Claim Review¶
A public claim can be reviewed by tracing its object, evidence role, scope, precision, provenance, and product decision. The visible map or summary is the starting point, not the complete proof.
Review Path¶
flowchart TB
Claim["public statement or feature"] --> Object["what exact object?"]
Object --> Role["which evidence role?"]
Role --> Scope["which geographic and product scope?"]
Scope --> Precision["which spatial and temporal precision?"]
Precision --> Authority["which governing evidence?"]
Authority --> Source["which upstream identity?"]
Source --> Decision["admitted, qualified, or excluded why?"]Questions That Must Resolve¶
| Question | Evidence to seek | Warning sign |
|---|---|---|
| What is the object? | sample, site, project, paper, source record, grid cell, boundary, or product member identity | one label used for several object types |
| What role does it play? | direct evidence, primary context, contextual domain, sampling context, comparator, or geographic framing | context described as direct support |
| Which scope governs it? | world, Europe-plus, Nordic, country, lake, or another named product | geographic selection presented as representativeness |
| How precise is it? | reported and normalized place/time plus basis and evidence class | decimals or narrow ranges without matching support |
| Who owns the fact? | governing evidence record and fact-ownership relation | reliance on a convenient downstream copy |
| Where did it originate? | dataset, version, accession, DOI, sample label, retrieval context, and source capture | citation without recoverable record linkage |
| Why is it visible? | admission, qualification, comparison, or framing decision | presence on a map treated as equal evidential weight |
Direct And Contextual Evidence¶
Direct evidence supports a claim about its governed sample or observation. Contextual evidence supports interpretation around that claim. Boundaries and hydrography frame scope or sampling. These roles can coexist in one product, but they cannot be substituted for one another.
For example, archaeological density around a lake can affect fieldwork prioritization without proving that one animal sample originated at that lake. A project description can guide recovery without supplying sample-owned chronology.
Precision Review¶
Spatial and temporal precision must be supported independently:
- a named region is not an exact point;
- a site coordinate is not automatically sample-owned;
- a project date is not a sample interval;
- a normalized interval retains the reported text and conversion basis;
- a missing value remains missing when no defensible transformation exists.
Decision Review¶
An admitted record satisfies the named product contract. A qualified record satisfies it only under an explicit caveat. A comparator is present for structured contrast. An exclusion records why a known candidate is absent.
None of these states implies that the underlying source program is complete. Review the relevant readiness, coverage, truth-posture, and exclusion surfaces before making collection-wide statements.
Cross-Surface Agreement¶
When a claim appears in several formats, compare them by identifier rather than by visual similarity:
- the manifest names the product scope and admitted members;
- JSON or CSV exposes structured values and relationships;
- GeoJSON carries geometry and feature-level evidence properties;
- Markdown or HTML explains interpretation, citations, and limits;
- warning and exclusion surfaces account for qualification and absence.
Counts, labels, roles, geography, and precision should agree across those surfaces. A disagreement is not resolved by preferring the most polished view; it is an unresolved publication defect.
Review Checklist¶
Use the checklist against one exact claim and product identity. A blank item is not implicit approval; it is either evidence to recover or a reason to narrow the verdict.
| Check | Pass condition | Stop condition |
|---|---|---|
| object | stable governed identity and observation unit resolve | label conflates samples, sites, projects, papers, or product members |
| source | versioned source member, retrieval context, and native key resolve | citation cannot be connected to the acquired object |
| role | direct, contextual, decision-support, and framing roles are explicit | co-location or presentation is doing evidential work |
| place | locality owner, coordinate basis, precision, and conflicts resolve | display geometry is more precise than governing evidence |
| time | chronology owner, reported wording, dating basis, and normalized posture resolve | contextual time is presented as sample-owned time |
| relation | every cross-record link names its basis and cardinality | a name or coordinate match silently creates identity |
| decision | product, eligible population, rule, disposition, and reason resolve | normalized presence is treated as admission |
| product | manifest, scope, member identity, caveat, and structured views agree | rendered output cannot be reconciled with its manifest |
| independence | supporting observations have distinct governing lineages | duplicated descendants are counted as corroboration |
| absence | capture, normalization, review, admission, publication, and view boundaries are distinguished | non-visibility is described as evidence absence |
Record the first failed governing link and every dependent statement it weakens. Continue far enough to identify affected descendants, but do not use downstream agreement to overrule the failed owner. The review packet is complete when another reader can reproduce the verdict from stable identities without relying on the current visual presentation.
Challenge The Claim Before Accepting It¶
Positive traceability is necessary but can still conceal an overclaim. Apply these counterchecks to the exact proposed wording:
| Challenge | If the answer is no |
|---|---|
| If contextual layers were hidden, would the direct claim still be supported by its owned evidence? | reduce the claim to context or recover direct evidence |
| If the map were unavailable, could the feature still be resolved from manifest identity to governing source? | treat the publication lineage as unresolved |
| Does every stated count name one observation unit and denominator? | separate samples, people, sites, sequences, cells, and source rows before counting |
| Would the claim survive replacement of a precise marker with the source-reported spatial precision? | narrow the geographic wording |
| Would the claim survive replacement of a display date with the governed temporal posture and caveat? | narrow or withhold the temporal wording |
| Are two supporting records independent at the observation-lineage level? | describe agreement as shared or unresolved context, not corroboration |
| Can absence be located at capture, normalization, admission, publication, or current view? | do not interpret non-visibility as evidence absence |
flowchart TD
Wording["proposed public wording"] --> Remove["remove contextual support"]
Remove --> Precision["restore governed place and time precision"]
Precision --> Count["name unit and denominator"]
Count --> Independence["test observation-lineage independence"]
Independence --> Absence["locate every absence boundary"]
Absence --> Verdict["supported, qualified, contextual, or refused"]These checks do not demand maximal evidence for every statement. They ensure that the wording does not gain strength from visual proximity, aggregation, formatting precision, or duplicated lineage.
Review Verdicts¶
| Verdict | Meaning | Reader action |
|---|---|---|
| supported | identity, role, scope, precision, lineage, and decision all resolve | use within the declared contract |
| supported with qualification | the chain resolves but one or more explicit limits narrow the claim | preserve the qualification in reuse |
| contextual only | the record informs interpretation but does not directly support the target claim | describe it as context |
| unresolved | a required link or owner cannot be recovered | narrow or withhold the claim |
| contradicted | governing evidence conflicts with the claim | reject the claim and follow the conflict record |
| out of scope | evidence may be sound but the named product or question does not admit it | do not infer absence from exclusion |
A verdict attaches to a specific claim and scope, not permanently to the underlying record. The same record can directly support one statement, provide context for another, and remain out of scope for a third.
Example: Reviewing A Map Point¶
For an animal ancient-DNA marker, identify the sample rather than stopping at the project or species label. Follow its bundle member to the governing sample record, then verify project and paper lineage, sample-owned locality, chronology, coordinate basis, evidence role, and the point-admission decision. Finally confirm that the map geometry and narrative qualification agree with those records.
If only project geography is available, the review may support a project-level regional statement while refusing an exact sample point. That is a successful review outcome: it preserves the supported claim without concealing the missing evidence.
Review Outcome¶
A claim is supportable when every link resolves and its wording stays within the weakest governing evidence. If identity, role, precision, provenance, or admission cannot be recovered, the appropriate outcome is a narrower claim or an explicit refusal.