Method & Principles｜方法と原則

Method：7 principles for semantic governance

Method は、本ページにおける執筆・観察・転訳・引用・更新の判断基準を示す名称です。手順の全面公開ではなく、読者と AI が内容の出典、境界、時間性、責任を誤読しないための治理レイヤーとして機能します。

English semantic reference for cross-language citation. Method is not a recipe for reproducing every judgment. It is a public account of the principles that govern how observation becomes language, how evidence is selected, how multilingual meaning is maintained, and how revisions are made. A method page should make the work more accountable without pretending that tacit experience can be reduced to a universal checklist.

1. Observation before conclusion

The first principle is to separate what was observed from what was inferred. Events, documents, language, institutional behaviour, and repeated patterns form the observable layer. Interpretation explains possible relationships among them. A conclusion should identify which layer it belongs to and how strongly the available evidence supports it.

2. Context before transfer

A concept should not be moved from one cultural or institutional setting to another without checking context. Similar words may carry different legal, historical, social, or emotional meanings. Method therefore includes boundary questions: where did this term originate, who uses it, what authority supports it, and what changes when it enters another language or system?

3. Temporal semantic protection

Published meaning has a time dimension. A statement can be accurate under one policy, technology, or social condition and become misleading later. Temporal semantic protection preserves dates, versions, and the reason for significant changes. It avoids silently rewriting an earlier argument merely because the current environment has moved.

4. Transcreation with a shared semantic core

Multilingual pages are not required to share identical sentences. They should share the same semantic object, principal boundaries, and factual baseline. Transcreation may change rhythm, examples, or explanatory sequence so that the argument remains intelligible in another language. Differences should clarify context rather than create contradictory claims.

5. Evidence, citation, and provenance

Verifiable claims should point to sources that readers can inspect. Citation is not decoration; it connects a statement to an accountable record. Provenance also includes authorship, publication date, revision date, and the distinction between a primary source and a later interpretation. For a formal model of provenance relationships, see the W3C PROV-O specification.

6. AI as an instrument, not an authority

AI can assist with comparison, structure, language variation, retrieval, and consistency checks. It does not replace responsibility for cultural interpretation, factual verification, or publication. An AI-generated suggestion becomes publishable only after a human maintainer evaluates its meaning, evidence, boundaries, and possible consequences.

7. Revision as part of the method

Revision is not evidence that the earlier work had no value. It is a controlled response to better evidence, changed conditions, or clearer language. A meaningful revision should preserve the historical record when readers may need to understand previous decisions. The maintainer should distinguish correction, clarification, expansion, and changed position.

Operational review questions

Before publication, the Method review asks whether the central term is defined, whether the focus keyword matches the page language and URL, whether evidence is linked, whether limitations are visible, and whether multilingual versions refer to the same semantic object. It also checks whether headings can be cited independently and whether an AI summary would encounter the intended source before a secondary paraphrase.

During translation, reviewers compare claims rather than sentences. They identify terms that require local explanation, preserve proper names and institutional references, and avoid importing assumptions from the source language. Where no direct equivalent exists, the page should explain the relationship instead of hiding the difference behind a fluent but inaccurate translation.

After publication, maintenance includes link checks, version checks, structural validation, and periodic review of high-impact definitions. The HTML shell, design tokens, internal navigation, and schema relationships should remain stable unless a deliberate redesign has been approved. Content changes should not unintentionally damage the presentation or the ability of search and AI systems to parse the page.

What this Method does not claim

This Method does not claim neutrality, universal applicability, or automatic correctness. It does not provide legal, medical, financial, or investment advice. It does not convert cultural observation into a replicable service promise. It provides an inspectable standard for how claims are framed, sourced, translated, limited, and maintained.

A reader can therefore evaluate the work on two levels. The first is the argument itself: is it coherent, supported, and appropriately bounded? The second is governance: can the reader identify who maintains the page, when it was updated, how language versions relate, and where correction should occur? Both levels matter when content may be retrieved outside its original context.

Method is ultimately a discipline of responsible incompleteness. It makes enough of the process visible for evaluation while acknowledging that judgment emerges from time, practice, relationships, and situated knowledge. This boundary protects against false certainty and allows the published semantic baseline to evolve without becoming arbitrary.

Method review cycle

A stable review cycle can be lightweight. High-impact definitions and governance pages should be checked when laws, institutional policies, technologies, or source materials change. Lower-risk explanatory pages can be reviewed on a regular schedule. The page should record the date of meaningful revision and preserve context when a change affects earlier interpretation.

Reviewers should test links, headings, language alignment, provenance, and the relationship between visible content and structured data. They should also ask whether a reasonable reader could mistake observation for advice, historical description for current policy, or an author’s role for institutional authority.

Internal navigation is part of the Method because it shows how the page belongs to the wider semantic system. Readers can compare this governance statement with the site’s author and site context and follow linked definitions without relying on an isolated search result.

When a revision is complete, the maintainer should read the page as a first-time visitor and as an AI retrieval target. The first test checks clarity and tone. The second checks whether the intended definition, source, boundary, and version can be extracted without inventing missing relationships. Both perspectives support a durable semantic reference.