Jump to text
On this page
Copy and source
Published source confirmed at last check

Source changed 2026-11-18 21:11:39 UTC · snapshot created 2026-11-18 21:12:35 UTC · last check 2026-11-18 21:55:20 UTC

A.10 - Evidence Graph Referring: Claim-Bound Evidence and Provenance Graph

Type: Kernel pattern Status: Stable Normativity: Normative

A.10:1 - Problem frame

Use this pattern when a source, carrier, result episteme, credential, dashboard, provenance label, generated explanation, model card, or review note is being relied on for a named claim or bounded action and the source-to-use account is still implicit.

Primary EntityOfConcern. The relied-on claim and bounded use. A.10 builds a descriptive evidence-provenance path that represents the independently established sources, carriers, work, result epistemes, provenance relations, currentness, and later-use relations needed to judge that use. The path is not a new world-side relation and its edges establish none of the facts they cite.

First useful move. Name the selected claim C in episteme E and its bounded use U. Recover the cited sources, carriers, and direct relations S, and currentness T when it affects that use. Use §4.5 to determine the bounded A.10 disposition D, or name what is missing for that classification.

When C states a local result R, retain the rule that defines or tests R and the episteme stating it. Add a dated U.Work occurrence only when that Work is itself a current claim. If the account says that Work returned, produced, or obtained a value, name the A.6.1 application binding, one local A.15.PROD claim, or a direct subject relation under its own pattern; otherwise keep the Work and value as separate facts. Mark any missing rule or relation as a gap.

What goes wrong if missed. Carrier presence becomes truth, provenance becomes approval, a result record becomes performed work, MethodDescription becomes a run trace, a graph edge becomes an obtaining relation, and a currentness or assurance decision is inferred from display styling.

What this buys. A source-to-use account that can be replayed, contested, refreshed, narrowed, or handed to the pattern that defines or tests an additional claim, while keeping the claim, carrier, performed work, local result, result episteme, provenance, currentness, reliance, assurance, and action distinct.

Not this pattern when. A.10 does not establish measurement, formal, causal, diagnostic, conformance, comparison, selection, acceptance, gate, permission, commitment, Work, or decision results. It does not establish representation correspondences. Use the pattern that defines or tests each result, A.15.1 and A.6.1 for performed Work and actual bindings, C.2.1 for the result episteme, G.11 for currentness, and B.3 only when a named assurance claim is current. Use A.6.3.RT for a same-EntityOfConcern representation transition, or the applicable transition pattern for another transition. Use C.29 only for an unresolved mathematical-lens choice, transfer or reliance question.

Use A.2.4 first when only the first evidence-use or status-use classification of an episteme is at issue. Enter A.10 when carrier identity, source recovery, provenance, currentness, rival explanations, or bounded reliance must remain replayable.

Here path means a path in a descriptive evidence/provenance graph, never a route of action or a universal evidence relation.

A.10:2 - Problem

Source-backed reasoning fails in recurring ways:

  1. the relied-on claim is not named;
  2. a carrier or publication face is substituted for the claim it represents;
  3. a method description, plan, signature, or stored reference is substituted for actual work and bindings;
  4. a local domain result is replaced by a generic evidence or result field;
  5. provenance, currentness, reliance, assurance, and authorization are collapsed; or
  6. a graph edge is asserted before the direct source, work, production, representation, participation, or use relation is known to obtain.

The practical effect is false authority and unreplayable decisions: a badge looks like permission, a dashboard looks like a gate decision, or a model output looks like an accepted conclusion.

A.10:3 - Forces

  • Minimality vs consequence. Orientation needs a small path; material reliance needs the fields that change the decision.
  • Carrier identity vs claim content. The same content can appear in several carriers and editions; a carrier can be authentic while the claim is false or stale.
  • Reusable method vs performed work. A method is a repeatable way of doing. Ordinary reading, orientation, or reliance can remain ordinary; a claim about performed U.Work requires the dated occurrence and bindings consumed by the claim.
  • Provenance vs result establishment. A.10 must make a result traceable without establishing the result itself.
  • Graph convenience vs ontic discipline. A graph can represent many relations compactly but cannot make them obtain.
  • Contestability vs confidentiality. Reliance must be challengeable while sensitive carriers may require scoped, redacted, hashed, or access-controlled views.

A.10:4 - Solution — recover the objects before drawing the path

A.10:4.1 - Start with the relied-on claim

Name the C.2.1 episteme and select the claim or proposition being relied on from its ClaimGraph. Recover that claim’s subject, interpretation basis, polarity or status when current, and uncertainty or qualification when relevant. If the claim is used as evidence for another target claim, name that target separately under A.2.4.

When the selected claim states a local result, use the pattern that defines or tests that result: C.16 for measurement, C.28 for causal support, A.19 for comparison or selection, G.4 for an acceptance-clause application, A.21 for a gate decision, C.11 for an option-set choice, and the applicable formal, diagnostic, conformance, identity, permission, or commitment pattern.

When role leaves the intended claim unresolved in a technical result, use E.10.ROLE to select the local system-role-kind classification, U.SystemRoleAssignment occurrence or state, relation among system-role kinds, declaration, participation, interface, or representation claim before citing its governor.

A source episteme may contain several claims. Keep an ordinary source citation ordinary; use C.2.1:4.2.5 ClaimAddress only when an intrinsic claim reference is needed and resolves uniquely. If that needed address does not resolve uniquely, follow C.2.1’s whole-episteme or separately identified claim-episteme alternative.

A carrier, citation, provenance entry, or A.10 classification does not constitute the result episteme or the domain result. When their identity is live, use C.2.1 for the result episteme and the subject pattern for the domain result; use A.15.PROD when inception through Work is claimed.

A.10:4.2 - Ground source, carrier, publication, and representation

Recover the selected source episteme and its edition and claim content. When availability, form, or carrier matters, separately recover the EpistemePublicationRelation occurrence, publication form, carrier, or face involved, together with any copy, extraction, or transformation between source and use. Use E.24.PUB for publication, form, and carrier relations and E.17 for multi-view publication. Use C.29 only when a separate mathematical-representation choice, transfer or reliance question remains. The descriptive graph points outward to those independently established objects and relations.

Carrier authenticity, integrity or provenance can support a named origin, history, build or transformation claim under its rule. The content may also support another claim through that claim’s evidence or constitutive relation. Carrier presence or provenance alone establishes no additional truth, approval, permission, assurance or Work occurrence.

A.10:4.3 - Separate method, work, participants, and local result

Under A.3.2, an already identified episteme is a U.MethodDescription when its EntityOfConcern is one admitted U.Method and it makes at least one substantive claim about that Method as a way of doing. Generic participants, parameters, effects—including intended effects—and operating conditions can supply such claims; a name, date, or approval alone cannot. A particular work-plan intention, actual-participant binding, or proof/test occurrence remains a separate claim. Generic participant meanings and method claims do not establish actual operation-application bindings or that Work occurred.

Source production, measurement, verification, interpretation, transformation, query, review, publication, or later reliance may be described ordinarily. When the current claim says that one of these is a dated U.Work occurrence, first recover each performer’s A.13 core and independently admit the occurrence through A.15.1 from its performance history, enacted Method, extent, and containing-System relation. Add F.6 afterward only when the evidence account also needs precise assignment-bound attribution.

The A.13 core always includes the obtaining assignment. A short Work account may omit an assignment identifier unused by the receiving claim only when the complete core and all consumed relations remain recoverable; a precise attribution claim additionally requires the F.6 facts.

