Methodology
How the record is built
AbrahamMoses is a structured record, not a commentary. This page sets out what every claim records, what each confidence level requires, how dependencies are evaluated, and the editorial rules that govern publication and correction.
Reproducibility
What every claim records
| Component | What is recorded | API field |
|---|---|---|
| Stable identifier | Permanent claim code and revision number | code, revision |
| Primary witness | Book, chapter and verse, folio or section of the text the claim reports | witnesses |
| Textual edition | The critical or standard edition each witness is cited from | editions |
| Translation | Translator, publication, edition and licence for any quoted wording | quotations[].translation |
| Claim classification | Quotation, textual proposition, exegetical lemma, philological or historical claim, inference, or doctrinal interpretation | classification |
| Evidence | Exact references, with notes, and any relevant textual variants | evidence, variants |
| Confidence | Four axes with written criteria, plus per-tradition acceptance — never one score | confidence, acceptance |
| Tradition | Explicit attribution to a tradition, or to scholarship, or none for a textual report | attribution |
| Counterviews | Rival readings with their evidence and the principles they rely on or deny | counter_models |
| Editorial history | Author, reviewers and dates, change note for every revision | editorial, revisions |
| Machine involvement | Whether AI contributed, how, and whether an editor has verified the result | machine_involvement |
Claim classification
Quotation
The verbatim wording of a witness, with its edition and translation recorded.
Grounding: its witness
Explicit textual proposition
What a text explicitly states. It is a report about the text, and it stands or falls with the text, not with any reading of it.
Grounding: its witness
Exegetical lemma
An intermediate reading step — what a text argues, or how one text reads another — recorded as its own claim so it can be examined and disputed.
Grounding: its witness
Philological claim
A claim about what a word or construction means, resting on lexical, grammatical and comparative evidence.
Grounding: its evidence
Historical claim
A claim about events, dates, documents or transmission, resting on historical and material evidence.
Grounding: its evidence
Doctrinal inference
A conclusion drawn from stated premises. It holds only while at least one of its recorded derivations holds.
Grounding: derived from other claims
Doctrinal interpretation
A reading held by a named tradition or school. It is attributed, never presented as the text, and it holds only while one of its derivations holds.
Grounding: derived from other claims
Never one score
Confidence: separate axes, written criteria
A claim can be textually secure and contested between traditions at once. The axes are recorded and shown separately and are never combined into a single figure; each level has a criterion a reader can check.
Textual certainty — Is the wording the claim relies on securely established in the witnesses?
| 5 · Secure | The wording is stable across all major witnesses of the cited edition; no known variant affects the proposition. |
|---|---|
| 4 · Strong | Minor variants exist in the witnesses, but none alters the proposition. |
| 3 · Probable | Variants affect the wording; the proposition holds on the reading the cited edition prefers. |
| 2 · Uncertain | A material variant bears directly on the proposition; the choice of reading is decisive. |
| 1 · Conjectural | The proposition depends on an emendation or on a poorly attested reading. |
| Not applicable | The claim does not depend on the wording of a particular witness. |
Philological certainty — How securely is the meaning of the wording established?
| 5 · Unambiguous | Lexically and syntactically unambiguous; standard lexica and grammars agree. |
|---|---|
| 4 · Established | The meaning is established; only nuances are debated. |
| 3 · Debated | Standard lexica or grammars differ; the proposition holds on the majority reading. |
| 2 · Contested | The proposition depends on a disputed sense that is not the majority reading. |
| 1 · Speculative | The proposed sense is unattested elsewhere or rests on etymology alone. |
| Not applicable | The claim turns on no contested word or construction. |
Historical support — What does evidence outside the claim’s own tradition support?
| 5 · Independently corroborated | Corroborated by multiple independent sources or by material and documentary evidence. |
|---|---|
| 4 · Corroborated | Corroborated by at least one source independent of the claim’s tradition. |
| 3 · Consistent | Consistent with external evidence, but not directly corroborated by it. |
| 2 · Tradition-attested only | Attested only by sources within the tradition that holds the claim. |
| 1 · In tension | In tension with external historical or material evidence. |
| Not applicable | The claim is about what a text says or means, not about an event, date or document, so external corroboration does not bear on it. |
Inferential strength — How strongly does the claim follow from its stated premises?
| 5 · Direct | The source states the claim; no inferential step is involved. |
|---|---|
| 4 · Strong | Follows from the stated premises without contested assumptions. |
| 3 · Defeasible | Follows if the recorded interpretive axioms are accepted. |
| 2 · Weak | Requires several contested axioms, or an analogy the premises do not compel. |
| 1 · Speculative | The premises are compatible with the conclusion but do not support it. |
| Not applicable | The claim is reported directly from a witness and involves no inference. |
Traditional acceptance — which traditions accept or dispute the claim
- Affirms — The tradition holds the claim as stated.
