C.22 - Task Typing and TaskSignature Assignment (Problem-CHR)
Status: Stable Type: Calculus (C)
Purpose. Declare a minimal, portable TaskSignature for selector-facing use after the problem-side episteme is stable enough for Principles-to-Work, eligibility, acceptance, or policy-governed choice. C.22.2 carries the first problem-framing episteme for a messy signal.
Use C.22 to constitute one CHR-grounded U.Signature and, when a receiving use is current, relate the exact problem-side episteme to that signature through TaskSignatureAssignmentRelation. Typed characteristics and unknowns stay visible. The declaration includes scope and only those basis, currentness, evidence-use, or crossing relations on which the receiving use relies; none adds a generic setting, carrier, or organization as a participant.
Body-level kind boundary. TaskSignature is a C.2.1 episteme and a species of existing U.Signature, conformant to A.6.0 direct declaration fields, Vocabulary, Laws, and Applicability. Its declaration may use ordinary language when the required content is recoverable; it is not a record format and introduces no new root U-kind. TaskSignatureAssignmentRelation is a separate obtaining relation among one exact problem-side episteme, one exact TaskSignature episteme, and one exact use episteme. ProblemCard is the C.22.2 problem-side episteme used before that assignment. KindSet contains C.3 U.Kind values for selected entities. Descriptor maps, telemetry hooks, policy ids, and selector fields remain signature vocabulary or projections unless an exact admission predicate and current subject assertion establish another kind.
Primary EntityOfConcern. This pattern defines or constrains one TaskSignature episteme. Inside it, EntityOfConcernRef identifies the exact task or work target declared for the receiving use; it does not identify the signature, TaskKind, carrier, organization, or publication. TaskKind, optional TaskFamilyRef, KindSet, characteristic bindings, and scope relations are declaration content. A later SelectorOutcome remains a downstream result.
C.22:0 - Use This When
Use this pattern when one stabilized problem-side episteme must be related to a selector-facing TaskSignature for eligibility, acceptance, or policy-governed selection. Typical cases include solver choice, method-family eligibility, QD archive selection, open-ended generator selection, or specialization claims that need a declared task family or work target.
The working moment often sounds like this: “We are about to compare possible ways of doing, but which facts about this problem make a method family eligible, comparable, or unacceptable for this use?” Construct the smallest A.6.0 TaskSignature that a later selector can consume without selecting a method in advance, then assign it to the exact problem-side episteme and receiving use.
If problem framing remains contested or stale, use C.22.2. If a sufficient signature and assignment already exist and the current question is selection, use G.5. If a method is selected and dated enactment is being prepared, use A.15.2.
What goes wrong if missed. Unstated selector inputs let eligibility drift; incompatible scales are combined, unknowns are coerced and acceptance criteria are mistaken for characteristic definitions. A reuse that depends on semantic correspondence can also proceed without its local senses, obtaining F.9 relation and separate bounded-use claim.
What this buys. The downstream selection question gets one separately constituted TaskSignature with typed vocabulary, laws, applicability, unknown handling, and ClaimScope. Evidence, freshness, and cross-semantic relations are added only when that receiving use needs them. Its assignment is replayable while publication and serialization can vary without changing the signature.
C.22:1 - Intent
Make the task distinctions relevant to a selector decision explicit in a typed TaskSignature. A problem reaches C.22 when its problem-side episteme is stable enough to constitute and assign that declaration without selecting a method in advance. The signature is the smallest CHR-typed A.6.0 declaration sufficient for eligibility, acceptance, and policy-governed selection without inadmissible arithmetic or silent coercions.
C.22:1.1 - Term split used in this pattern
TaskSignatureassignment means one obtainingTaskSignatureAssignmentRelationamong an exact problem-side episteme, exact TaskSignature, and exact receiving-use episteme; it does not pre-bind a method.ScopeSlice(G)means the exact A.2.6U.ClaimScopevalue used by this declaration; it is not an evidence-path slice, baseline-set slice, container, or assignment participant.thresholdhas three distinct uses here:
Name and kind map for code-shaped heads. The names below identify different structural positions; capitalization does not make them peer kinds.