Affected or evaluated referents, resources, and participants enter only through direct subject relations or A.6.1 operation-application bindings. Capability, authority, and responsibility remain separate predicates. A compatible signature, plan, description, log schema, or graph node establishes none of those bindings.

For every cited result, recover its defining or testing rule and identify its result episteme separately when that identity matters. A PatternID can locate the rule without requiring a new pattern application. The provenance path may represent Work, participants, entities, domain results, result epistemes, and outcomes only after their direct relations are established.

Relate Work to a returned or produced value only through an A.6.1 application binding, one local A.15.PROD claim, or a direct subject predicate under its own pattern; otherwise show the two facts separately and return the reason-specific non-assertability result if that connection is needed.

A.10:4.4 - Build a descriptive evidence-provenance path

The minimum A.10 path records only what the bounded use needs:

FieldRequired content
Relied-on claimThe selected claim or proposition in one identified C.2.1 episteme; a separate target claim when the evidence use distinguishes it
Bounded useThe ordinary orientation, learning, action, or reliance use being judged, and its premise, reference, decision-use, operation-argument, or other direct use relation. Add a later U.Work occurrence only when it is independently current.
Sources and carriersSelected source epistemes and editions; publication occurrences, forms, carriers, and faces when material; transformations; and direct provenance/citation relations
Work and bindingsOnly independently current dated U.Work, its complete basis, and the performers, Methods, resources, direct relations, and A.6.1 bindings used by the claim. A claimed Work-to-value link also needs an A.6.1 application binding, one local A.15.PROD claim, or a direct subject predicate defined by its own pattern.
Result ruleThe defining or testing rule for each asserted local result, and its result episteme when that identity is used
Time/currentnessSource and result windows plus G.11 currentness when currentness affects use
ChallengeA live rival explanation when it needs discrimination; a locally grounded, plausible unsupported use when excluding it changes reliance; an available contest/redress path when a party is affected; and the stop or reopen condition applicable to this use

For every unsupported-use guard in this pattern, require an independent local ground for the rejected reading, a plausible intended-reader error, and a change in truth, understanding or use. State only the smallest correction needed under F.19. A live rival explanation remains a substantive alternative to test, not a mandatory invented opponent.

Graph nodes retain their admitted kinds. Each edge cites one independently established direct relation; no generic evidences, verifiedBy, validatedBy, measuredBy, producedByWork, or criterion-participant relation is minted as a fallback. A project may label display edges for navigation, but the label has no ontic force.

A.10:4.5 - Classify bounded reliance

The canonical local RelianceDisposition member set is exactly: pass, degrade, abstain, reopen, evidence-needed, assurance-needed, and blocked-current-use. pass supports only the bounded use; degrade supports only the named narrower or reversible use.

assurance-needed says that A.10 alone cannot support the attempted use because a direct domain rule or receiving decision requires a separately stated assurance claim. It creates no assurance claim and does not open B.3 until that claim is current. No disposition is claim truth, CV.Status, gate decision, selector outcome, approval, permission, release, assurance, or Work authorization.

When a named assurance claim is current, use B.3 for that assurance question. A.10 continues to supply the source and provenance paths but does not issue the assurance result. Consequential evidence use without such a claim stays with the direct safety, access, status, gate, permission, release, responsibility, or controlled-action pattern.

A reliance limitation qualifies the attempted use. When advice must also say what to do next, use C.11.DUA to compose a feasible continuation from that limitation. Further inquiry is one possible continuation. Retain the unsupported claim boundary when choosing a narrower use, another action or a stop. Use an already adequate C.11 choice directly.

A.10:4.5a - Route unlike exploratory inputs without changing their kind

When an observation, objective, former cue, novelty characterization, or similar item is proposed as a premise for an exploratory or creative move, recover the item under its direct owner before applying this bounded reliance classification. Do not rename every item signal or cue, and do not create a second premise-disposition vocabulary.

Incoming itemSource/result recoveryA.10 useReceiving choice
Evidence-bearing measurement, assessment, experiment, inference, or capability resultThe selected claim in an identified C.2.1 episteme and the measurement, capability, experiment, inference or other rule that defines or tests its result.State the relied-on claim, evidence-provenance path, bounded premise use, and existing RelianceDisposition; exclude a locally plausible unsupported use when that distinction changes reliance.When an option or probe comparison is current, C.11 compares the available alternatives and emits its ChoiceResult.
Objective, reward, utility term, loss, preference, or heuristic that is not evidenceThe objective, evaluation, preference, Method, or source-local construction.Apply A.10 only to a separate evidence or source-reliance claim about that construction or its bounded transfer; the numeric objective is not self-authenticating evidence.Enter the term as the declared EvaluativeMeasure, PreferenceOrder, or ChoiceRule input that it supplies, with assumptions and limits visible.
A former pre-articulation cue that has now been articulatedA.16.1 no longer owns the articulated result. Use B.4.1, B.5.2, or the endpoint claim owner selected by the articulation.Qualify reliance only when that articulated claim is used as a premise.C.11 owns any current option/probe comparison; the earlier cue pack neither selects nor evidences the move.
A non-evidential C.17 novelty, surprise, use, or creativity characterizationC.17 owns the characteristic claim, its scale/basis, and its limitations.Apply A.10 only when evidence or source reliance for that characteristic claim is current; characterization is not evidence merely by being decision-relevant.C.11 may consume the bounded characteristic together with other premises and still choose, reject, probe, or reroute.

The composition retains the source result and an existing RelianceDisposition when bounded reliance is current. A later ChoiceResult is required only for an actual option/probe comparison under C.11. If the source claim is already qualified and the current C.11 record can use it directly, stop; no intermediate premise record is required.

A.10:4.6 - Currentness, actual use, and graph limits

Source availability and source currentness are distinct. Record issue/effective windows, supersession, revocation, source-order rules, and the G.11 currentness result when a use depends on them.

Actual reliance requires one premise, reference, decision-use, operation-argument, or other direct relation to the relied-on episteme. When that reliance is also claimed as dated U.Work, recover the Work independently. Storage, indexing, citation, graph membership, visibility, or co-location establishes neither reliance nor performed Work.

Part-whole, temporal, production, publication, representation, provenance, participation, and reliance relations must each be established separately. The A.10 graph may cite them together for replay but never substitutes one for another.

A.10:4.6a - Authority-reliance use of ordinary A.10 evidence-provenance paths

Use this subsection when an authority-looking carrier is being relied on. The A.10 path represents one named claim, its sources and direct relations, and one bounded use; it is not an authority relation. If the Work occurrence, gate decision, speech act, commitment, permission, system-role assignment, assignment-state assertion, or other required relation already exists in a project-side source, recover that object by value and let the graph cite it.

Start with A10-lite for source-finding, orientation, learning, and bounded reversible probes. It is sufficient only when these fields supply the evidence required by the claim and its rule; reversibility alone does not establish sufficiency:

FieldRequired content
claim or effectThe claim, effect, or source-backed reliance use the evidence carrier is being asked to evidence for the named work occurrence or reliance use.
evidence carrierThe display, badge, credential, attestation, dashboard tile, copied text, generated text, log, trace, source file, report, or other SymbolCarrier/publication carrier.
producer, issuer, verifier, or source contactName the producer, issuer, verifier or source contact and the direct relation through which it bears on the claim only when the evidence use needs them. A recognizable ordinary actor designation needs no further expansion. If dated Work is asserted, first recover each performer’s A.13 core and independently admit the Work under A.15.1. The core includes the obtaining assignment; add an F.6 link through that same assignment only when this evidence path consumes precise assignment-bound attribution.
method use or Work occurrenceName the ordinary source-finding or method use. Add admitted measurement, verification, review, build, attestation, copy, extraction, generation, query, trace, or log U.Work only when independently current. If that Work is said to have returned, produced, or first constituted the carrier or result, cite an A.6.1 application binding, one local A.15.PROD claim, or a direct subject predicate under its own pattern; otherwise keep the facts separate.
time windowIssue time, effective window, decay, supersession, revocation, policy or gate version, and reopen condition, each when the selected source or bounded use depends on it. Name a missing required value rather than inventing one.

