Claims and interpretations

A claim history wraps an observation with attributed claims. It keeps what a source says, what an evaluator concludes and what was measured distinguishable. Explore the complete claim-history structure.

Three claim shapes

Assessed outcomes are reported, confirmed, absent or inconclusive. They are the evaluator's declarations. An assessed absence is different from an unassessed claim. Missing confidence remains omitted; it is never filled with a default.

Every claim has an ID, recording time, topic and subject. The current subject contract supports the observation or a named measurement. It cannot directly target an arbitrary witness, specimen or contextual field. Topics are descriptive strings, not validated membership in an observable or origin vocabulary.

Sources, locators and input references

Source statements identify a source or product through provenance.sourceRef. Page, time-range and JSON-pointer locators can identify a passage. attributedTo preserves the original voice; extractedBy preserves the person or system that extracted it.

Assessment inputs can cite a claim, observation source, product, value assertion or measurement. The claim: namespace is distinct from the observation's assertion: namespace. References must resolve locally, and method versions must match declared methods.

An assessed claim may have an empty input list so an unsupported declaration can still be preserved. Parsing it successfully does not establish research eligibility. The document, testimony and physical-sample profiles check particular supporting material; they do not automatically endorse claims.

Parse a real example shape

The downloadable synthetic testimony example includes one attributed source statement and an observation with unknown time and position. From the extracted example folder:

import { readFileSync } from "node:fs";
import { parseExperimentalClaimHistory } from "@disclosureos/records/experimental/v2";
const history: unknown = JSON.parse(
  readFileSync("testimony/history.json", "utf8"),
);
const result = parseExperimentalClaimHistory(history);
console.log(result.success, result.checks);
if (result.success) console.log(result.currentClaimRefs);

The archive pins @disclosureos/records@2.0.0-beta.1 directly. Use the same explicit dependency when importing it in your own project. The browser inspector also accepts observation and claim-history JSON.

Expected for this history: structural and semantic checks pass; profile and external checks remain not_checked. The current claim list contains claim:statement.

The JSON Schema is exported at @disclosureos/records/experimental/v2/claims/schema, with identifier urn:disclosureos:experimental:claim-history:0.1.0.

Revisions and disagreement

A revision uses a new ID and an explicit supersedes list; predecessor claims remain in the document. Revisions preserve kind, topic and subject. An assessed predecessor can be superseded only by an assessed claim bearing the same evaluator label. A different evaluator supplies an independent assessment.

A revision cannot be recorded before its predecessor, and an evaluation cannot occur after its own recording time. Inputs and supersession must remain acyclic. Array order does not select a winner. If two revisions supersede one predecessor, both remain current until an explicit revision resolves that branch.

currentClaimRefs means “not superseded within this document.” It does not mean accepted or independently supported. The stateless parser cannot prove that a submission retained everything from an earlier submission; storage must enforce that history separately.

What validation establishes

The parser checks shape and local semantics without fetching URLs, executing methods, authenticating identities or verifying source bytes. Structural failure skips semantic checks. Errors carry codes and JSON Pointers; observation diagnostics retain their /observation prefix.

The assessment guide explains the separate questions addressed by profiles. A profile pass has its stated scope; it never silently changes the meaning of a source statement or supplies a scientific conclusion.

Context subjects

Claim history 0.2.0 can address a particular context entity, field or assertion in an exact document snapshot. Use parseContextClaimHistory and evaluateContextClaimHistory for this workflow. The 0.1.0 examples and profiles above retain their explicit version. See Event and setting for snapshot requirements and evaluation boundaries.

Reviews of witnesses and accounts

For witness, account, procedure or witness-group subjects, use claim-history 0.3.0 with parseResearchClaimHistory and evaluateResearchClaimHistory. Declare exact entity snapshot references and the assertions used by each assessment. A witnessReview retains the evaluator's factors, rationale and uncertainty; it never assigns an automatic credibility score.

The testimony guide includes the explicit exports, runnable example and the limits of supplied-document review. Claim histories 0.1.0 and 0.2.0 retain their own parsers and applicability. The browser inspector dispatches by the declared document version; it does not project new fields into older contracts.

Reviews of document editions

Claim-history 0.4.0 uses research-entities 0.2.0 for exact editions, digital artifacts and custody declarations. Use parseArchivalClaimHistory and evaluateArchivalClaimHistory. Edition citations retain the entity snapshot, edition, artifact and locator. Source-reported artifact reviews remain distinct from assessments made by a named reviewer. See Documents and provenance for a runnable example and checked scope.