第 7 章 · 22 分钟

Reverse audit: checking the empirical claims of chapters 1–6

The audit covers chapters 1 to 6; the evidence is one person and one firm, self-reported.

概念地图

Reverse audit

Counting the empirical claims of chapters 1 to 6; the evidence is a sample of one.

深色为本章概念，浅色为其他章

陪练模式

不想只是读？让它带你走一遍。

导师把这一章拆成小步，每一步先问你，再讲；你用自己的话答，它先认下对的部分再纠偏。用你自己的 AI，对话只存在这台设备上。

对话只存在本机

Locus01

延伸阅读

Dalio · the opening demand · do not adopt anyone's principles unthinkingly

The book opens by asking the reader to think independently rather than take a ready-made package on trust. It argued against uninspected adoption; the cost is the tension with the two hundred-odd ready-made entries that follow, and that the book gives the reader no method for inspecting them — this chapter supplies that method.

Turn

The 2011 free PDF and the 2017 trade book

Two writings of the same material, with different numbering and wording. Any reading that turns on a phrase must say which version it uses; this course works throughout from the 2011 version, and that is the standard for this chapter's textual test.

Foundational

The evidential structure of self-help and management books · the general case

The evidence in this genre is normally the author's own experience and that of their organisation: a sample of one, written after it succeeded. It argued against leaving experience sealed inside a person; the cost is that survivorship cannot be removed from such material — not because the author is dishonest, but because the material contains no counterexamples.

Counterexample

Chapters 1 to 6 of this course

The object of this audit. Block 3 (derivation) and block 5 (cracks) of each chapter are the entry points; what matters is whether that chapter's empirical claim stands, and how much this course added.

Object of audit

机制02

What definition in disguise looks like here

Chapter 1's 'writing rules is cheaper', chapter 2's 'patching lets it recur' and chapter 6's 'weighting beats averaging' come out of one shape — a close relative of the immunised claim: true by definition and written as a discovery. The reason for flagging them is practical: each comes out of the meaning of one word — 'reusable', 'patch', 'weighting' — and all three are among the most quoted lines. Left unflagged, a reader carries the force of the delivery over onto the content.

机制A proposition true by definition, through being written as a discovery about the world, lends the definition's certainty to the conclusion.

可迁移性测试

Move it to job levels: define 'senior' as five years at this company, then announce that this company's senior engineers are all experienced. It sounds like a judgement and repeats a definition. The isomorphism breaks in that a levelling standard sits in a handbook and can be challenged, while these three definitions hide inside everyday words.

机制03

What this kind of material can carry

Every claim in this book leans on the same kind of material, and this course books it as self-report evidence. What such material can and cannot carry is clear enough: it proposes hypotheses, supplies operable actions and points at mechanisms invisible elsewhere; it supplies no degree of support for anything, because those who used the same practices and failed do not write books, and nobody keeps their records.

机制A record written only by those who succeeded, through lacking failure samples, can propose hypotheses but cannot support them.

可迁移性测试

Move it to a test report that keeps only the passing cases: it tells you which paths have run green, not what the pass rate is. The isomorphism breaks in that the failing cases can still be rerun, while the people who never wrote a book cannot be found.

机制04

How much this course added

In the other courses the derivations mostly rearrange arguments the original already made. Here it is different: this book gives entries and practices and rarely gives arguments, so this course added structure in several places, each marked in its chapter — chapter 1 added the test that a principle must be violable, chapter 3's derivation carries one premise that is ours (the stricter test governs), chapter 4 added the acceptance standard 'write what it predicts', and chapter 5 replaced an unfalsifiable formula with a conditional stand-in. None of that was read out of self-report evidence; the marking is not modesty but an audit convention, so the reader knows which sentences to take up with the author and which with this course.

机制Adding structure, through supplying checkable tests for a practice, makes it inspectable and also adds what the original lacked.

可迁移性测试

Move it to writing a verbal briefing up as a specification: only once written can it be reviewed, and the writing necessarily adds judgements the briefer never made. The isomorphism breaks in that the briefer can be asked to confirm, while here there is nobody to confirm with.

Derivation05

Counting the categories in chapters 1 to 6

This section totals the category of every derivation step in the first six chapters, and reaches one conclusion: this course's observational risk sits in six claims, none of which the original ever tested.