Minimum evidence-provenance path for routine reliance:

FieldRequired content
evidenced claim or effectApproval, permission, gate passage, local system-role-kind classification, system-role-assignment occurrence or state, relation among system-role kinds, status currentness, work occurrence, evidence relation, assurance input, or other claim named by value or effect being attempted. Use E.10.ROLE only when a technical role claim remains ambiguous.
evidence carrierThe visible or recovered carrier, with enough identity to reopen it.
issuer, performer, trust root, status register, and source-side predicatesName every object this path uses: the admitted System performing source-side Work; the trust-root or status-register episteme, register, or service; and the issuer, publication, registration, status-source, source-maintenance, trust, acceptance, or currentness predicate by which that object bears on the relied-on claim. For admitted Work, first recover every performer’s A.13 core and independently admit the Work under A.15.1. The core includes the obtaining assignment; add an F.6 link through that same assignment only when this path consumes precise assignment-bound attribution. If no current pattern defines or tests the needed predicate, return the A.6.RCD missing-governor result. Authority and source-maintenance responsibility remain separate relations.
affected entity and relying contextThe release, service, model, person, admitted System and any separately obtaining assignment, policy subject, work target, claim, audience, tenant, environment, or other entity for which reliance is attempted.
time window and freshnessRetain each issue/effective, decay, supersession, revocation, policy/gate-version, or reopen value that the selected source or bounded use consumes; report a missing required value.
relevant Work occurrence or method traceAny independently current production, verification, query, generation, review, or other U.Work, plus the method trace when the method matters. Connect that Work to the carrier or result only through an A.6.1 application binding, one local A.15.PROD claim, or a direct subject predicate under its own pattern; otherwise record them separately.
evidence relation and rival explanationWhich source claim or content bears on the target claim, through which evidence relation, and which live rival explanation must be distinguished—for example, stale status, a spoofed badge, a copy changing the claim, or a scope change defeating the use. Recover the relevant episteme from its carrier.

Expanded fields are collected only insofar as they decide the current reliance question. Evidence depth follows consequence severity, reuse, contestability, cross-context movement, and the evidence relation required for the attempted claim. Do not expand a source-finding note into a full evidence dossier, and do not collect every expanded field merely because a carrier is copied, generated, credential-like, provenance-like, or cross-context.

Adversarial misuse guard. When a carrier appears to support a claim, name what it appears to establish and any live competing explanation. An apparently current credential may have an authentic carrier but a stale displayed status; a copied approval may be genuine but concern a different scope or window. Test the source, issuer, or currentness predicate and relying context that distinguish that explanation; include source-side Work and precise attribution only when the claim uses them. Authenticity and provenance can contribute to their named claims, but appearance or provenance alone does not establish additional truth, currentness, or authority. If the required predicate has no governor, return that A.6.RCD gap.

Data-minimization and privacy boundary. Preserve the source, provenance and direct-relation account sufficient for the intended use. Use redacted, hashed, scoped or access-controlled references when raw material would disclose secrets, unnecessary personal or tenant identifiers, privileged records, or sensitive security, incident, model or data provenance. Redaction creates no source relation; it must preserve enough recoverability for the relying context.

Expanded fieldWhen it is needed
method trace or work traceThe selected provenance, attestation, generated/copy/dashboard/rollback source relation, or Work claim depends on how the method was applied or the Work occurred. A source relation alone need not assert Work.
evidence-carrier integrityA plausible spoof, stale source, copy, transformation, rendering, redaction, or context shift could change the relied-on claim or its use. Check the integrity property that discriminates that risk.
identity or holder bindingThe claim depends on a credential holder, admitted System, separately obtaining assignment holder, acting holon, issuer, performer, delegate, revoker, verifier, or relying party.
verifier context, relying-party context, and acceptance ruleThe evidence relation is accepted only for a verifier, audience, tenant, environment, release line, policy subject, operational mode, or consumer-side policy or gate rule that accepts the evidence for this use.
proof, cryptographic-signature, or status verification resultThe selected credential, provenance, attestation, authenticity, revocation, or currentness claim requires that verification result under its source specification or verification/use policy. Retain every check required by that regime.
policy version, gate version, and decision sourceThe attempted permission, release, rollback-authority, policy-authorization, or other use depends on that policy or decision. Gate version and gate-decision source are required for a gate-dependent use.
source-chain transform notesEvidence relation passed through extraction, copy, rewrite, representation shift, explanation rendering, summary, export, redaction, or another transform step before reliance.
source order and supersession ruleMultiple source candidates disagree or freshness or priority may defeat the visible publication face, publication carrier, rendering, or cue. Include the direct register or status-source-order relation when a register entry is the source for a system-role-assignment occurrence, status assertion, permission, commitment, or gate state.
minimum disclosure boundaryRaw evidence would expose secrets, personal data, tenant identifiers, privileged logs, tokens, security-sensitive traces, or unnecessary identities.

Case repairs:

CaseEvidence repair
Stale credential badge or status displayName the issuer or trust-root object and its direct issuer or trust relation; name the status register, entry, and direct registration or status-source relation when one exists; then show the verifier and relying-party context and the proof, status, freshness, window, entry-version, and integrity facts needed to resolve the stale-source claim. Include holder or subject binding when the claim or regime requires it, validity limits when present or required, and revocation/status checks when the mechanism is present or the selected verification/use policy requires them. Display presence is not an obtaining system-role-assignment occurrence, status assertion, or permission.
Verifiable credential, credential view, or register excerptTreat it as an A.10 carrier. Name the issuer or trust-root object and relation; the status register, entry, and registration or status-source relation when present; the selected source U.Episteme and edition and, when availability matters, its EpistemePublicationRelation; verifier, relying context, acceptance rule, and the proof/currentness facts required by that rule. Include holder or subject binding only when required by the claim or verification regime, validity limits when present or required, and status/revocation checks when that mechanism is present or the selected verification/use policy requires it. Passing the applicable checks may evidence credential currentness for that bounded use; it does not imply a holder-bound claim where none was established. A strong grant, exercise, weak non-prohibition or non-violation finding, or conflict requires A.2.8.PER; a commitment requires A.2.8; an issuing act requires A.2.9; a system-role assignment requires A.2.1; a status assertion requires its status pattern; an entry predicate requires its defining pattern and A.6.B boundary classification; and gate passage requires A.21. Display presence creates none of them.
Copied approval or review summaryShow the original A.2.9 SpeechActRef or issuing act when approval or authorization is claimed, or the original reviewed source when only review-content currentness is claimed. Add the copy relation, currentness, scope, and window. Add a separately identified dated U.Work only when it is current, and connect it to the copy or result only through an A.6.1 application binding, one local A.15.PROD claim, or a direct subject predicate defined by its own pattern. State separately whether the claim concerns an A.2.8.PER grant, finding, exercise, or conflict result; an A.2.8 duty, recommendation, or prohibition commitment; or another Work relation. Copy evidence is not approval by itself.
Provenance, authenticity, or attestation labelShow the bounded origin, history, build, or process claim; selected source U.Episteme, the EpistemePublicationRelation occurrence when availability is material, or evidence carrier; the method/Work trace, source-specific proof, and carrier-integrity facts that this claim and its verification regime require; the verifier or relying policy that accepts them; and any live rival that changes reliance. Provenance does not show truth, safety, approval, release, gate passage, permission, or assurance unless another FPF relation named by value carries that additional claim or effect.
Dashboard status tileRecover the dashboard query and the source relation or source-bearing record it uses, with the time, window, currentness, source order, freshness policy, and live rival relevant to the claim. For a gate-dependent use, cite the current A.21 GateDecisionResult, directly or through its DecisionLogRef, with gate profile, gate version, release target, and work target. For a release claim not dependent on a gate, cite that claim’s own rule and decision source; do not invent a gate result. A.10 records this source-to-use account. A status display is not gate passage or Work occurrence by itself.
Rollback command-like cueRecover whether the claim concerns a command, an issuing speech act or only an A.6.A action invitation. Name the actor, target, scope and window the claim uses; recover the command record or issuing act and any authorization relation only when the claimed use depends on them. A command cue is not performed-work evidence.
Rollback performed-work resultRecover the independently admitted A.15.1 U.Work occurrence and the Method, performance evidence and time window relevant to the performed-work claim. Include logs or outcome evidence when that claim uses them. Performed-work evidence is not approval, assurance, or gate passage by itself.
Generated explanationRecover the source or source-to-target relation claimed. Use E.17.EFP only when an unresolved explanation-use ambiguity changes the interpretation or use. For reliance, recover each operative claim’s source basis or independently established inference/result under its rule, including any source passage, carrier, relationFunctionClaimRef or authoritySourceRef that the claim uses. A.10 can then qualify reliance on that claim. Fluent wording alone supplies no evidence, approval, authorization, gate passage, performed Work or assurance.
Model card or datasheet used as evidenceShow documented bounded-use statement or external intended-use field, version, window, evaluation condition, limitations, evidence carriers, and whether a B.3 assurance claim is being made. Documentation does not become readiness or assurance by presence.
Extracted source-to-use path to gate or release claimName the selected source U.Episteme ref and, when availability is material, the EpistemePublicationRelation occurrence ref; the source-bearing relation or pattern reference that identifies the rule carrying the claim; the transformation chain and any loss or non-commutativity, identifying its first step when present; the FPF relation or pattern that defines or constrains each relevant transform (A.6.3.CR, A.6.3.RT, A.6.3.CSC, E.17.EFP, E.17.ID.CR, or E.18 where applicable); and the bounded inference relation after the transform. A lossless chain retains those relations without a lossy step. When the claim consumes a relation-function or authority-source claim, name its relationFunctionClaimRef or authoritySourceRef; otherwise retain the direct source reference already recovered. State what change reopens the use and which source, transformation, evidence relation or rule must then be rechecked. Identify any gate or release claim whose use remains blocked by an unresolved dependency.
Conflicting source relationsWhen display, source publication carrier, decision log, recency signal, freshness signal, copied summary, generated summary, credential status, provenance label, or assurance evidence disagree, name the visible source relation, rival source relation, source-order rule, decision-source relation, freshness policy, and supersession rule. Do not choose by color, visual salience, confidence wording, copied wording, or apparent recency; the work claim or reliance claim is contested until the source-order question is resolved.
Sensitive evidence-provenance pathUse redacted, hashed, scoped, or access-controlled carrier refs when raw carriers expose secrets, personal data, security-sensitive traces or data, privileged logs, tenant identifiers, or unnecessary identities. Redaction does not create a source relation; it must preserve enough recoverability for the relying context.
Pointer or proof-status evidence-provenance pathUse a hash, proof or status verification result, selected source U.Episteme, EpistemePublicationRelation occurrence when availability matters, source or source-currentness relation, scoped pointer, disclosure receipt, or access-controlled view instead of copying raw sensitive carriers or payloads when that pointer preserves enough recoverability for the relied-on claim or effect. Do not copy raw secrets, tokens, privileged logs, personal identities, or tenant details merely to make the path look fuller.

If the path is incomplete, name the missing source, carrier, Work, result rule, direct relation or needed currentness fact and narrow or block only the dependent use. Select a feasible continuation under C.11.DUA when one is still needed: recover or verify the source, refresh a query, narrow reliance, choose a permitted reversible probe, or stop. An explicit WorkPlan is needed only when planning that continuation requires it; reversibility alone establishes no permission or evidential sufficiency.

Missing source-relation repair assignment. If the relying actor cannot recover or verify the source relation, first name the missing relation or source-bearing record and the affected use. Source-finding, a request, narrowing, or a stop may be the complete continuation.

If the selected continuation allocates or requests repair, recover the recipient and the independently obtaining project-side responsibility, allocation, permission, or commitment relation that the instruction consumes. An assignment to an admitted System must identify that System and its assignment basis. Plan or request the future repair under its applicable planning or assignment rule; this is not an assertion that repair Work has occurred.

The source-side objects and relations identified above remain separate facts, not responsibility by source label or form. Return an A.6.RCD missing governor only when the selected assignment claim requires that missing predicate. Source exposure and the affected party’s challenge remain available independently of who may later perform repair.

ViewpointPrompt
Relying actorWhich claim named by value or effect needs an evidence relation, and what is the minimum carrier, source-bearing record or relation, time, and evidence-provenance path for that claim or effect?
Issuer, verifier, or status relation maintainerWhich facts in the selected source relation and verification regime must be exposed or repaired for this claim?
Auditor or technical reviewerCan the carrier and selected source relation be recovered, together with the method/Work trace, time, and live rival that this reliance question needs?
Security reviewer or compliance reviewerWhich applicable verification, source-order, supersession, and disclosure conditions decide this reliance question?
LLM user or tool userWhich generated or copied operative claims map to source passages or carriers, and which claims remain only source-finding?
Author of model documentation or data documentationWhich intended-use, evaluation-condition, version, window, limitation, and evidence carriers bound the model documentation or data documentation?

Repeated source failure. When repeated stale, contested or missing-source findings suggest a shared source or display defect, check that explanation before selecting a repair. If the defect is established and correction is needed, repair the affected source relation or display: for example, expose the decision-source reference, add a needed currentness check, preserve claim-to-source correspondence through copying, or show the scope of an attestation. Recheck uses that depended on the repaired source. Repetition does not require each relying user to reconstruct the path manually.

Display the claim, recoverable source and carrier references, relying context, and the time/currentness values that affect use. Add a guard only under §4.4’s F.19 conditions. A display of source availability must identify that limited claim; approval, permission, gate passage, Work and assurance retain their separate rules.

Incident-learning fields for evidence and currentness overread: visible carrier or publication face, intended claim or effect, missing evidence-provenance field, evidence carrier named by value, source-side predicate used, ordinary method trace or admitted Work trace, and needed time relation; any live rival relevant to the incident; current safe disposition; and the smallest upstream repair to the implicated source relation, instrumentation, or publication. Identify the selected source episteme and the publication occurrence, form, carrier, or source display when that object needs repair; the field definitions above determine which facts must remain recoverable.

Contestability and redress relation: when an evidence-provenance path or source-currentness relation affects person or team status, access, responsibility, a compliance relation, or a release decision, the A.10 result names the disputed claim, evidence carrier, affected use or harm, available challenge, review, redress, communication, source, publication, register, access, or contact relation, allowed evidence or argument, possible disposition change, outcome route, reopen trigger, and safe interim disposition.

Source exposure remains independent of who may later perform review or repair Work. Name a responsibility, allocation, commitment, permission, or authority relation—or its missing governor—only when the current claim consumes that relation, including when allocating future review or repair; its absence does not close the challenge.

