STD-R — The Research Standard

Subordinate to Codex 1 and the Codices named in §1. See 00-index.md for authority, normative vocabulary, citation format, and waiver rules, which this Standard does not restate.

1. Purpose

This Standard governs how Anabasis produces and publishes knowledge: what may be cited as evidence, how strongly a claim may be stated, what must be disclosed before publication, how member data may and may not be used, how a result is reproduced, reviewed, corrected, retracted, and archived. It converts Codex 0 Chapter 16's measurement doctrine and Codex 1 Article IV into obligations a reviewer who did not do the research can check.

Codex 1 Art. IV / Codex 0 ch.16  ─────►  what research must be honest about
STD-R (this file)               ─────►  what a reviewer checks before publication
Editorial review record         ─────►  the evidence that the checks were done
Published finding                ─────►  carries its lineage and its correction history forever

2. Scope and non-scope

In scope. Every publication issued under the Research Institute (DIV-03): journal articles, indexes, working papers, capability and outcome studies, and any internal finding cited externally or used to justify an institutional metric (Codex 0 §16.3). Source classification, claim-strength language, citation format, methodology disclosure, anonymization and consent for member data used in research, reproducibility, peer and editorial review, pre-registration, correction and retraction, definitional stability for recurring measures, AI-assistance and conflict-of-interest disclosure, and archival obligations.

Not in scope. The institutional metric set's own definitions and thresholds (Codex 0 Chapter 16, which this Standard's §11 inherits rather than restates); the annual report's mandatory contents (Codex 0 §16.5); dashboard presentation of a figure (STD-X, Codex 0 Chapter 12); the Research Institute's internal workflow stages (Codex 0 Chapter 09, WF-03).

3. Source classes and weight

3.1 Every source cited in a publication is classified into exactly one class, stated adjacent to the citation or in an accompanying methodology note:

3.2 A reviewer checks that every cited source in a publication carries its class, visibly, either inline or in a methodology appendix; an unclassified source is an incomplete citation, not a minor omission.

3.3 A claim resting on a lower-weight source class must not be presented with the same confidence as a claim resting on a higher one (§4); mixing classes to imply a stronger evidentiary base than any single source supports is a defect.

4. Claim strength tied to evidence strength

4.1 Every claim of fact in a publication is expressed at one of four strength levels, and the level used must match the evidence class and sample supporting it:

4.2 A reviewer checks each factual sentence in a publication against this table: does the verb and hedge used match the evidence actually cited two sentences prior or in the methodology section. A mismatch — an "established" claim resting on preliminary-tier evidence — is a blocking defect, corrected before publication, not softened after the fact by a footnote.

4.3 A claim must never be strengthened between draft and publication without new evidence added to support the stronger language; a reviewer checks the editorial review record (§9) for the evidence justifying any strength escalation.

4.4 No claim may outrun its stated sample size or population; a reviewer checks that any claim quantified by "most," "many," or a percentage is traceable to a stated N and population in the same publication.

5. Citation format and traceability

5.1 Every claim of fact carries a citation a reader can reach: a stable link to the source (an external DOI or URL, or an internal reference to a published methodology note or dataset description), never a citation resolvable only to internal, unpublished material a reader cannot access.

5.2 Where the underlying data is internal and cannot be published in full for privacy reasons (§6), the citation points to a published methodology note describing the data's shape, collection method, and aggregation, satisfying traceability without disclosing individual records.

5.3 A reviewer checks that every factual sentence in a publication resolves, by following its citation, to a source that actually supports the specific claim made — not merely a source on the general topic. A citation that supports a weaker or different claim than the one it is attached to is a defect (a "citation mismatch").

5.4 Citation format is consistent within one publication (author-date, numbered, or footnote, stated once at the top of the document) and stable across the Research Institute's output so that cross-referencing between publications is mechanical, not guesswork.

6. Anonymized-data-only rule and aggregation thresholds

6.1 Research conducted on members, scholars, founders, or clients uses anonymized or de-identified data only (Codex 1 Article VI.2 restated for research use; Codex 0 §04.4.8). Research code paths and datasets must not carry a direct identifier (name, email, member ID resolvable to a person) once a study begins; a reviewer checks the described dataset for the presence of identifying fields and treats any as a blocking defect.

