。ACX 的介绍页也说明，重要内容免费；部分额外内容需要付费订阅。本站不出售会员，不代收订阅费。 / Yes. The three classics above are free to read. ACX’s About page says important content is free, with some extras for paid subscribers. This site sells no membership and collects no subscription payments.

AN INVITATION, NOT A CONVERSION

## 不必换一个阵营。 只需多问一个好问题。 / You don’t need a new tribe. You need one better question.

挑一篇。找出最不服的一点。 然后问自己：什么证据，会让我改主意？ / Pick an essay. Find the part you resist. Then ask: what evidence would change my mind?

[从《小心一项研究》开始 / Start with “One Study”↗](https://slatestarcodex.com/2014/12/12/beware-the-man-of-one-study/)