Positive repaired evidence-use statement. When the source account is complete, write the smallest bounded statement: named relied-on claim; carrier and source; direct provenance and citation relations; ordinary bounded use; and RelianceDisposition. Include currentness, a locally plausible unsupported use, and a stop or reopen condition when each changes that reliance decision.

Add producing or interpreting U.Work, each performer’s A.13 core, independent A.15.1 admission, Method, actual bindings, and later Work only when those facts are current. Add F.6 afterward only when the receiving account needs precise assignment-bound attribution. A claimed Work-to-value link needs an A.6.1 application binding, one local A.15.PROD claim, or a direct subject predicate under its own pattern. Add authority or responsibility only when a relation is independently required by the use. A short Work statement may omit an unused assignment identifier only when the complete A.13 core, including its obtaining assignment, and every consumed relation remain recoverable; an assignment is never the authority or responsibility result.

A.10:4.6b - Local evidence-use classifier and RelianceDisposition for source-bearing carrier or display reliance

Use this subsection when a visible carrier, publication face, selected source U.Episteme ref, EpistemePublicationRelation occurrence ref when availability is material, source relation ref, or display is being relied on for a named claim or act. First recover the claim kind, the pattern or source rule that defines or tests the claim, the source/provenance path, and the bounded use. Broad words such as source, metric, confidence, conformant, safe, ready, certified, approval, or permission are recovery prompts, not relation names.

This is a local reliance-use classifier, not a Core evidence-kind ontology. Use only the row that decides the attempted use. The path represents direct relations and the RelianceDisposition records one bounded A.10 judgment; neither becomes a general evidence or authority relation. Affordability: ordinary orientation or source-finding can stop here; bounded reliance states one evidence use and the currentness window, unsupported-use boundary, and stop or reopen condition that apply to it. If a domain rule requires assurance, state the assurance claim and then use B.3. Plain wording remains ordinary unless it changes one of the named source, evidence, gate, assurance, Work, decision, or control claims.

Stop when the bounded claim and use, source, evidence relation and applicable currentness and stop/reopen conditions are recoverable. Reuse already established subject results when their grounds still apply. Open B.3, A.21, B.2.5, a Work pattern or another supplier only for an unresolved question under its entry conditions; an official-looking, quantitative, generated, credentialed or safety-related display does not itself create that question.

Plain disposition palette: RelianceDisposition=pass means proceed only inside the bounded evidence use; degrade means use only a narrower or reversible version; abstain means do not decide yet; reopen means a changed or contested evidence relation defeated the previous classification; evidence-needed names missing evidence at the decision point; assurance-needed says a separately stated assurance claim is required before this attempted use can proceed; blocked-current-use blocks the attempt until its evidence-provenance path or source relation changes.

Source-looking evidence use or attempted useFirst A.10 actionEscalation triggerForbidden overread
Ordinary source-backed report, record, citation, observation, model card, datasheet, data card, or publication excerptName the claim, carrier, producer or Method trace when relied on, evidence-provenance path, and bounded evidence use. Add currentness, a locally plausible unsupported use, and a stop or reopen condition when they change reliance.Open B.3 only when a named assurance claim is current; open A.21 for a relied-on gate decision, A.15 or A.15.1 for Work, or another pattern only when that relation is claimed.Evidence presence as approval, gate passage, assurance, release permission, Work authority, control authority, or safety acceptance.
Confidence, calibration, prediction interval, abstention reason, or selective-action cueName the act and the assumptions of the selected inference or calibration method. Retain its calibration population and exchangeability when that method requires them, and the window, shift, applicability, and stop condition that bound this evidence use. Use pass or degrade only for that use; exclude a locally plausible unsupported wider use when the distinction matters.Open C.27 or G.11 when timing, expiry, refresh, distribution shift, monitoring, or applicability changes the act; open B.3 only for a named assurance claim.Confidence as global permission, trust, readiness, safety, release reliance, or engineering justification.
Generated explanation, generated summary, or didactic reconstructionRecover the source or source-to-target relation claimed; use E.17.EFP only when explanation-use ambiguity still changes the interpretation or use. For a claim used beyond source-finding, recover its source basis or independently established inference/result and its A.10 evidence-provenance path.Apply the rule defining or testing that operative claim; B.3, A.21 or A.15 applies only to its own current question.Treating fluent wording alone as evidence, assurance, approval, gate passage, Work or permission.
Conformance label, CV.Status, benchmark result, score, semantic-fidelity marker, or CV-looking publication near releaseRecover the declared relation: measurement or marker relation, A.20 step-local CV status, A.21 gate check, E.19 admission/refresh review result, E.21 pattern-quality evaluation result, C.16 characterization, or external-rule source named by value. A score label alone selects neither E.19 nor E.21.Open A.21 only when a named gate applies its profile rule to the effective check-application results and returns a GateDecisionResult; open B.3 only when an assurance claim is being made.Conformance or a score as an unrestricted value, adequacy, release, safety, trust or gate conclusion outside the declared relation.
Provenance, authenticity, C2PA-like credential, SLSA-like attestation, build record, or status-register displayState the bounded origin, history, build, or currentness claim being evidenced and the source relation carrying it. Retain the method/production trace, holder, status, verifier rule, and relying-context facts required by that claim or verification regime.Open the record or relation that carries truth, permission, safety, release, gate passage, work occurrence, or assurance only when that relation is being claimed by value.Provenance, authenticity, or status-currentness as truth, safety, approval, permission, release, gate passage, or assurance.
Contest, redress request, challenge, appeal, or conflicting source relationName the contested claim, evidence carrier, source-order or currentness issue, affected use or harm, available challenge or redress relation, allowed evidence, possible disposition change, outcome route, and reopen trigger. Add a review-responsibility relation or missing governor only when the claim assigns future review Work.Open neighboring system-role, assignment-state, commitment, gate, control, assurance, Work, or representation patterns only when those effects are claimed by value.Appeal-channel presence, challenge form, or redress workflow presence as truth, compliance proof, social-effect acceptance, completed redress, gate passage, or work authorization.

For A.10 use, RelianceDisposition is a local disposition over the evidence-provenance path and the bounded reliance use. Outside a table column already headed RelianceDisposition, write the qualified form RelianceDisposition=... and bind it to the named bounded evidence use, with currentness and window, an unsupported-use boundary, and a reopen or stop condition when applicable; it is not CV.Status, GateDecisionResult, selector result, or ProblemCard@Context state.

Observed-effect or consequence evidence supports what happened or is credibly recorded. If the claim says that the source caused, prevented, would have changed, or causally contributed to the effect, use C.28 for that causal question while retaining its A.10 evidence path and any separately required Work or assurance relation. A normative responsibility or allocation claim uses its own rule; the phrase ‘responsible for’ must be resolved from the source claim before selecting either branch.

If a proxy marker, benchmark, confidence value, dashboard metric, or score becomes the primary driver for action, release, resource allocation, people status, team status, or P2W priority, check whether the claim being made also raises an E.13 proxy-to-objective question. Do not open E.13 for every metric; open it only when the proxy is being used as the target or decision driver.

Use C.26.1 when publication, observation, or another interaction changes the represented state while its output is being used as if it were a passive read, export, comparison, or decision input. First identify what the interaction changed and which use of the output must change. This is the probe-coupled question for a dashboard, warning, label, or public status display; ordinary influence without the false passive-read use does not activate it.

