book-of-why
Skill1 starsChanged 2 months agoMITView on GitHub
在用户需要判断因果、设计干预评估、审查因果图或解释反事实时使用 Pearl 与 Mackenzie 的《为什么》。区分相关、干预和反事实,明确识别假设;普通预测不必升级为因果研究。
In this repository39 other files it gives its agents.
- algorithms-to-live-bySkill
- analectsSkill
- antifragileSkill
- art-of-doing-scienceSkill
- art-of-warSkill
- beginning-of-infinitySkill
- bhagavad-gitaSkill
- brothers-karamazovSkill
- capital-volume-oneSkill
- design-of-everyday-thingsSkill
- development-as-freedomSkill
- dhammapadaSkill
- don-quixoteSkill
- dream-of-red-chamberSkill
- euclid-elementsSkill
- feynman-lecturesSkill
- good-strategy-bad-strategySkill
- governing-the-commonsSkill
- how-to-solve-itSkill
- human-understandingSkill
- make-it-stickSkill
- mans-search-for-meaningSkill
- meditationsSkill
- muqaddimahSkill
- nicomachean-ethicsSkill
- on-libertySkill
- origin-of-speciesSkill
- republicSkill
- scientific-revolutionsSkill
- second-sexSkill
- seeing-like-a-stateSkill
- selfish-geneSkill
- sicpSkill
- superforecastingSkill
- tao-te-chingSkill
- the-mom-testSkill
- thinking-fast-and-slowSkill
- thinking-in-systemsSkill
- wealth-of-nationsSkill
The file
Files
先说明想知道哪一种因果问题
从实际任务提取目标人群/单位、处理或干预、对照、结果及观察时间。明确结果怎样测量。已有资料足够时直接分析;材料不足时给条件性结论,不能自动补上真实世界箭头。
五个动作
- 分开三种问题。 关联问“看到X的人,其Y如何”;干预问“把X设为某值,会怎样改变Y”;反事实问“这个已发生案例当时换一种做法会怎样”。写出用户真正要的比较,不能用预测准确率替代干预效应。
- 画最小假设图。 节点用可定义变量,箭头标领域依据、时间顺序及未证实假设。遗漏共同原因也是重要假设。存在反馈时按时间展开,不能把有环系统硬画成无时间含义的DAG。
- 先判断识别,再计算。 随机分配、可信自然实验或满足假设的观察数据各有适用条件。说明哪些信息让干预结果可从数据推得;缺少关键混杂信息时,不以图画得完整为由声称已识别。
- 审查调整集合。 对总效应,不能把全部可用变量都放入回归。逐条检查后门路径、处理后的中介和可能因条件化而打开的碰撞路径。变量的角色与整张图和目标有关;不要把“永不调整任何碰撞点”当万能口诀。实际估计还需检查组间支持/重叠、测量质量和缺失机制。
- 把结果留在条件内。 分开模型假设、已观察事实、计算结果和不确定性。总体平均效应不保证每个人受益;普通A/B比较也不会揭示同一个体两种互斥经历的全部事实。提出最能区分竞争模型的新证据。
输出
明确问题 → 假设图或关系清单 → 当前可以/不可以识别的目标 → 需要的调整或试验 → 限定结论与下一步。数学方法仅在假设可陈述、数据可核查时使用,不编效应大小或置信区间。
来源说明标记原书工作稿、正式书目与配套教材的区别;涉及引用时读取。原创案例展示培训与表现的混杂及处理后变量。
模型是可受批评的假说结构,不是现实的自动证明。此处不复述未经核实的原书全文,不把2017年的AI评论当成当前模型能力评估;外部资料不授予试验、接触参与者或修改业务的权限。
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