6.2 Any published figure derived from member data is subject to the institutional aggregation threshold (Codex 0 §16.7) recursively across breakouts; a reviewer checks that no combination of published cross-tabulations in the same work permits re-identification of an individual by intersection, even where each individual breakout clears the threshold alone.

6.3 A cohort smaller than the threshold is reported qualitatively or merged with an adjacent cohort, never disclosed as a small-N figure (Codex 0 §16.8.3, restated for research publications specifically).

6.4 Re-identification risk is assessed and stated in the methodology disclosure (§8) for any dataset involving fewer than a stated multiple of the aggregation threshold's population, even where the headline figures individually clear it.

7. Consent for member data used in research

7.1 Member data enters a research study only where the member has consented to research use specifically, distinct from consent to receive the institution's services (Codex 0 §04.7.5, restated). Consent bundled into an unrelated agreement is void per Codex 6's prohibition on bundled consent.

7.2 A reviewer checks the study's data-source description states which consent record authorises the data's use, and confirms the consent record's scope (research use) matches the study's actual use, not merely the member's general terms of service.

7.3 A member may withdraw research consent; a reviewer checks the study's data pipeline states how a withdrawal is honoured for already-collected data (removal, irreversible anonymization prior to any withdrawal mechanism, or a stated reason removal from an already-published aggregate is not possible).

7.4 No data sale, rental, or brokerage occurs in connection with any research relationship (Codex 1 Article VI.2, absolute); a reviewer checks that any external funding or data-sharing agreement tied to a study does not transfer member-level data to the funder or partner.

8. Methodology disclosure before publication

8.1 Every publication carries, before or alongside its findings, a methodology section stating: the question asked; the data used, its source class (§3), and its size; the method applied (statistical technique, qualitative approach, or model); the limitations of the method and data, stated plainly, not minimised; and any conflict of interest (§13) held by an author, editor, or funder.

8.2 A reviewer checks each of the five elements in §8.1 is present, specific, and not a placeholder sentence ("this study has some limitations" is not a disclosed limitation; naming the limitation is).

8.3 A publication with an incomplete methodology section does not proceed to editorial review (§9); the section is a precondition of review, not a formality attached afterward.

9. Peer and editorial review, and reviewer independence

9.1 Every publication passes editorial review by at least one reviewer who did not author or directly supervise the work, before publication (Codex 0 §16.4.3 "editorial independence incidents," restated as a positive obligation here).

9.2 A publication with a material claim (Established or Supported tier, §4) additionally passes review by at least one subject-matter reviewer external to the authoring team, or, where no qualified internal reviewer exists independent of the work, an external peer reviewer.

9.3 A reviewer's independence is checked by disclosure: the reviewer states, in the review record, any prior collaboration with the authors, any funding relationship with the study's sponsor, and any reporting relationship to the authors. A reviewer with an undisclosed conflict is not independent, and the review does not count toward §9.1–9.2.

9.4 The editorial review record is retained alongside the publication: reviewer identity (or role, where anonymity is preserved), date, findings challenged or requested changed, and the disposition of each. A publication with no retrievable review record is treated as unreviewed regardless of any claim otherwise.

9.5 Editorial independence is a named institutional metric (Codex 0 §16.4.3); a confirmed instance of sponsor influence over a finding is logged as an editorial independence incident under that metric, not handled informally.

10. Pre-registration for outcome studies

10.1 A study measuring outcome realization (Codex 0 §16.3, "Outcome realization rate," "Longitudinal capability retention," or any study designed to test whether a program produced a claimed effect) is pre-registered before data collection begins: the hypothesis, the population, the outcome measure, and the analysis method are recorded and timestamped prior to seeing outcome data.

10.2 A reviewer checks the publication states its pre-registration reference and date, and checks that the published analysis matches the pre-registered plan; any deviation (a changed outcome measure, an added subgroup analysis) is disclosed explicitly as exploratory, not folded silently into the confirmatory result.

10.3 An outcome study published without pre-registration is labelled exploratory in its title or abstract and its claims are capped at "Preliminary" strength (§4) regardless of sample size, until replicated with a pre-registered design.