RelianceDispositionA.10 classificationMinimum A.10 statement
RelianceDisposition=passThe evidence relation named by value is present and current for the named use, evidence of the required kind is present, the source relation is current enough for that use, and the evidenced use is bounded.State the evidenced claim, act, work occurrence, review claim, or P2W carry-through use and its evidence-provenance path. Add the applicable window and any locally plausible unsupported use that must be excluded.
RelianceDisposition=degradeThe source relation carries only a narrower claim, smaller audience, reversible local act, lower assurance input, or shorter window.State the narrowed bounded evidence use, the proposed or locally plausible wider use it excludes, and the applicable stop condition.
RelianceDisposition=abstainEvidence is insufficient, stale, out-of-context, uncalibrated, conflicted, or not tied to the claimed relation, while immediate rejection is not justified.State the claim not decided and the missing evidence or relation needed before use.
RelianceDisposition=reopenA contest, changed representation, changed selected entity, stale source, expired window, changed profile, conflicting source, retargeting, or new evidence defeats the previous evidence-provenance path.State the source or relation to reopen and the previous use that is no longer evidenced.
RelianceDisposition=evidence-neededThe visible carrier, selected source U.Episteme ref, publication-occurrence ref when availability is material, source relation ref, or display may matter, but evidence of the required kind or the source-currentness relation is absent.State the missing evidence kind, the pattern or source rule that defines or tests it, and the decision point so delay does not become indefinite.
RelianceDisposition=assurance-neededA domain rule or receiving decision requires a separately stated assurance claim before the attempted use may proceed, and the current A.10 basis alone cannot supply it.State the required assurance claim and its domain basis. Apply B.3 only after that claim is current; until then block or narrow the attempted use.
RelianceDisposition=blocked-current-useNo current evidence-provenance path carries the evidence relation needed for the attempted act, work, claim, gate, release, assurance, review, control-bearing feedback, or P2W use.State the blocked use and the neighboring pattern or project record required before a new attempt.

Minimum contest relation with possible redress: a contest relation exists when the affected party can identify the disputed claim or source, affected use or harm, an available challenge, review, redress, communication, source, publication, register, access, or contact relation, evidence or argument allowed in challenge, possible disposition change, outcome route, and reopen trigger. A system-role label, assignment, feedback channel, complaint form, or appeal label without those recoverable values is not enough to change the disposition. Responsibility for future review Work is a separate claim.

Affected-party contestable minimum: even when raw evidence stays restricted, the contesting party must be able to see enough of the claim, source class, disposition, affected use, available challenge or contact route, and allowed challenge evidence to challenge the result. Privacy, security, or privilege can narrow disclosure; they cannot erase the challengeable minimum while still claiming contest or redress.

False-negative reliance guard: a blocked, abstained, or evidence-needed use is not final if challenge evidence, missing affected-party evidence, changed source relation, changed selected source U.Episteme edition, changed EpistemePublicationRelation occurrence when availability is material, changed publication form, changed evidence carrier, changed representation, or redress can materially change the disposition. If refusal is based on missing evidence, name the missing evidence kind and decision point rather than closing the dispute by vagueness.

Sensitive evidence boundary: use scoped, hashed, redacted, or access-controlled evidence refs when raw carriers would expose personal data, secrets, tokens, privileged logs, tenant identifiers, incident details, security-sensitive traces, or unnecessary identities. A redacted path must still preserve enough recoverability for the relied-on claim, disposition, and contest relation.

Worked source-overread slices:

SliceA.10 usable classificationUnsupported lift
Software supply-chain attestation is cited near a release conversation.The attestation may evidence its named bounded origin, build-method, or production-trace claim. Recover the verifier-rule, holder, and currentness facts that this claim or its verification/use regime requires.Runtime safety, release approval, gate passage, or assurance unless B.3, A.21, or another relation that carries the asserted use is established for that use.
A verified provenance credential, watermark, or authenticity mark appears on a publication face.The mark may evidence where the carrier, signature, assertion, or manifest came from under the verifier regime.Truth of the represented world-state, safety, permission, or adequacy by provenance alone.
A confidence interval or calibration result is used for one reversible act.State the act, context, calibration condition, and bounded evidence use, with the applicable window, unsupported-use boundary, and stop condition.Global readiness, trust, safety, release reliance, or engineering justification.
A generated explanation or summary says a result is reliable.Keep source-finding or explanation use until the operative claim has a recovered source basis or independently established inference/result under its rule and bounded A.10 reliance.Evidence, approval, gate passage, Work or assurance inferred from fluent wording alone.
Contest or redress is claimed after a source relation, selected source U.Episteme, publication occurrence, publication form, or evidence carrier is challenged.State the disputed claim, affected use or harm, available challenge or redress relation, allowed challenge evidence, possible disposition change, outcome route, and reopen trigger. Add future-review responsibility only when that stronger claim is current.Claim truth, compliance proof, completed redress, or social-effect acceptance by appeal-channel presence.
A harmed party gives challenge evidence that could change the disposition, but the receiving System answers “evidence insufficient” without naming the missing evidence kind or decision point.Treat the refusal as RelianceDisposition=reopen or invalid RelianceDisposition=evidence-needed; name the missing evidence kind, decision point, available challenge route, and possible disposition change.Closed refusal, completed redress, or RelianceDisposition=blocked-current-use by vague insufficiency.

A.10:4.6c - Route a Changed Claim Across Several Actual Receiving Uses

Use A.10.1 when a later or replacement source may materially change a claim and the receiving uses must still be found across a bounded frame or closed across several actual uses. A.10 continues to govern each one-use source-to-use account, direct-use relation, and RelianceDisposition. Applying A.10.1 bounds the receiving-use search frame, states source-outward and receiver-oriented coverage and gaps, classifies found candidates as depends, mentions only, or unresolved, and identifies the depends branches that need subject judgment. This includes a conclusion whose scope may change while the decision stands.

If one already-known bounded reliance use is the whole question, apply A.10 and the direct subject guidance. A citation, carrier, declared edge, or graph-reachable node does not become affected merely because the source changed. The completed A.10.1 account cites each independently obtained subject result afterward; it neither replaces that result nor changes A.10’s disposition set.

A.10:4.7 - Causal support in evidence-provenance paths

An A.10 path used for a causal claim cites the C.28 components it carries; it does not compress them into one alternative-valued “support basis”.

causalSupportComponentRefs?: CausalSupportComponentRefs
causalUseSupportResultRef?: CausalUseSupportResultRef

CausalSupportComponentRefs remains defined only by C.28. A.10’s ordinary evidence-provenance path still identifies the source, carrier, Work or data, provenance relations, and bounded use required by this pattern. Inside the C.28 contract, cite only the components the causal claim relies on. When C.28 admits another specialist component, A.10 can cite it through that contract.

Examples:

  • an observational cohort path cites the observation and measurement Work plus observationalOrNaturalBehaviorData; an intervention-effect statement still needs a C.28 identification or design result;
  • a randomized estimate path cites assignment and Work evidence, its identification or design result, estimate, uncertainty, and limits;
  • a prospective counterfactual-sampling path may cite the realizability result, its decision Method, the sampling construction or obstruction required by its status, and any unresolved question; counterfactual-quantity bounds use a separate identification result, and the prospective path claims no performed sampling or data;
  • a performed counterfactual-sampling path cites independently admitted dated sampling Work and the resulting sample or data; add precise assignment-bound attribution only when the receiving support claim uses it. Only the complete Work/data path may support realizedCounterfactualSamplingData;
  • a simulation path cites model output, assumptions, validation, and bounded model use; it does not become realized or interventional evidence by relabeling;
  • a target-trial emulation path cites its TargetTrialMappingResult, including the observational source, protocol-to-data mappings, gaps, residual-confounding assessment, and sensitivity mappings; reporting completeness alone establishes neither identification nor low bias;
  • an off-policy or causal-RL path cites its OffPolicyCausalEvaluationResult with the question, behaviour and evaluation policies, overlap check, supported and unsupported use, and reopen condition. It also cites optional details about history or horizon, confounding, changed endpoints or transport, the estimator, and uncertainty only when they change what the path supports, where that support came from, whether it is current, its bounded use, or when it reopens;
  • a causal-representation path cites its CausalVariableRepresentationRecord with the question, source representation, selection or abstraction Method, representation assumptions, intervention-validity result, supported and unsupported use, and reopen condition. It also cites optional invariance, fidelity, query-preservation, uncertainty, or shift-limit results only when they change what the path supports, where that support came from, whether it is current, its bounded use, or when it reopens;
  • a transport path cites every changed endpoint, assumptions, overlap evidence, formula or comparator, uncertainty or unresolved assumptions, and bounded use carried by its transportability result; and
  • an observationally identified estimate cites both the evidence path and separate identification and estimate results.

