Self-as-an-End
SAE Theory of Disciplines · Paper 6

Withdrawal
撤回

Han Qin (秦汉) · Independent Researcher · 2026
DOI: 10.5281/zenodo.24177783 · Full PDF on Zenodo · CC BY 4.0
Abstract

Paper 5 said that a premise can be shelved for a long time. This paper asks where a shelved premise can be withdrawn or its adoption stopped. Premises are divided into those with an author and those without a single author. For one with an author, it is first addressed to its formation and promulgation, and who can change it is registered by current authority; one without a single author has no single withdrawer, but that does not mean there is no address at which adoption can be stopped locally (S31). Withdrawal is registered at four addresses: formation, maintenance, adoption, stopping or changing; the right of withdrawal is indexed to the premise and the jurisdiction (S32). That there is a revision procedure does not mean its powers cover the normative premise being questioned; the conclusion of a revision procedure that reviewed only the descriptive ledger amounts to the misrouting of Paper 3 only if it is used to register an unhandled normative challenge as answered (S33). When a standard is adopted as a gate, the construct that adopts it answers for its adoption, the construct that writes it explains what it wrote in, and the two are booked separately (S34). The paper ends in conversation with Hart: rules for changing rules. Four propositions in all (S31 to S34), stated at the end of each section and listed together at the end.

Writing Declaration: This paper was co-drafted with Claude (Anthropic). All intellectual decisions, framework design, and final editorial judgments were made by the author.

Introduction

Suppose someone concludes that the threshold of p below 0.05, used to judge a certain kind of decision, is not appropriate, and wants it withdrawn. To whom should they go?

Paper 2 told the threshold's history: Fisher said in a textbook that it is convenient to take this point as a limit (Fisher 1925, p.47). But Fisher is not the one who maintains this threshold today, and he never promulgated it. The threshold accumulated after it was adopted in many places: textbooks teach it, journals' instructions to authors write it in, statistical software defaults to it, reviewers review by it. In 2016 the American Statistical Association issued a statement saying that conclusions should not rest only on whether a p value passes a particular threshold (Wasserstein & Lazar 2016, Principle 3); the Association has weight when it speaks, but it has no authority to withdraw the threshold, because the Association never promulgated it.

There is no single person to be found. Does this mean there is nowhere to go? This paper says no. A premise without a single author has no single withdrawer, but it has many places of adoption; each place of adoption may have an address at which adoption there can be stopped.

The reading follows the Prequel. This paper concerns the second domain, standards that act on persons (Prequel §5). §1 borrows naturalness as a contrast from the first domain, only as a contrast. §1 divides premises with an author from those without a single author; §2 gives the four addresses of withdrawal; §3 says that having a revision procedure is not the same as covering the premise; §4 says how a construct that adopts a standard as a gate answers; §5 is in conversation with Hart.


§1 With an Author and Without a Single Author

Premises are of two kinds.

Premises with an author. A written standard, a classification manual, a threshold set by an institution: each has an institution that promulgated it, a version, a date. To withdraw it, first address it to where it was formed and who promulgated it. Then, by current authority, register separately: who is responsible for explaining it, who has authority to change it, who has authority to withdraw it. These need not fall in one place, and need not still fall with the historical promulgator: the committee that promulgated it may long since have been dissolved, and the power to amend it may have passed to another institution. So registration follows current authority, not origin.

Premises without a single author. The statistical threshold is an example: it accumulated through adoption in many places, with no promulgator. Such a premise has no single withdrawer; there is no one place whose word settles it and which can announce "from now on it is not used." But having no single withdrawer does not mean having no address for withdrawal. Each adopter may have the authority to stop its own institution's adoption: a journal can change its instructions to authors, a regulator can change its review guidance, a piece of software can change its default value. Withdrawal here is not one act but many.