| Head used in this pattern | Recoverable kind or position | Direct governance boundary |
|---|---|---|
TaskSignature | C.2.1 episteme and species of U.Signature; this pattern’s primary EntityOfConcern | C.22 governs its A.6.0 direct fields, Vocabulary, Laws, and Applicability; C.2.1 governs constitution; E.17 supplies reader-facing publication forms, and E.24.PUB governs publication and carrier relations. |
ProblemSideEpistemeRef and ReceivingUseEpistemeRef | Participant designations in an assertion or description of TaskSignatureAssignmentRelation, not content or identity positions of TaskSignature | C.22.2 or the direct problem-side pattern defines or constrains the first episteme; the receiving-use episteme states the use but does not prove that an assignment obtains. |
TaskKind | TaskSignature position filled by one exact C.3 U.Kind value that types the current task or work target | C.3 governs the kind value; the field does not introduce U.Task. |
TaskFamilyRef | Optional reference position for the comparison-relevant task family | C.22 and C.22.1 govern task-family anchoring; the reference is not the family or a selected method. |
ProblemProfile | C.2.1-conformant U.Episteme that describes the stabilized problem and may reference the TaskSignature assignment | TaskSignature membership requires its A.6.0 declaration content; the assignment is a separate obtaining relation. One episteme may serve both descriptive uses only when their content and C.2.1 constitution conditions hold. |
ScopeSlice(G) | Local position whose filler is the exact A.2.6 U.ClaimScope value that bounds claims about the exact EntityOfConcernRef | A.2.6 governs membership; the position is not an E.18 path slice or a new slice kind. |
CHR field heads in 5.1 | TaskSignature positions filled by characteristics, scales, units, polarity values, scope values, evidence relations, and currentness conditions | C.16 and each subject-pattern locator identify the exact definitions and constraints for the fillers; C.22 states why the selector-facing use needs them. |
QD and OEE extension heads in 5.1 | Optional TaskSignature positions filled by exact characteristic-space, archive, policy, telemetry, generator-family, validity-region, and transfer-rule values or references | C.18, C.19, G.5, G.11, and the named direct patterns keep authority over those fillers. ArchiveConfig, TelemetryHooks, and GeneratorIntent do not become root kinds here. |
C.22:1.2 - ProblemCard relation
ProblemCard is the C.22.2 C.2.1 episteme used to stabilize one problem-side representation before downstream Principles-to-Work.
A ProblemCard can prepare TaskKind, scope, and characteristic bindings for a candidate TaskSignature. Identify which adequate signature was adopted for this problem-side episteme and receiving use under §5.2. Other adequate descriptions may be available. If adequacy or adoption is unresolved, retain the candidate signatures and identify what would settle it.
Assigning a TaskSignature does not transfer the card’s claims into it. The signature keeps only its A.6.0 declaration content; the card remains the reviewable problem-side episteme that explains why this problem can proceed to characterization, comparison, search, refresh, retirement, or another subject pattern.
The corresponding claims remain with their named subject patterns.
C.22:2 - Problem Frame
Selector-facing problem case
For selector-facing C.22 use, a problem case applies when the problem-side episteme is stable enough to construct a minimal TaskSignature and assert its TaskSignatureAssignmentRelation for eligibility, acceptance, or policy-governed selection. Method absence or contestability is a common reason for this downstream use; it does not define what counts as a problem.
When the live question remains a symptom, contested framing, stale ReferenceScheme or ClaimScope, set-derived candidate, opportunity cue, or preselected work item, use C.22.2 before asserting one assignment. When selection becomes current, cite the A.19.SelectorMechanism relation and exact G.5 policy refs rather than moving selection policy into the signature.
Unknown-first discipline. In the TaskSignature, preserve a live unrecovered trait as unknown when its value rule permits it. Name the exact downstream policy that interprets a live unknown for the receiving use. C.22 introduces no universal outcome set; C.23, G.4, G.5, or another direct pattern defines or constrains the resulting eligibility, acceptance, or selection disposition.
When the receiving selector cannot recover the relevant task distinctions, it cannot filter or abstain on a justified basis. Acceptance criteria may leak into scoring, or cross-semantic reuse may proceed by a shared name instead of an obtaining relation.
The receiving use needs a use-bounded TaskSignature that:
- establishes MM-CHR admissibility for Scale and applicable Unit and polarity before aggregation;
- carries tri-state unknowns explicitly; and
- states the exact scheme, scope, and only the basis and currentness relations needed by the current use.
Keep the direct support required for the claim; open A.10 when its evidence, provenance or bounded reliance needs a replayable account.
Open B.3 only when a named assurance claim about an exact target claim is current for this receiving use, including when a material-reliance rule requires that assurance claim; its assurance result is not a TaskSignature field.
C.22:3 - Problem
Without recoverable descriptor meanings, Eligibility and Acceptance can depend on inadmissible operations, such as ordinal means or unconverted mixed units. Without the F.9 split, a cross-scheme comparison can also equate unlike local meanings or import an assurance penalty even though no current assurance policy selected it.
C.22:4 - Forces
| Force | Tension |
|---|---|
| Parsimony vs sufficiency | Fewer fields to limit declaration work vs enough to drive admissible gating. |
| Unknowns | Many traits are unknown in the initial problem record → tri-state semantics propagate to Acceptance without silent coercions. |
| CHR admissibility | No mean on ordinals; no unit mixing; aggregation is admissible only after polarity and scale type are declared. |
| Locality vs portability | The declaration is use-bounded. Cross-semantic reuse first resolves two exact local senses and tests whether an F.9 Bridge obtains; the proposed use and any evidence reliance or assurance remain separate claims. |
C.22:5 - Solution — Problem CHR, TaskSignature, and assignment relation
Local TaskSignature mantra. Stabilize the problem; name the receiving selection question and task kind; keep only traits that can change eligibility, acceptance, or selection; type each live trait; preserve unknowns, scope, and any basis or currentness relation the use relies on; declare the TaskSignature, assign it to the problem-side episteme for that use, and stop before selecting a method.