11. Definitional stability for indexes and recurring measures

11.1 A recurring measure or index (Codex 0 §16.3's institutional metric set, or a Research Institute index published on a cadence) uses a fixed definition, formula, and data source once published, changed only through a decision record naming the old definition, the new definition, the reason, and the effective date (Codex 0 §16.8–§16.9, inherited here for any index the Research Institute maintains independent of the annual report).

11.2 A reviewer checks that a changed index shows the old-definition and new-definition figures side by side for the year of transition (Codex 0 §16.9.3), and that historical figures are never silently restated (§16.9.4).

11.3 Index methodology revisions and their materiality are themselves counted and disclosed annually (Codex 0 §16.4.3, "Index methodology stability"); a reviewer checks the Research Institute's own annual disclosure names every revision made that year.

12. Reproducibility

12.1 Every publication with a quantitative or empirical finding preserves, for the life of the institution: the dataset or a description sufficient to reconstruct it under §6's privacy constraints; the exact analysis code or method steps; the software or tool versions used where version affects the result; and the publication's own stated limitations.

12.2 A reviewer checks that a named person or system could, from the preserved materials alone, re-derive the headline figure without contacting the original author; where full data cannot be shared for privacy reasons, the preserved methodology note (§5.2) must be sufficient for an independent party to reproduce the method on an equivalent dataset.

12.3 "Published finding reproducibility" is itself a named institutional metric (Codex 0 §16.4.3); a reviewer checks that a sample of the year's publications was actually tested for reproducibility, not merely asserted reproducible.

13. AI assistance and conflict-of-interest disclosure

13.1 Any use of an AI system in drafting, analysing, or summarising a publication is disclosed in the methodology section: what the system did (drafting prose, running an analysis, generating a summary), what a human verified, and what a human decided (Codex 1 Article VI.3–4, restated for research). Human oversight of the finding itself is never delegated to the system.

13.2 A reviewer checks the AI-assistance disclosure is specific enough to know which parts of the work a person is answerable for; "AI was used" without specifying the task is an incomplete disclosure.

13.3 Every author, editor, and funder conflict of interest — financial interest in the finding's outcome, a reporting or personal relationship to a study's subject, a funding source with a stake in the result — is disclosed in the methodology section, regardless of whether it is judged to have influenced the outcome; a reviewer checks the disclosure names the relationship and its nature, not merely states "no conflicts" without confirming that was checked against a specific list (funding source, employment, equity, personal relationship).

14. The unflattering-result obligation

14.1 A result is not withheld, delayed, or reworded because it reflects poorly on the institution, a division, or a program (Codex 0 §16.2.3, restated for research specifically as a publication obligation, not only a reporting one).

14.2 A reviewer checks, where a study's design was pre-registered or its methodology committed in advance (§10, §8), that the actual published conclusion matches what the pre-committed method would produce, regardless of whether that conclusion is favourable; a conclusion softened relative to the pre-committed analysis is a blocking defect.

14.3 The Research Institute's editorial review record states explicitly, for any study whose result is unfavourable to the institution, that the result was published as found; a pattern of unfavourable studies not reaching publication is itself examined as a defect in the editorial process (Codex 10, Institutional Critic).

15. Correction, erratum, and retraction

15.1 A factual error discovered after publication is corrected via one of three instruments, chosen by severity: a correction (a minor factual fix that does not affect the finding), an erratum (a fix to a material element — a figure, a citation, a methodology detail — that does affect interpretation but not the core finding), or a retraction (the finding itself is unsupported and withdrawn).

15.2 Every correction, erratum, or retraction is published visibly attached to the original work — never as a silent edit — dated, and stating what was wrong, what is now correct, and how the error was found.

15.3 A retraction is issued within 30 days of the error being confirmed; a correction or erratum within 14 days. A reviewer checks the dated gap between confirmation (recorded in the editorial record) and public posting against these limits.

15.4 A retracted or corrected finding that was cited in an institutional metric (Codex 0 §16.3) or the annual report triggers a review of that downstream figure under Codex 0 §16.9's restatement discipline; a reviewer checks that this handoff occurred where applicable.

15.5 The public-facing version of the work (index page, listing, citation record) links to the correction, erratum, or retraction notice permanently; it is never possible to reach the original publication without also reaching its correction history.