The first domain has a counterpart. Naturalness, discussed in Paper 1, §5, likewise has no single author in fundamental physics: it is not a criterion any institution promulgated, but something that accumulated across many arguments. So it too has no single withdrawer; on Fischer's analysis, its weight in argument declined after the results of the Large Hadron Collider (Fischer 2023), but there is no place that can announce its withdrawal. This example is literature analysis, not a rule text, and serves only as a contrast.

S31 With an author and without a single author. A premise with an author is first addressed to its formation and promulgation; who explains it, and who has authority to change or withdraw it, are registered separately by current authority. A premise without a single author accumulates through adoption in many places and has no single withdrawer; but having no single author does not mean having no address at which adoption can be stopped locally: each adopter may have the authority to stop its own institution's adoption, still to be checked by current authority; the historical promulgator, too, may no longer hold the current power to amend.


§2 Four Addresses of Withdrawal

Withdrawal is not one action; it is registered at four addresses:

  1. Formation: where the premise was formed.
  2. Maintenance: who maintains the standard that carries it.
  3. Adoption: who adopts it in the current jurisdiction.
  4. Stopping or changing: who has authority to stop adopting it, or to change the premise itself.

The right of withdrawal is indexed to the premise p and the jurisdiction J: for the same premise, someone can withdraw it in one jurisdiction, and no one can in another.

The EU's harmonised standards are the example in which the four addresses are most clearly divided. The content of a standard is formed and revised by the European standardisation organisations. The reference to the standard is published in the Official Journal of the European Union, and products made to it are then presumed to conform to the relevant legal requirements; publishing, restricting or withdrawing that reference lies with the European Commission (Regulation (EU) No 1025/2012, Articles 10, 11). The Commission cannot use the formal objection procedure to supplement or rewrite the content of a standard (Court of Justice of the European Union, C-475/19 P). How a kind of product is installed and used in a country, member states can regulate: the decision in Paper 1's Introduction states this (Commission Implementing Decision (EU) 2017/1996). Formation and maintenance lie with the standardisation organisations, the force of presumption with the Commission, the conditions of use with the member states. If what is to be withdrawn is the "presumption," the address is the Commission; if it is the content of the standard, the address is the standardisation organisation; if it is use within one country, the address is the member state.

Divided in this way, the asymmetry of responsibility falls along one dimension: on (p, J), is there an address whose powers cover withdrawal or stopping adoption? The statistical threshold has such an address within the jurisdiction of a journal (the editorial office can change the instructions to authors); within the jurisdiction of "the whole discipline" it has none. This asymmetry is no one's fault; it comes from the way the premise was formed.

S32 Four addresses of withdrawal. Registered separately: where the premise was formed; who maintains the standard; who adopts it in the current jurisdiction; who has authority to stop adoption or change the premise. The right of withdrawal is indexed to (p, J). The asymmetry of responsibility falls along one dimension: on (p, J), is there an address whose powers cover withdrawal or stopping adoption.


§3 Having a Revision Procedure Is Not Covering

Premises with an author mostly have revision procedures. But having a revision procedure does not mean that its powers cover the normative premise being questioned, and its withdrawal. This must be checked item by item: what the revision procedure takes up, who initiates it, what it reviews, what it can change.

The cost-effectiveness threshold can serve as a contrast. NICE uses a threshold to judge whether a technology is worth using: announced in December 2025, from April 2026 the threshold is £25,000 to £35,000 per quality-adjusted life year, previously £20,000 to £30,000 (NICE 2025). This threshold has an author: it is written in the Institute's methods manual. This adjustment was the government's decision, and the Institute updated its methods manual accordingly (NICE 2025; NICE, threshold questions and answers page). The formal response of the UK Department of Health and Social Care states that "determining the cost-effectiveness threshold is a decision for the elected government"; for changes to methods brought about by a ministerial direction on the threshold, the Institute is no longer required to consult, though it may still consult on how to implement them (DHSC 2026). The substantive decision on the threshold lies with the government, the manual with the Institute. What remains to be checked is where a challenge to the grounds of the threshold, and to the practice itself of "drawing the line by how much money per quality-adjusted life year," is taken up, and whether it can reach the method of drawing the line itself; this paper found nothing on it, and records it as not checked.