Apply that formula as follows:
- Confirm that the problem-side representation is stable enough for selector-facing use; otherwise use
C.22.2. - Name the receiving eligibility, acceptance, or selection question and the
TaskKind, optional task family, or work target that the signature will declare. - Read the eligibility conditions, acceptance clauses or comparison criteria of the receiving use. For each condition, identify the task-side fact or characteristic whose value the condition consumes, and recover it from the problem-side account or its source. Include that trait and its interpretation; leave unrelated optional extensions absent. If two feasible tasks fit the same description but require different eligibility or acceptance decisions, identify the characteristic that accounts for the difference. When a method must be chosen before work starts, establish what the chooser can learn by then. A supported conditional choice or a declared rule for an unknown value can suffice. Keep an unresolved interpretation visible.
- Identify each live Characteristic and Scale, its Unit and polarity when applicable, and the admitted comparison relation. Declare ReferencePlane when interpretation or the receiving comparison contract requires it. Apply these conditions before aggregation or comparison.
- Preserve a live but unknown value as
unknown; include or reference the exact scope, evidence relation, freshness or edition condition on which later use relies. When cross-semantic reuse is current, resolve the two local senses and test the F.9 Bridge separately; a shared label, scheme, or plane does not establish it. - Close with one minimal
TaskSignature. Pass later eligibility and acceptance claims toC.23andG.4, and method-family selection toG.5; do not put their outcomes back into the signature as if they were problem traits.
C.22:5.0a - Positive closure, bounded non-use, and local return
Close the C.22 use positively when the direct TaskSignature fields, Vocabulary, Laws, and Applicability are complete and the assignment obtains under C.22:5.2. Its assertion identifies the exact problem-side episteme, TaskSignature, receiving-use episteme, effective ReferenceScheme, ClaimScope, and current qualification conditions.
Each live characteristic has its Scale, applicable Unit and polarity, required ReferencePlane, admitted comparison relation, and value or admitted unknown; every relied-on evidence, freshness or edition relation is named. When cross-semantic reuse is current, its exact local senses, obtaining F.9 Bridge, and separate bounded-use claim are recoverable.
A downstream selector can now consume the assigned signature without guessing, but no eligibility verdict, acceptance result, method recommendation, selector outcome, WorkPlan, or dated Work is claimed.
Close by bounded non-use when problem framing is not stable enough for a TaskSignature declaration, when no selector-facing receiving use is current, or when the current question has already become eligibility, acceptance, selection, planning, or performed work. A non-current optional extension remains absent. Report a particular assignment when the evidence establishes adoption of that adequate signature under §5.2. Keep unresolved alternatives as candidates.
Return to the smallest affected TaskSignature position when its receiving question, exact target, effective ReferenceScheme, ClaimScope, TaskKind, task-family reference, characteristic meaning, scale, unit, polarity, reference plane, unknown status, evidence-use relation, freshness condition, or edition changes. Recheck the F.9 endpoint senses, Bridge predicate, bounded-use claim, or reliance result only when that exact dependency changed. Keep the upstream ProblemCard and downstream selection history unchanged unless that exact change invalidates them under their own subject patterns.
Worked local repair. A machining TaskSignature originally records surface finish as an ordinal visual grade. The named use later adopts measured roughness Ra on a ratio scale in micrometres with a named measurement and evidence relation. Repair the affected characteristic head, scale, unit, admitted comparisons, and evidence relation. Keep the machining TaskKind, unaffected constraints, scope, and prior Work history. Reopen eligibility, acceptance, or Method-family selection only when its earlier result relied on the replaced finish head; state the new result under the exact downstream predicate with its subject-pattern locator.
C.22:5.0b - Apparatus proportionality
Use the lightest signature declaration and assignment relation that the named receiving use can consume:
- Minimal selector-facing use. Materialize one TaskSignature with only the live fields needed by the current eligibility, acceptance, or selection question. This is the ordinary positive result of C.22.
- Reliance-bearing use. Add an addressable
ProblemProfileepisteme only when delayed feedback, audit, transfer, automation, expensive reversal, or another named use relies on replay beyond the local assignment relation. Pin the exact problem-side episteme and edition, TaskSignature edition, receiving use, every relied-on field-basis relation with its subject pattern, qualification window, review trigger, and any current evidence or currentness relation. When this use crosses local meanings, test F.9. Add a Bridge and separate bounded-use claim only if its predicate is true; otherwise keep the local values separate and stop that reuse. - Extension-bearing use. Add QD, OEE, archive, generator, parity, or specialization positions only when that exact downstream relation is current and its direct pattern requires those values.
If no selector-facing receiving use needs a TaskSignature, finish without creating one.
C.22:5.1 - Minimal CHR fields (tri‑state aware).
Selector-side field boundary. The fields below are live only after problem framing has been stabilized enough to ask eligibility, acceptance, selection, Method-family, or policy-constrained choice questions. They are not a universal problem-framing checklist and do not replace the C.22.2 Thin ProblemCard pass for a messy signal. Each live characteristic field is CHR-typed by its Characteristic and Scale, with Unit and polarity where applicable under MM-CHR discipline. A live predicate may preserve unknown only when its exact value rule permits it; the cited downstream policy states what follows. This aligns G.4 and G.6 without making their results C.22 values.