- Qualifies — The tradition holds the claim in a modified or restricted sense, recorded in the note.
- Disputes — Voices within the tradition contest the claim; there is no settled position.
- Rejects — The tradition’s recorded position denies the claim.
- Not addressed — The tradition does not ask this question in these terms. This is a finding about non-comparability, not a gap in the record.
Acceptance is a matrix, never a percentage. A report of what a text says is not disputed between traditions as a report; disagreement is recorded on the readings built on it.
Dependency semantics
Every conclusion records its derivations — independent ways of arriving at it — and each derivation records its necessary and supporting premises and the reading principles it requires or is strengthened by. A conclusion is not established under the selected assumptions only when no derivation survives; a textual report keeps its standing when the reading built on it is set aside; circular support establishes nothing. The full rules, and the explorer that applies them, are on the Hermeneutics page.
- Axiom stance Holds — The tradition reads with this principle.
- Axiom stance Qualifies — The tradition holds the principle in a restricted form, recorded in the note.
- Axiom stance Rejects — The tradition does not read with this principle.
- Axiom stance Sets aside — A method that sets the principle aside for the purpose of analysis, without judging it.
- Case response Addresses the question — The tradition asks this question in these terms and gives a recorded answer.
- Case response Asks a different question — The tradition engages the material through its own categories. The comparison is shown, but the questions are not the same.
- Case response Not comparable — The tradition does not address this question; forcing an answer onto it would misrepresent it.
Publication states and immutable revisions
Draft
Work in progress. Never served by the public site, API or MCP server.
Never public
Provisional
Awaiting editorial review. Public and citable, but its sources, references and wording have not yet been verified by an editor. Shown with a notice everywhere it appears.
Public
Published
Verified against the cited editions by an AbrahamMoses editor, who is named with the date in the editorial history. A revision without a named reviewer is never published.
Public
Externally reviewed
Published and additionally reviewed by a scholar outside the editorial team, named in the editorial history.
Public
Withdrawn
No longer asserted. The claim stays resolvable so existing citations do not break, and the reason is recorded.
Public
A public revision never changes. Correcting one means appending a new revision with a change note; the earlier revision stays readable at its own address, so a citation always resolves to what was cited. Every public revision’s content hash is pinned in the repository, and a build that alters one fails. Drafts are never served by the site, the API or the MCP server.
Published means a named reviewer. A revision is published only when the editor who verified it is recorded, with the date, in its editorial history; the integrity check fails the build otherwise. In the review sequence Draft → Source-validated → Editorially reviewed → Externally reviewed, a provisional revision is a public draft awaiting source validation and editorial review. Records carried over from the earlier Claim Ledger, where no reviewer was recorded, are provisional for that reason. Their one status correction (2026-10-20) is listed on the changelog.
Editorial standards
Governing rules
- Scripture is the source. Primary text is recorded as text. Nothing is rewritten to fit a conclusion.
- Interpretation is attributed. Every reading names the tradition, author or source that holds it.
- Analysis is identified. Machine or editorial analysis is labelled as analysis, never as source material.
- Provenance is preserved. A finding that cannot show its chain does not enter the record.
- Dispute stays visible. Counterviews and divergence are recorded beside the claim, not hidden behind it.
- Questions are asked in their own terms. A tradition is not made to answer another tradition’s question; where the questions differ, the record says so.
Review
A revision is published when an AbrahamMoses editor has checked its references against the cited editions, its quotations against their translations, and its confidence levels against the written criteria. It is externally reviewed when a scholar outside the editorial team, named in the editorial history, has reviewed it. Research cases are to be reviewed by scholars of the tradition whose question they ask before they leave provisional status.
Machine involvement
Every revision discloses machine involvement: no machine involvement; drafted with ai assistance; machine generated; not recorded at publication. AI assistance may draft structure and metadata; it never publishes. A revision drafted with assistance stays provisional until an editor verifies it. Evidence assessments under Evidence Intelligence record the model, prompt version and provider that produced them.
Corrections
Anyone may report an error with a reference to the claim code. A correction is made by appending a revision whose change note says what changed and why; withdrawn claims stay resolvable with their reason. The changelog lists every revision.
Limitations
- The record holds 74 claims in four research cases. It is a demonstration of method at scale small enough to verify, not a comprehensive theology.
- Revisions added in October 2026 — including every new case — were drafted with AI assistance and are provisional until an editor verifies them. Named editors and external reviewers have not yet been published.
- Tradition stances on reading principles summarise mainstream positions; internal diversity is noted where recorded.
- A tradition lens removes only the principles a tradition is recorded as rejecting; it is a simulation, not a statement of the tradition’s view on every claim.
- Coverage gaps are listed on the corpus page.