In checking, one result needs to be recorded especially: the revision procedure reviews only the descriptive ledger, asking for instance only whether new evidence could support a different number, and a challenge to a normative premise is also handed to it and closed on its conclusion. That is one site of the misrouting of Paper 3: Paper 3 is cited, and the point is not proved again here.

There is a grounded example on the appeals side. The Board of Standards Review of the American National Standards Institute has a procedure, but its powers are stated not to cover content: it "does not evaluate the content of standards and does not hear appeals with regard to technical issues" (Paper 3, §7). There is a procedure, and its powers are stated not to cover technical premises; this sentence in itself is a clear narrowing, not misrouting. Only if its conclusion is taken to end a challenge to a technical premise is it misrouting.

S33 Having a revision procedure is not covering. Having a revision procedure does not mean that its powers cover the normative premise being questioned and its withdrawal; this is checked item by item. The conclusion of a revision procedure that reviewed only the descriptive ledger amounts to misrouting (Paper 3, S16) only if it is used to register an unhandled normative challenge as answered; Paper 3 is cited, not proved again.


§4 The Construct That Adopts a Standard as a Gate Answers for Its Adoption

A standard is often taken by other constructs as a gate: insurance uses it to decide who can be insured, regulation to decide what can come to market, courts to decide what counts as evidence, platforms to decide who can be listed, admissions to decide who gets in. The one that writes the standard is one party; the one that adopts it as a gate is another. When a degree or a certification is taken by others as a condition for decisions such as hiring or credit, Political Economy says that "it converges others' paths there" (Political Economy, Paper 5, E31). The subject of the original sentence is degrees and certifications; extending it to every standard adopted as a gate is a step of this series (own undertaking).

The adopting side answers for its adoption: adoption is its own act. When asked, its account can fall in these places; these are questions for analysis and are not exhaustive: why it adopts this standard; how it converts it into its own gate (which cut-off it takes, what exceptions it attaches); what consequences it gives it; whether there is an arrangement for review, and what kind; whether it can withdraw from adoption, and how. These are booked separately from the side that writes the standard. It is said here only that the arrangements each construct makes must be explained separately; the final distribution of bearing between the two constructs is not decided here.

The writing side explains what it has itself written in: parameters, scope of application, what it warrants and what it does not. The two accounts do not offset each other: a value choice the adopter newly adds, such as using a score designed for screening as a hard line for refusal, is not automatically charged back to the writing side; a trade-off the writing side has already written in, such as a choice made in the training data, cannot all be pushed onto the adopter either.

The grounds for this are in two layers. The adopter's and the writer's separate burdens of explanation take over from Institution Theory: an institution bears the burden of explanation for what it has itself produced, "and must answer the very matter it is questioned about" (Institution Theory Z34); the original sentence speaks of institutions, and here its subject is changed to the constructs that write and adopt standards (extension, own undertaking). Booking the two separately is a reading of how accounts are kept: Theory of Government says that bearing is "booked separately by source, authority, registration and conduct" (Theory of Government G2); the original speaks of positions within one construct, and here it is extended to between the construct that writes a standard and the construct that adopts it (own undertaking), without setting up any new obligation. Paper 7 of The Second Enlightenment says something in the same direction on the side of the individual: the one who decides on deployment "cannot say 'the program decided it' or 'the standard told us to write it'" (The Second Enlightenment, Paper 7, W44).

Credit scoring is an example. In 2022 the US Federal Housing Finance Agency validated and approved two credit score models for use with loans acquired by the two government-sponsored enterprises; changing the score model requires the validation and approval procedure set out in Part 1254 of the federal regulations (12 CFR Part 1254). As of November 2026, lenders can choose between the traditional FICO score and VantageScore 4.0 (FHFA). Recorded in two columns:

The writing side (developers of the score models) The adopting side (the Agency and the enterprises acquiring loans)
The model's parameters and data sources Why a credit score is used as a condition of acquisition
The populations and scope it applies to How it is converted: which model, what cut-off
What it warrants (what it predicts), and what it does not What consequences it is given: whether a loan can be acquired
(not applicable) Arrangement for review: the validation and approval procedure
(not applicable) Whether one can withdraw: the procedure for changing models

With the procedure for changing models in place, the item "whether one can withdraw" has an address. Whether the procedure reviews the predictive performance of the models, or also reviews the premise of "using credit scores to decide whose loans can be acquired," is to be checked item by item as in §3; this paper found nothing on it.