Optional extension absence rule. If QD, OEE, archive, generator, parity, specialization, or another optional relation is not live for the current case, the corresponding optional fields are absent, not unknown. Use unknown only for a live field whose value is currently unknown. An absent non-live extension triggers no downstream disposition.
-
DataShape— data regime and admissible transforms (e.g., tabular, sequence, graph; density; stationarity claims). -
NoiseModel— uncertainty class and robustness envelope (e.g., iid Gaussian; heavy‑tailed; adversarial budget). -
ObjectiveProfile— objective heads (Scale, applicable Unit and polarity, and ReferencePlane when interpretation or the receiving contract requires it), admissible order relations (lexicographic, Pareto), and medoid or median operations where admissible. Weighted sums across mixed scale types are inadmissible; ordinal heads use order-only guards. For QD tasks, enumerate quality heads, diversity or descriptor-space heads, and any policy-authorized QD contribution heads; see DominanceRegime below. Do not introduce a default QD score. If a scalar or set-scalarization policy is live, cite the governing CAL policy and keep the uses of dominance and telemetry explicit. -
RegularityTraits— method-relevant structure (convexity, differentiability, separability, monotonicity) as CHR-typed predicates with guard macros (for example,ORD_COMPARE_ONLY,UNIT_CHECK,POLARITY_CHECK). IncludeConditionClasssuch as stiffness or kappa proxies where applicable. -
Constraints— explicit hard and soft constraint classes (feasibility predicates; ResourceEnvelope and RiskEnvelope). Declare acceptance-gate thresholds in G.4 AcceptanceClauses. A signature or implementation may reference those clauses for its use; it does not redefine them as Characteristic properties or hidden constants. -
ShiftClassand stationarity — CHR‑typed claims about regime stability (iid | covariate‑shift | concept‑drift | adversarial). When this field is live but unrecovered, its value isunknown. The cited acceptance or selector policy governs the consequence; omit the field when the receiving use does not need it. -
Evidence and assurance (conditional). Cite the direct evidence-use, provenance or subject relation that supports the receiving claim. Use A.2.4 when an episteme’s evidence-use or status-use relation needs classification, and A.10 when evidence, provenance or bounded reliance needs a replayable account. State source edition, currentness, or freshness only to the degree that reliance requires.
Open B.3 only for a named assurance claim about an exact target claim and receiving use, including when a material-reliance rule requires it, and use only the assurance lanes and fold that its declared policy requires. A TaskSignature by itself requires neither all TA/VA/LA lanes nor a Gamma-fold.
-
ScopeSlice(G)— the A.2.6U.ClaimScopevalue that bounds this declaration’s claims aboutEntityOfConcernRef(discipline governance in CG‑Spec; Domain is a catalog mark only). -
SizeAndConditionProfile— size and condition proxies (n, m, kappa, sparsity) with units declared where applicable; a unit mismatch makes the current comparison unsupported until the direct acceptance or selector policy supplies its governed result. -
Freshness(conditional) — the validity window for a descriptor only when the receiving use relies on its currentness. -
Missingness— MCAR, MAR, or MNAR (or mapped equivalents) per CHR.Missingness; Acceptance and flow use preserve the declared missingness semantics. -
KindSet— selected C.3U.Kindvalues for the entities addressed by the TaskKind; separates EntityOfConcern kind from Scope (USM).
QD and Illumination extensions (normative; ties to C.18 and C.19).
Select these fields from the receiving use. A QD search constructs and maintains diverse candidates under its descriptor, archive and emitter rules; that use activates the QD fields below. Retaining exploration evidence under PortfolioMode=Archive, comparing a front, or returning several alternatives does not by itself require QD. An open-ended generator use activates GeneratorIntent and its own required fields; add the QD fields when that use also performs QD. Keep the source collection, its retention policy, the comparison and the emitted selector outcome distinct under G.5:4.4b–4.4e.
CharacteristicSpaceRef— reference to the exactU.CharacteristicSpace, with its dimension and CHR-typed coordinates as required by the declared descriptor and comparison rules. Declare ReferencePlane where coordinate interpretation or the receiving contract requires it. Pin its defining episteme separately withcharacteristicSpaceDescriptionRef: U.EpistemeRefand resolve both to the same space under A.19:5.1.1.1.ArchiveConfig— archive topology (grid, CVT, or graph), resolution (bins or centroids), K‑capacity,InsertionPolicyRef(elite replacement, dedup, or novelty), andDistanceDefRef.edition(declare metric or pseudometric status and invariances; normalisation is admissible only when the applied scale transform is admitted by CG-Spec); admissibility follows CG‑Spec.EmitterPolicyRef— reference to the emitter policy governed by C.19 and applicable to this TaskSignature; edition id recorded.DominanceRegime—{ParetoOnly | ParetoPlusIllumination}. Default =ParetoOnly(illumination remains report‑only telemetry unless CAL explicitly authorisesParetoPlusIllumination, policy‑id cited).IlluminationSummary— a telemetry summary overDiversity_P; reported by default; excluded from dominance unless a CAL enablesParetoPlusIllumination(policy‑id cited).IlluminationMap(parity-run) — parity-run publication is complete when an IlluminationMap publication (grid, CVT, or graph perArchiveConfig) records coverage per niche or cell withDescriptorMapRefandDistanceDefRef.edition. A single-score leaderboard does not satisfy this comparison use; compare under the declared CG-frame.PortfolioMode— resolve its values and default under G.5CC-G5.23.Archivepreserves exploration evidence. It neither activates a QD search nor names the emitted selector result; G.5:4.4e distinguishes retention, source set and outcome.Budgeting— evaluation, time, and batch budgets, including E/E‑LOG exploration budget id; units declared (CG‑Spec).TelemetryHooks—PathSliceIdonly when an E.18 path-slice reference is current, plus decay and refresh policy ids, edition counters, descriptor-map updates, and policy-id updates upon illumination gains.GeneratorIntent(OEE) — optional intent to use a registeredGeneratorFamily(G.5), with pointers toEnvironmentValidityRegion,TransferRulesRef, and coverage and regret reporting expectations.