What changes in practice: the path exposes where every relied-on component came from and may cite the C.28 support-result episteme. A.10 creates none of the C.28 components, the causal verdict, or downstream authority by carrying their refs.

A.10:5 - Archetypal Grounding

Runtime acceptance from a measurement result. C.16 dated measurement work obtains a pressure measurement result with uncertainty under a named model and calibration; a distinct C.2.1 episteme states it. If the claim that this Work first constituted that episteme is current, A.15.PROD recovers one local entity-inception claim. Separate evaluation work applies the declared G.4 pressure clause through A.6.1 bindings and obtains unknown; another C.2.1 episteme states that verdict. Later C.11 decision work uses the verdict episteme as a premise and defers.

A.10 can trace the source publications, calibration and measurement work, result episteme, evaluation work, clause declaration, bindings, provenance, and any currentness or live rival that affects reliance.

The supplied G.4 unknown and C.11 deferral are separate results, not an A.10 disposition. The bounded claim/use and the evidence basis needed by §4.5’s seven-value rule are not specified here, so the A.10 reliance classification remains unresolved. Select that claim and use and recover their required basis before classifying them; neither unknown nor deferral fills the gap. No ledger edge establishes measurement, verdict, decision, or use.

Complete the bounded classification. Continue the constructed case with claim C in Verdict-e7: “The evaluation of PressureClause-v3 for Batch-B17 returned unknown.” The intended use U is to report that clause verdict for Batch-B17. Stipulate that Verdict-e7 and the current clause are recoverable, the independently checked evaluation bindings select that batch and return unknown, their source and provenance relations hold, and the relevant window covers this use. No unresolved rival changes that reporting claim. These facts supply the missing basis; they are not inferred merely from the word unknown. A.10 returns RelianceDisposition=pass for reporting C within U. The pressure-acceptance verdict remains G.4 unknown, and the separately established C.11 deferral remains unchanged. A pass for reporting the verdict does not authorize acceptance of the batch.

Meta-analysis. Source study publications, datasets, analysis code, inclusion work, statistical method, and synthesis work are recovered by their direct relations. The application of the statistical method and the result relation establish the pooled estimate and uncertainty; the relied-on claim in the result’s C.2.1 episteme states that pooled estimate and uncertainty. A.10 records source identity, transformations, coverage, provenance, currentness, and the bounded clinical or policy use, not a generic validatedBy relation.

Credential display. A credential view can support credential currentness when the issuer or trust-root object and relation, verifier, relying use, and applicable verification facts are recoverable. Retain validity limits when present or required, status-source/register and revocation facts when the mechanism is present or the selected verification/use policy requires them, and holder binding when that claim or regime requires it. A bearer credential need not identify its holder, and a holder need not be the credential subject; neither case relaxes the selected policy.

Permission remains with A.2.8.PER, commitment with A.2.8, an issuing act with A.2.9, and a system-role-assignment occurrence with A.2.1. A status assertion requires its status pattern; an entry predicate requires its defining pattern and A.6.B boundary classification; gate passage remains with A.21. Display presence creates none of them.

A.10:6.1 - Bias-Annotation

A.10 corrects carrier-authority bias and graph-authority bias. A polished badge, attestation, dashboard, generated explanation, or provenance mark can make an unsupported claim look settled; a tidy graph can make an ungrounded edge look like an obtaining relation. The repair is to recover the claim, source, carrier, direct relations, bounded use, and disposition. Add Work and its local result, defining rule, and result episteme when asserted, and currentness or a live rival when it changes reliance.

A.10:6 - Conformance Checklist

  1. Claim: the selected claim or proposition and its C.2.1 episteme are named, together with a separate target claim when the evidence use distinguishes it. An asserted local result retains its defining or testing rule.
  2. Result rule: each asserted local result retains the rule that defines or tests it. Cite the applicable FPF pattern or domain source when that reference is needed. Use E.10.ROLE only for a remaining ambiguity about a technical role claim.
  3. Carrier/source: the selected source episteme and edition, any material publication occurrence, form, carrier, or face, the copy/transform chain, and direct provenance or citation relations are recoverable.
  4. Work: whenever production, interpretation, transformation, evaluation, or reliance is asserted as dated U.Work, point to its basis: the A.13 core including the obtaining assignment first, independent A.15.1 Work admission second, and F.6 afterward only for precise assignment-bound attribution. Add direct relations, A.6.1 bindings, and resource-use facts only when the receiving claim uses them. Ordinary source-finding action need not be admitted as U.Work.
  5. MethodDescription boundary: when an episteme is identified as a U.MethodDescription, it meets A.3.2’s positive membership rule. Its generic method claims, including intended effects, establish no actual participants, occurrence, use, proof/test event, or result.
  6. Result boundary: domain result, result episteme, carrier, provenance entry, outcome, and later action remain distinct.
  7. Graph boundary: every asserted edge names an independently established direct relation; no edge establishes work, participation, production, result, currentness, reliance, or representation by graph membership.
  8. Time/currentness: edition, window, supersession, revocation, source order, and G.11 result are explicit when they affect use.
  9. Reliance: bounded use and local RelianceDisposition are present. A live rival and stop/reopen condition are included when they affect reliance; any unsupported-use guard satisfies §4.4’s full F.19 test. B.3 opens only for a named assurance claim.
  10. Contest/privacy: the affected party can challenge the claim and disposition, while sensitive carrier access is minimized without erasing recoverability.

A.10:6.2 - Common Anti-Patterns and How to Avoid Them

  • Carrier as truth. Recover the claim and direct source relation; authenticity or availability is not truth.
  • MethodDescription replaced by a plan or run trace. Recover the substantive generic method claims, including any intended effect, separately from a particular plan, dated Work, and actual bindings.
  • Generic result field. Recover the domain result and its defining or testing rule; identify the result episteme separately when its identity matters.
  • Edge as fact. Establish the direct relation first; then let the graph represent or cite it.
  • Provenance as assurance or permission. Enter B.3, A.2.8.PER, A.21, or the pattern that defines or tests the additional claim only when that claim is live.
  • Citation as actual use. Ground the premise/reference/argument relation and any later Work claimed as such.
  • Full dossier by default. Collect only fields that decide the bounded use, consequence, contestability, and reopen condition.

A.10:7 - Consequences

Benefits. Reliance becomes replayable without turning A.10 into an authority over the results it cites. The same path can expose stale sources, hidden transformations, ungrounded work, incompatible currentness, or an unsupported lift from provenance to action.

Trade-offs. Subject-pattern recovery takes more effort than a single evidence edge. The gain is that later users can challenge the claim, work fact, source relation, currentness result, or reliance boundary that failed.