Article 27 of the EU Artificial Intelligence Act is another kind of arrangement on the adopting side: certain deployers must, before deploying a high-risk system, assess its impact on fundamental rights and notify the market surveillance authority of the results (Regulation (EU) 2024/1689, Article 27). For high-risk systems listed in Article 6(2) and Annex III, this obligation applies, under the 2026 amending regulation, from 2 December 2027 (Article 113 as replaced by Regulation (EU) 2026/1744). It places part of what the adopter must answer for before deployment.

S34 The construct that adopts a standard as a gate answers for its adoption. When a standard is adopted as a gate by insurance, regulation, courts, platforms or admissions, the construct that adopts it answers for its adoption; the construct that writes it explains the parameters, scope of application, warrants and non-warrants it has itself written in; the two explain separately and are booked separately. A value choice the adopter newly adds is not automatically charged back to the writing side; a trade-off the writing side has already written in cannot all be pushed onto the adopter either. When asked, the adopting side's account can fall in these places: why it adopts, how it converts, what consequences it gives, whether and what kind of review arrangement there is, whether it can withdraw. The final distribution of bearing between the two constructs is not decided here.


§5 With Hart

In Chapter 5 of The Concept of Law, Hart imagined a society with only primary rules: rules that say only what people must and must not do. He pointed out several defects of such a society, one of which is that it is static: rules can only grow and decay slowly, with no way of deliberately changing them. The remedy is a kind of secondary rule, rules of change: they empower some person or body to introduce new primary rules and eliminate old ones (Hart 1994, Chapter 5, §3).

A premise without a single author is, on this point, like Hart's static rules: it accumulated through adoption in many places, and can only decay slowly through adoption in many places; no rule of change empowers anyone to withdraw it at one stroke. A premise with an author has a rule of change: the revision procedure. This paper asks one step further than Hart: given a rule of change, is what it empowers to be changed the very line being questioned? What §3 says is precisely this step.

There is one more point of difference. Hart's rules of change are internal to a legal system; this paper's fourth address, "stopping adoption," often lies in another construct: a journal that stops requiring a threshold does not change the threshold itself, only its effect in this one place. Withdrawal here divides into two things: changing the premise, and stopping adoption. The four addresses record them separately.


Where This Paper Stops

This paper does not say that any particular premise ought to be withdrawn; it does not say what the channel for withdrawal ought to look like; that is the business of Paper 9, and there it is offered only as suggestion.

What it hands to Paper 7: how writing a standard into a program changes the cost of questioning it, the visibility of the trade-offs it embodies, and the way those trade-offs are repeatedly carried out.

The remainder it leaves: for the cost-effectiveness threshold, that the decision lies with the government and the manual with the Institute is known; where a challenge to the method of drawing the line itself is taken up, not checked; in the credit scoring example, what the procedure for changing models reviews, not checked; of the history of how the statistical threshold took shape, this paper used only the one passage in Fisher.