Admissibility. Before any numeric comparison or aggregation, establish CSLC admissibility for Scale and the applicable Unit and polarity, and cite CG-Spec.Characteristics. Record ReferencePlane when its interpretation or the receiving contract requires it. Preserve unknown for the downstream policy; do not coerce it to 0 or false, and do not invent a C.22-local disposition.
C.22:5.2 - TaskSignature declaration and assignment
TaskSignature is a C.2.1 episteme and a species of A.6.0 U.Signature. It uses A.6.0 identity and declaration content directly rather than a flat record schema. Add an optional SignatureManifest only when dependency replay requires imports and provided names; SignatureId and edition are designators and currentness handles, not substitutes for identity.
TaskSignature <: U.Signature
EntityOfConcernRef = exact task or work target declared for the named receiving use
effectiveReferenceScheme = exact scheme in which TaskKind, TaskFamilyRef?, KindSet, and characteristic values are interpreted
SubjectKind = TaskKind, one exact C.3 U.Kind value
RangedValueKind = U.Entity
SliceSet = declared A.2.6 claim-scope slices
ExtentRule = entities admitted by KindSet inside those slices
ResultKind = absent; selector outcomes are not TaskSignature values
Vocabulary:
TaskFamilyRef?
KindSet
characteristic bindings with Scale, applicable Unit and Polarity, required ReferencePlane, admitted comparison relation, and value or admitted unknown
constraint relation references
evidence-use relation references only when the receiving use relies on them
optional QD, OEE, archive, generator, parity, budget, telemetry, and specialization vocabulary only when current
Laws:
include only positions that can change eligibility, acceptance, or selection for the declared use
preserve admitted unknown and distinguish it from absent non-current vocabulary
apply CHR scale, unit, polarity, ReferencePlane, and comparison legality before aggregation
keep acceptance verdicts, selector outcomes, selected methods, plans, Work, and performed results outside the signature
keep each reliance-bearing field connected to its exact basis relation and subject pattern
Applicability:
exact U.ClaimScope and any required A.2.6 membership
declared qualification or use window when current
qualification, freshness, edition, and evidence-use conditions on which use relies
exact F.17 local senses, an obtaining F.9 Bridge, and a separate bounded-use claim only when cross-semantic reuse is current
The field families in C.22:5.1 are projections of Vocabulary and Applicability. They are not extra conceptual rows and do not redefine A.6.0.
The assignment is a separate relation with exactly three direct participants:
TaskSignatureAssignmentRelation <: U.Relation
ProblemSideEpistemeSlot = <ProblemSideEpistemeSlot, U.Episteme, U.EpistemeRef>
TaskSignatureSlot = <TaskSignatureSlot, U.Signature, U.EntityRef constrained to TaskSignature>
ReceivingUseEpistemeSlot = <ReceivingUseEpistemeSlot, U.Episteme, U.EpistemeRef>
The relation obtains while that exact TaskSignature is adopted as the task-typing declaration for the exact problem-side episteme and the use stated in the exact receiving-use episteme, under the stated scheme, scope, and qualification conditions. Co-publication, a card field, a shared label, or one record row does not make it obtain.
One occurrence is identified by the three participants plus its maximal continuous actual assignment extent. A participant change yields another occurrence; actual withdrawal and later readoption yield distinct occurrences even when the same three participants return.
TaskSignature identity and publication. The tuple <declaration content, EntityOfConcernRef, effectiveReferenceScheme> determines TaskSignature episteme identity under A.6.0 and C.2.1. SignatureId and edition designate and track that episteme. A semantic change to direct declaration fields, Vocabulary, Laws, Applicability, the exact target, or the effective scheme identifies another episteme. Admit it as a TaskSignature only if it satisfies A.6.0; relate it as a revised signature edition only when C.2.1:4.5’s source-use and continuation conditions obtain.
Two E.17 publications, database rows, cards, or files may present the same edition when they resolve to the same tuple and add no new claim. A ProblemProfile may cite the signature and assignment or include the declaration content. Test TaskSignature membership and C.2.1 identity from that content; the assignment relation remains separate.