16. Archival and permanence

16.1 A published finding, its methodology note, its review record, and any correction or retraction are retained for the life of the institution (Codex 1 Article VII, hundred-year horizon), in a form that resolves at a stable, citable URL that does not change on a content or design refresh (Codex 0 §11.4.3, slugs restated for research).

16.2 A reviewer checks that a publication's canonical URL is stable and that any prior URL, if changed, redirects permanently rather than breaking external citations, including citations from credential verification records or third-party work.

16.3 Archival includes the dataset description and analysis method (§12.1) in a durable store, not only the rendered publication page, so a re-derivation remains possible independent of the website's own uptime.

17. Governing Codex clauses

Codex 0 Chapter 16 (measurement philosophy, honesty requirement, aggregation thresholds, restatement discipline), §16.4.3 (DIV-03 metrics: reproducibility, index stability, editorial independence). Codex 1 Article IV (institutional test — knowledge and trust), Article V (the member is not the product), Article VI.2 (no data sale), VI.3–4 (human oversight and AI explainability), VI.6 (auditability), VI.10 (doctrine amended in writing). Codex 0 §04.4.8 (research reads anonymized data only).

18. Conformance checklist

• [ ] Every cited source carries a stated class (§3) and no claim rests on a lower-weight source than its stated strength implies. — yes / no / not-applicable-because
• [ ] Every factual claim's language (established / supported / preliminary / speculative) matches the tier of evidence actually cited (§4). — yes / no / not-applicable-because
• [ ] Every claim of fact resolves, via its citation, to a source a reader can reach that actually supports that specific claim (§5). — yes / no / not-applicable-because
• [ ] The methodology section states question, data, method, and limitations in specific, non-placeholder language, before publication (§8). — yes / no / not-applicable-because
• [ ] Member data used is anonymized, consented specifically for research, and every published breakout clears the aggregation threshold recursively (§6, §7). — yes / no / not-applicable-because
• [ ] The dataset description, analysis method, and tool versions are preserved so the finding could be re-derived independently (§12). — yes / no / not-applicable-because
• [ ] Editorial review is complete, the reviewer is independent and discloses any relationship, and the review record is retrievable (§9). — yes / no / not-applicable-because
• [ ] Any outcome study is pre-registered, or is labelled exploratory and capped at preliminary strength (§10). — yes / no / not-applicable-because
• [ ] Any changed index or recurring measure shows old and new figures side by side for its transition year (§11). — yes / no / not-applicable-because
• [ ] AI assistance and every author/editor/funder conflict of interest are disclosed specifically, not generically (§13). — yes / no / not-applicable-because
• [ ] The result was published as found, including where unfavourable to the institution (§14). — yes / no / not-applicable-because
• [ ] Any correction, erratum, or retraction is dated, visible from the original work, and issued within the stated time limit (§15). — yes / no / not-applicable-because
• [ ] The publication's URL and its archived materials are stable and permanently retrievable (§16). — yes / no / not-applicable-because

19. Interfaces to other Standards

STD-X governs the presentation of a published research route — its metadata, register, and states — as a surface; this Standard governs the truth-content and evidentiary process behind that surface. STD-D governs how the underlying entities (ENT-07, ENT-18) are modelled and retained; this Standard governs how they may be used in a study once modelled. STD-S governs access control over the anonymized datasets this Standard requires; this Standard states that anonymization and consent are preconditions, not how they are technically enforced. STD-I governs AI system behaviour generally; §13 states only the disclosure obligation specific to research authorship. STD-Q is the final gate a publication passes before it counts as shipped.

20. Open questions

20.1 The specific numeric aggregation threshold referenced throughout (§6, inherited from Codex 0 §16.7) remains unset pending Executive and Stewardship Office review (Codex 0 §16.15.1); this Standard's clauses apply the threshold once fixed.

20.2 The independent verification mechanism for "outcome realization" (Codex 0 §16.15.2) — third-party audit, sampled follow-up, or partner attestation — is undecided, and §4's evidence table cannot be finalised for that specific metric until it resolves.

20.3 Whether pre-registration (§10) should be filed in a public registry visible before publication, or an internal, dated record sufficient for institutional purposes, is undecided.

20.4 Whether pieces published under the Research Institute's less formal "ideas" track (/research/ideas) carry the full methodology and pre-registration obligations of the "journal" track, or a lighter obligation set explicitly labelled as such, is open.