Chapters 1 to 6 have 4 derivation steps each, 24 in all
Counting convention: only the steps inside block 3 of each chapter, not the cracks or debate blocks.
12 steps are tagged [identity] and 6 are tagged [assumption]
Definition unpacked carries no observational risk; an assumption is a premise the author chose, and removing it breaks the argument.
6 steps are tagged [empirical], exactly one per chapter
This is the only class observation can overturn, and where this course actually places its bets.

推导 · 0 / 4

裂缝06

Cracks in this reading

争议地形07

本书主张

This course's position: read the book as a device that can be lifted off it, mark line by line what is the author's and what this course added, and for every claim write what would have to happen for it to be overturned.

另一种看法

Another line holds that the value of a book like this lies in doing what it says, not in being verified; taking it apart into claims and demanding tests turns a manual that changes behaviour into a pile of hypotheses, and in the years spent waiting for tests the people who simply followed it have already benefited.

分歧扎在

The disagreement is rooted in what you mean to take away. For changing behaviour now, dismantling is a drag; for carrying the structure elsewhere and knowing where it fails, not dismantling means carrying nothing. This course wants the second, and that is a choice determined by the goal, not a refutation of the other reading.

什么证据能裁决

What would settle it is transferable output after reading: have one group follow and one group dismantle, then ask each, of a new practice, what it forbids and how it could be overturned. If the dismantling group is markedly more accurate, this course's position holds on its own terms.

No such comparison exists, so this course's position counts only as an untested methodological choice. And one strength of the book has to be granted: it writes a great many practices at a grain fine enough to act on directly, which is more than most of its genre manages and more than a dismantling reading achieves. This course delivers tests, not the capacity to execute.

Falsification08

Turning 'none of the 6 was ever tested' into a testable specification

What is under test is this chapter's own statement: none of the 6 empirical claims of chapters 1 to 6 is given a test with a failure condition in the 2011 original.

Data source: the full text of the 2011 free PDF, located by section title; the 2017 trade book kept in a separate column
The two columns may never be counted together; this course works from the 2011 version throughout.
Sample range: the 6 claims named by step 4 of block 3 in each of chapters 1 to 6
Specified in advance and matched one by one, not picked after reading.
Time window: not applicable; replaced by a search standard — every place in the book where the corresponding claim appears must be located
A textual test has no time window, so completeness of search stands in: the whole book, not a sample of it.
Decision threshold: if 2 or more of the 6 turn out to carry a test that could fail (stating what result would count against the claim), this chapter's statement is weakened
Fixed in advance.
Failure condition: any one of them is found with an explicit failure condition and test design
That counts as this statement being refuted, and it must be rewritten claim by claim.

推导 · 0 / 3

接口09

Which slot it hangs on

挂在哪个槽位The conclusion you already hold is probably 'skim books like this and take whatever you can use' — a judgement about what reading yields. This chapter goes at how that is cashed out.

Does this chapter (a) replace 'take whatever you can use', or (b) constrain it: before adopting a practice, write what it forbids and what result would make you drop it?

慢变量Register two slow variables: how many practices you have adopted from books like this, and for how many you wrote down the condition for dropping them. Look at the second first.

小结10

本章小结

01Of the 24 derivation steps in chapters 1 to 6, 12 are definition unpacked and 6 are assumptions; only 6 empirical claims carry risk.

02Three of the most quoted lines are true by definition and written as discoveries.

03The evidence throughout is one person and one firm, self-reported: it can propose hypotheses and cannot support them.

04This course added structure in chapters 1, 3, 4 and 5, each marked in its own chapter.

05None of the 6 claims carries a failure condition in the original, so none has been tested and none can be refuted.

提取练习 · 合上书，先自己答一遍。

?Is 'risky steps equal total minus identities minus assumptions' (a) the arithmetic of the classification or (b) an empirical claim observation could overturn?

?Is 'weighting by believability beats averaging by headcount' (a) a step true by definition or (b) an independent discovery?

Forced choice11

A · Identify the mechanism

Which of these is a mechanism?

二选一

Forced choice12

D · Judge the interface

Which one hangs on a slot in your existing structure?

二选一

读到这里，把它记为已读。