Minimality rule. Include only declaration positions needed to determine eligibility, acceptance, or admissible selection for the named use. Additional traits remain outside Vocabulary until a later use makes them current.
Values are CHR-typed and tied to the exact measurement, evidence-use, source-use, representation, or scope relation that justifies their use when such a relation is current. Each reliance-bearing field basis names that relation and its subject pattern; generic provenance or support wording is not a replay basis. Unknowns preserve their direct missingness semantics.
TaskSignature invariants. A positive assignment satisfies all six conditions:
- The TaskSignature exposes its exact
EntityOfConcernRef, effective reference scheme, direct declaration fields, Vocabulary, Laws, and Applicability. - The exact problem-side episteme, exact TaskSignature, exact receiving-use episteme, obtaining conditions, and occurrence extent of the assignment relation are recoverable.
- Every live field has an admitted filler kind or scale discipline and, under reliance, an exact basis relation with a subject pattern.
- A live but unrecovered value is
unknownonly where the field’s exact value rule permits it and a downstream policy states how the named use handles it. - A non-current optional extension is absent; absence and unknown are not interchangeable.
- Eligibility verdicts, acceptance results, selected methods, selector outcomes, WorkPlans, and Work occurrences are absent from the TaskSignature and remain with their direct patterns.
C.22:5.2a - Lowering and withdrawal conditions
Stop relying on the assignment for the current use when its problem-side episteme, TaskSignature, receiving-use episteme, scheme, scope or qualification conditions cannot be recovered. Actual withdrawal of adoption or failure of an obtaining condition ends the occurrence under §5.2; missing access to its account does not by itself rewrite the assignment history. The TaskSignature may remain valid for another assignment. Use C.22.2 only when the problem-side representation itself is no longer stable enough.
Identify the resulting episteme and apply C.22:5.2’s signature-membership and edition-continuity checks when a direct declaration field, Vocabulary, Law, Applicability claim, EntityOfConcernRef, or effective reference scheme changes. Lower or remove one vocabulary position when its filler kind, scale, unit, polarity, reference plane, direct basis relation, or subject pattern cannot support the claimed use. Preserve unknown only when the position remains live and admitted. Split any selected method, selector outcome, acceptance result, plan, or Work occurrence into its subject pattern.
A changed or invalid signature position reopens an earlier downstream result only when that result relied on the changed position. The downstream pattern repairs or supersedes its own result. A revised signature does not imply that the actual Problem disappeared or that prior Work did not occur.
C.22:5.2b - Evolution and currentness boundaries
Use C.22 to revise the smallest affected identity or declaration-content component; identify the resulting episteme and claim a new TaskSignature edition only under the identity, membership, and continuity conditions in C.22:5.2. A changed problem formulation requires C.22.2 before a replacement assignment is made.
G.11governs relied-on source edition, freshness, decay, telemetry, and currentness relations; its result may trigger signature review but does not rewrite the signature by itself.C.18andC.19govern archive, front, lineage, and live-pool evolution.G.5governs selected-set and method-family selector results.E.23governs repeated object-version improvement.
C.22 introduces no local refresh object and does not rewrite earlier selector results or dated Work without an explicit dependency.
TaskKind fills SubjectKind. TaskFamilyRef? names one comparison-relevant family in Vocabulary when specialization, transfer, or parity is live. KindSet and A.2.6 scope slices determine the ranged extent. They remain declaration content, distinct from a selected method or selector result.
DesignRunTag hygiene. Do not mix DesignRunTag positions in one signature edition. If design-side information is reused in run-time Work, identify the actual receiving Work and its relations independently. Cite an E.18 structural GateCrossing only when a selected transformation-flow use contains that occurrence; it does not require a package and does not establish an F.9 Bridge.
C.22:5.2.1 - Specialization-claim reference discipline (normative)
A specialization claim identifies the holder, dyad, team, or scoped portfolio, the task family or TaskSignature, the work target, and the criterion and grounds for usable performance. Identify what the holder did and what assistance was available during the performance. A claim of acquisition needs evidence of change from the relevant prior state under the support conditions of the intended work. When the receiving use compares acquisition or adaptation, C.22.1 supplies the adaptation budget, prior-exposure and threshold-crossing conditions; retain freshness and provenance to the extent reuse depends on them. A method may be selected, refined, or retired in this process, while the specialization claim concerns the stated holder or portfolio. Preserve the task-family and work-target references required by the C.22.1, G.5 or G.9 receiving use.
Low-human-overlap or newly discovered task families remain admissible when those task-family or signature references are explicit by value.
When a low-human-overlap comparison matters to the receiving use, identify the target task family and a bounded reference set of documented human practice. Define the task correspondence through the inputs, objectives, constraints and conditions relevant to that use. State the direction, weighting or coverage rule and the local threshold meant by “low”. These choices make the comparison reproducible; an ordinary task-family declaration needs no such comparison.