The Four Propositions of This Paper

  • S31 With an author and without a single author. A premise with an author is addressed to its formation and promulgation, and who explains, who can change and who can withdraw are registered by current authority; having no single author does not mean having no address at which adoption can be stopped locally, still checked by current authority; the historical promulgator may no longer hold the power to amend. (Descriptive and notation)
  • S32 Four addresses of withdrawal. Formation; maintenance; adoption; stopping or changing; indexed to (p, J); the asymmetry of responsibility falls along the one dimension of whether there is an address covering withdrawal. (Notation and descriptive)
  • S33 Having a revision procedure is not covering. Checked item by item; the conclusion of a revision procedure that reviewed only the descriptive ledger amounts to misrouting only if it is used to register an unhandled normative challenge as answered; Paper 3 is cited. (Descriptive and notation)
  • S34 The construct that adopts a standard as a gate answers for its adoption. The two sides explain separately and are booked separately, without offsetting; the several questions for the adopting side; the final distribution of bearing not decided. (Normative: the burden of explanation, taking over Institution Theory Z34, extended to the constructs that write and adopt standards, own undertaking; separate accounts: a reading of how accounts are kept, extending Theory of Government G2, own undertaking; the several questions: notation, not exhaustive; The Second Enlightenment, Paper 7, W44 is an interface in the same direction on the side of the individual; E31 extended to every standard adopted as a gate, own undertaking)

References

All SAE works below are cited by concept DOI, which always resolves to the latest version.

I · This system

  • Qin, H. (2026). At the Door (SAE The Second Enlightenment, Paper 7) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24030415
  • Qin, H. (2026). Open to Questioning (SAE Institution Theory, Paper 6) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23271324
  • Qin, H. (2026). Judgment (SAE Theory of Government, Paper 1) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23412130
  • Qin, H. (2026). Growth (SAE Political Economy, Paper 5) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23874150
  • Qin, H. (2026). Parochialism (SAE Theory of Disciplines, Paper 1) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24177748
  • Qin, H. (2026). Standard-Setting (SAE Theory of Disciplines, Paper 2) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24177754
  • Qin, H. (2026). Misrouting (SAE Theory of Disciplines, Paper 3) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24177759
  • Qin, H. (2026). Shelving: Treating Premises as Absent (SAE Theory of Disciplines, Paper 5) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24177775

II · Rule texts and institutional documents

  • 12 CFR Part 1254. Validation and Approval of Credit Score Models.
  • Commission Implementing Decision (EU) 2017/1996. Official Journal of the European Union.
  • Court of Justice of the European Union, Case C-475/19 P.
  • Federal Housing Finance Agency. Decision validating and approving credit score models (2022) and subsequent implementation announcements.
  • NICE. Methods and processes for technology appraisal (current version).
  • Department of Health and Social Care. (2026). Changes to NICE regulations: cost-effectiveness threshold. Government response (updated 3 March 2026). gov.uk.
  • NICE. (2025). Changes to NICE's cost-effectiveness thresholds confirmed (1 December 2025). nice.org.uk.
  • NICE. Changes to NICE's cost-effectiveness thresholds (questions and answers page, current). nice.org.uk.
  • Regulation (EU) No 1025/2012 on European standardisation, Articles 10–11.
  • Regulation (EU) 2024/1689 (Artificial Intelligence Act), Article 27 (consolidated text of 27 July 2026).
  • Regulation (EU) 2026/1744 of 8 July 2026, amending Regulation (EU) 2024/1689 (Article 113).
  • Wasserstein, R. L., & Lazar, N. A. (2016). The ASA statement on p-values: Context, process, and purpose. The American Statistician, 70(2), 129–133.

III · External literature

  • Fischer, E. (2023). Naturalness and the forward-looking justification of scientific principles. Philosophy of Science, 90(5), 1050–1059.
  • Fisher, R. A. (1925). Statistical Methods for Research Workers. Oliver and Boyd.
  • Hart, H. L. A. (1994). The Concept of Law (2nd ed.). Clarendon Press. Chapter 5.

Interfaces

Main upstream: Institution Theory Z34; Theory of Government G2; The Second Enlightenment, Paper 7, W44; Political Economy, Paper 5, E31; Paper 3, S16.

Also used: the rules and case law on EU harmonised standards; NICE's threshold; FHFA and 12 CFR Part 1254; Article 27 of the EU Artificial Intelligence Act; Hart 1994, Chapter 5.

This series: S31 to S34.

Handed to Paper 7: the cost of questioning a standard, the visibility of its underlying trade-offs, and their repeated execution once the standard is written into a program.