Failure containment. Missing source, work, direct binding, rule for a cited result, currentness, or use relation blocks or narrows only the affected reliance use. It does not authorize a universal evidence or result relation.

A.10:8 - Rationale

Evidence use is a relation-specific claim about why one later use may rely on one episteme. Provenance records make the source history recoverable; they do not create the source facts, local result, truth, work, or use. Keeping the descriptive graph outward-facing keeps each cited result with the rule that establishes it while still making complex source chains inspectable.

A.10:8.1 - SoTA-Echoing

These sources contribute distinctions to source-to-use recovery. Apply their relevant contracts to the selected claim and verification/use regime; identify FPF objects and direct relations independently.

SourcePractical lesson for A.10Limit on reliance
W3C PROV-O, Recommendation 30 April 2013Qualified provenance descriptions and stable source/activity/agent references help recover source, copy, and transform identity. In an A.10 path, every edge cites an independently established direct relation between the appropriate FPF objects.A PROV-shaped graph, wasGeneratedBy label, or qualified relation does not establish FPF Work, participation, result, truth, currentness, or later use.
W3C Verifiable Credentials Data Model v2.0, Recommendation 15 May 2025Distinguish issuer, represented subject, holder, verifier, validity, status, proof, and relying context. Recover the verifier rule and the currentness facts needed for the bounded use and its local disposition. Check validity limits when present or required, status when the mechanism is present or the selected policy requires it, and holder binding when the claim or verification regime requires it. Retain any stronger requirement of the selected policy.The holder need not be the subject; bearer credentials need not identify the holder. A conforming or cryptographically verifiable credential does not by itself create transitive trust, permission, a U.SystemRoleAssignment occurrence, gate passage, assurance, or truth of every represented claim.
SLSA specification v1.2 and in-toto Attestation Framework v1.2, with its Statement/v1 contractDistinguish artifact subject, predicate type, producing context, source inputs, authenticated envelope, verifier expectation, and versioned attestation. Recover the bounded build/source claim and the producing Work or System, verifier rule, holder, and window required by that claim or regime. Keep provenance separate from the reliance decision and exclude a locally plausible unsupported wider use.A signed attestation, SLSA level, or verification summary is not runtime safety, release approval, gate passage, assurance, or proof that an uncited Work/result relation obtains. SLSA and in-toto versions are separate; Statement/v1 alone does not identify an in-toto minor specification version.
C2PA Content Credentials Technical Specification 2.4, April 2026For claim-bound content attribution, distinguish asset/manifest identity, claim generator, assertions, ingredients/actions, signature validation, trust policy, and specification version. Recover the carrier, manifest/assertion, transformation, verifier/trust regime, edition, and currentness window relevant to that claim.A valid manifest, repository receipt, authenticity mark, or visible Content Credential does not establish truth of the represented world state, authorship beyond its assertion, permission, safety, or adequacy.
Mitchell et al., Model Cards for Model Reporting, FAT* 2019, and Gebru et al., Datasheets for Datasets, CACM 64(12), 2021Use intended-use, evaluation-condition, performance/limitation, motivation, composition, collection, and maintenance disclosures to find sources. For every operative claim relied on, recover its source, carrier, and bounded use, plus Work and the local result when asserted.A model card, datasheet, polished summary, or disclosed limitation is not evidence for an unstated claim, performed evaluation, assurance, approval, or deployment permission.

Credential, attestation, documentation and provenance content must be interpreted under the rule for the claim. Their format alone supplies no additional authority.

A.10:9 - Relations

  • Builds on: C.2.1 for claim and result epistemes; E.24.PUB for publication, form, and carrier relations and E.17 for multi-view publication; A.13 for every performer’s local core, A.15.1 for independent dated Work admission, and F.6 only for a current precise assignment-bound attribution; A.6.1 for actual operation bindings; A.15.PROD when entity inception through Work or production completion is current; and the selected A.2-family rule for a technical role claim, with E.10.ROLE only when its meaning remains unresolved.
  • Coordinates with: A.2.4 for first-use evidence/status classification; G.11 for currentness; C.29 for an unresolved mathematical-representation choice, transfer or reliance question; B.3 for assurance; C.16 for measurement; C.28 for causal use; A.19 for comparison/selection; G.4 for acceptance declarations and applications; C.11 and A.21 for option-set choice and gate results, respectively.
  • Constrains: provenance and reliance descriptions only. A.10 does not create another pattern’s result, occurrence, participation, representation, currentness, assurance, permission, commitment, gate, or decision.

A.10:10 - Older source text interpretation and neighboring-pattern notes

Treat legacy names such as manifest, creator, observer, symbol register, SCR, RSCR, MIC, verifiedBy, validatedBy, or evidence path as recovery prompts, not decisions about object kinds or current relations.

  • For a manifest or source register, recover what the claim selects: for example, claim-bearing content, a source register or service, a publication occurrence or form, a carrier, or a provenance description. Recover the source, edition, claim, and direct relations relevant to that use; the old name does not decide among these objects.
  • For a creator, observer, producer, verifier or maintainer claim, recover the System and the participation relation asserted. A recognizable ordinary actor designation needs no further expansion. If dated Work is asserted, recover every performer’s A.13 core and independently admit the Work under A.15.1. Add F.6 only for precise assignment-bound attribution, and other direct relations or A.6.1 bindings only when consumed. A label or assignment alone establishes neither participation nor authority or responsibility.
  • Keep an ordinary method-instantiation note as such. If it asserts performed Work, recover the Method, occurrence, enactment, ordering, participants and result that the claim consumes under their rules; the note’s presence alone establishes none of them.
  • A work result, measurement result, validation result, or verification result label requires the domain result and a separate C.2.1 episteme; the legacy field name establishes neither.
  • Read a resource roster’s claims about resources. Distinguish those claims from the resources and resource-use relations described, and from the carrier or provenance account through which the claims are encountered. Use C.2.1 or E.24.PUB when those distinctions change the use; an ordinary roster citation can remain ordinary.

When older text also claims approval, permission, gate passage, assurance, causality, comparability, representation, publication effect, or decision, use the pattern that defines or tests that additional claim and let A.10 retain only source recovery, provenance, bounded reliance, and contestability.

A.10:10a - Evidence carriers for quantum-like statements

Use A.10 when a quantum-like statement is being relied on. Name the minimal claim, selected source episteme and edition, any material publication occurrence, form, or carrier, bounded use, and RelianceDisposition. Include time/currentness, a live rival, and a locally plausible unsupported-use boundary when they change reliance. Add producing or interpreting dated U.Work, Method, actual bindings, and a Work-to-value predicate only when those facts are independently current. Use C.16 for a measurement claim, the relevant C.26 pattern for actual probe or frame effects, and C.29 only when a separate mathematical-lens choice, transfer or reliance question remains. Use F.9 when the claim needs a Bridge between senses in different contexts; then distinguish that relation from the judgment of its bounded use and tolerated loss. Use B.3 only for a named assurance claim. Consequential uses without one retain their direct safety, permission, gate, release, or other governing rule.

The quantum-like label has no evidence weight. A descriptive graph may represent the source and use relations only after those relations are established.

A.10:10b - C.29 mathematical-lens use relation

Recover the mathematical correspondence and application conditions on which the evidence use depends. Use an adequate domain result directly; use C.29 only when a separate representation, transfer or reliance question remains. A.10 may cite the resulting episteme and record its provenance, currentness, bounded reliance, and later use; an A.10 graph edge does not establish the correspondence. Use C.16 when measurement construction is the question and B.3 only when a named assurance claim is current.

A.10:End

Referenced in the corpus

1178 literal mentions in other sections. Read their context to establish the relation.