For example, suppose one of five equally weighted target task classes has a match in a completely examined finite reference set. Directional coverage is 1/5, below a declared threshold of 1/4. If one class is matched, two are known not to match and two remain unclassified, possible coverage instead ranges from 1/5 to 3/5, so that evidence cannot settle the same low-coverage claim. The result concerns this reference set and task correspondence. Task admissibility and the holder’s specialization retain their own tests.
C.22:5.3 - Provenance, schemes, and planes
Record the effective reference scheme, U.ClaimScope, and any ReferencePlane needed to interpret a relied-on value. A difference in scheme or plane opens a comparison question; it does not by itself establish a Bridge, forbid use, or impose a penalty.
For cross-semantic reuse, recover the two exact F.17 local senses and test the direct F.9 predicate. Cite a Bridge only when that predicate is true. Then state the proposed use separately: the action, direction, correspondence rule, tolerated loss, and polarity. If no Bridge obtains, keep the local values separate and return the missing comparison or translation question instead of asserting correspondence.
Keep the grounds required for that bounded use, and open A.10 when its evidence, provenance or reliance needs a replayable account.
Open B.3 only for a named assurance claim about an exact target claim and bounded use, including when a material-reliance rule requires it; only that B.3 result may use edge-scoped CL and the policy it declares. An optional local F.9 CL note is evidence shorthand, not a use threshold or automatic penalty.
A card, gate check, reusable package, or publication is required only when its own receiving pattern independently needs it. The assignment receives no generic setting participant, and a domain, organization, location label, or shared carrier supplies none of these relations.
C.22:5.4 - Attachment & use.
The bullets below state which TaskSignature fields and relations each downstream use reads. C.22 does not execute eligibility, acceptance, selection, archive treatment, or generator-family choice. Their verdicts and returned sets remain results of the named direct patterns.
- Eligibility gates read TaskSignature against each MethodFamily.Eligibility (C.23) and CG‑Spec.MinimalEvidence for referenced characteristics.
- Acceptance clauses (G.4) use these fields for acceptance-gate threshold predicates (acceptance-gate thresholds live in Acceptance only).
- Selection kernel (G.5.S3) applies an admissible order (often partial); weighted sums across mixed scale types are inadmissible. A non-dominated front or retained archive supplies candidates. G.5:4.4b–4.4e governs the emitted outcome: retained alternatives use
ShortlistorRankedShortlist, while joint inclusion and other outcomes keep their own conditions. Preserve the source reference, comparison basis and any unresolved ties.PortfolioMode=Archivechanges evidence retention, not that result family. In an active QD use, illumination enters dominance only when CAL authorisesDominanceRegime=ParetoPlusIlluminationwith its policy id; otherwise the telemetry is reported separately. - When
GeneratorIntentis present, G.5-governed selection may use a registeredGeneratorFamily(POET-class); the selection domain becomes pairs{environment, method}, with Environment guarded byEnvironmentValidityRegionandTransferRulesRef(C.23 wiring). When the active branch or receiving use requiresIlluminationSummary, obtain that retained-set report under C.17:5.1, including itsDiversity_Pmeasurement policy, Scale, descriptor or feature basis, and distance or covering definition. Generation alone requires no such report. Illumination remains report-only unless its governing policy changes dominance participation as above.
C.22:5.5 - Unknowns.
An identity position needed for positive closure cannot be replaced by unknown. A live characteristic or predicate may preserve unknown only when its exact value rule permits it. The TaskSignature cites the downstream policy that defines or constrains the consequence; C.22 performs no implicit coercion and declares no universal outcome set.
C.22:5.6 - Publication.
When a named receiving use needs a reader-facing publication of the ProblemProfile specified in C.22:5.0b, publish that accepted C.2.1-conformant episteme through E.17. Use E.24.PUB when that use depends on publication availability, its declared boundary, or publication-occurrence identity. The profile references the bound TaskSignature and only the evidence, currentness, F.9 bounded-use, and representation relations on which that use relies.
Apply F.18 and F.17 Name Cards when a durable new name is being admitted; do not create a card merely because a local field or Bridge is present. Keep vendor or tool examples in Plain explanatory use rather than letting them become normative selector inputs.
When no replay beyond the local assignment relation is needed, the TaskSignature closes without a separate ProblemProfile.
C.22:5.7 - Open‑Ended tasks (GeneratorFamily) (normative).
When open-ended generation of tasks or environments is current, the TaskSignature supports that use only when it includes GeneratorIntent with pointers to EnvironmentValidityRegion (admissible region for generated environments), TransferRulesRef (cross-environment transfer constraints), and coverage and regret telemetry expectations. G.5 declares its outcome over {environment, method} pairs under G.5:4.4b–4.4e; the generating archive remains a source.
Report coverage and regret under their declared policies. Add IlluminationSummary when the active branch or receiving use requires that retained-set report, with the measurement basis specified in C.17:5.1. These telemetry results remain excluded from dominance unless a CAL policy promotes them (policy id recorded in SCR; see DominanceRegime).
The SCR change record retains current editions and actual edition increments of the definitions on which the active generator or telemetry use depends. These include TransferRulesRef.edition and any characteristic-space description, descriptor map, distance or covering definition, and emitter policy used by that branch. An illumination increase retains its governing policy id; a generator use without illumination requires no illumination record.
C.22:6 - Archetypal Grounding (Tell–Show–Show)
The examples show which TaskSignature fields supply eligibility or acceptance inputs. Keep that dependency recoverable when a downstream decision relies on it—for example, an ordinal service level permits order comparison, while budget arithmetic requires aligned units. The compact field projections below illustrate declaration content; §5.2 governs the complete signature and assignment.
A. Differential equations (continuous systems, solver choice).
TaskSignature fields. DataShape=ODE, stiff?=unknown, SizeAndConditionProfile={n≈10^3}, ObjectiveProfile={↓error@ratio, ↑throughput@ratio}, ConstraintRefs={budget-envelope relation, safety-predicate relation}, RegularityTraits={Lipschitz known?=unknown, Jacobian sparsity=high}, Missingness=MAR.
Attachment. Selector consumes TaskSignature; eligibility filters MethodFamilies whose eligibility conditions require known stiffness or differentiability, with unknown yielding degrade or abstain per family. Acceptance tests its safety_gate condition on the declared ordinal safety level using order comparison (ORD_COMPARE_ONLY), and uses unit-aligned sums for ratio-scale budgets. The selector declares a Shortlist retaining the non-dominated candidates and its comparison basis; no cross-ordinal weighting.
B. Mixed‑integer optimisation (planning and scheduling).
TaskSignature fields. DataShape=MIP, NoiseModel=deterministic, ObjectiveProfile={↓cost@ratio, ↑service_level@ordinal}, Constraints={SLA hard, workforce soft}, RegularityTraits={convex_relaxation=available}, SizeAndConditionProfile={vars~10^5}, Missingness=MCAR.
Attachment. CG‑Spec forbids means over service_level (ordinal); Acceptance holds acceptance-gate thresholds; Eligibility checks convex-relaxation availability; Selection applies the lexicographic guard (assumption-fit before evidence-fit before resource). If this receiving use makes a named assurance claim about an exact target claim, including when a material-reliance rule requires it, use B.3 to obtain its separate assurance result from the relied-on evidence relations and the declared policy; otherwise this example adds no assurance fold. If the admissible comparison remains partial, the selector can declare an unordered Shortlist from the non-dominated candidates, with the comparison basis and unresolved ties.
Practice example: the cited SciML Problem Interface constructs an immutable problem value before solver use and supports explicit
remakewhen problem fields change. C.22 adapts only that problem-before-selector separation; it does not import Julia types as FPF ontology.
C. Quality-Diversity archive and declared set (illumination).
TaskSignature fields. DataShape=policy‑search; ObjectiveProfile={↑reward@ratio, ↑coverage@ratio (report‑only)}, DominanceRegime=ParetoOnly, PortfolioMode=Archive, CharacteristicSpaceRef(d=3, characteristics=CHR‑typed), characteristicSpaceDescriptionRef=CS-description-1, ArchiveConfig(grid, res=32×32×16, K=1, InsertionPolicyRef=elite‑replace, DistanceDefRef.edition=v1), EmitterPolicyRef=v2, Budgeting{eval=1e6}, TelemetryHooks{PathSliceId=…}. Include PathSliceId only when an E.18 path slice is current.
Selection result. The QD archive remains the declared source. G.5 returns a Shortlist of retained alternatives from it, with exact members, ordering and basis pins under G.5:4.4b–4.4e. Coverage and illumination are reported separately and excluded from dominance under the declared ParetoOnly policy. Changes to DistanceDefRef.edition or the emitter policy are editioned and logged in SCR.
D. Open‑ended environment generation (POET‑class).
TaskSignature fields. GeneratorIntent{GeneratorFamilyRef=…, EnvironmentValidityRegion=… (CHR‑typed), TransferRulesRef=…, CoverageMetric=…}, PortfolioMode=Archive.
Selection result. G.5 declares a Shortlist over the eligible {environment, method} pairs and cites the source, ordering and selection basis. TransferRules govern cross-environment policy reuse. This use reports its declared coverage and regret with their edition and policy basis, separately from the selector outcome. It requires neither a QD search nor an illumination report; a later receiving use that needs illumination obtains its measurement basis under C.17:5.1.
E. Physical manufacturing method-family eligibility.
Problem-side episteme. PartFamilyFinishingProblemCard-E2 : U.Episteme is the exact C.22.2 ProblemCard for a shop that must finish AlloyPartFamily-17 on one machine under ShopInspectionScheme-E4 and a production-window ClaimScope. The receiving question is which available finishing-method families can be compared without presuming one of them.
TaskSignature. SurfaceFinishingEligibilitySignature-E1 declares EntityOfConcernRef=AlloyPartFamilyFinishingTarget-17, effectiveReferenceScheme=ShopFinishing-Scheme-A, TaskKind=surface-finishing work, ScopeSlice(G)=AlloyPartFamily-17 during [2026-09-01T00:00Z, 2026-10-01T00:00Z), ObjectiveProfile={surface roughness Ra@ratio in micrometres with downward polarity, throughput@ratio}, ConstraintRefs={geometric-tolerance relation, heat-distortion relation, resource-envelope relation}, and material hardness as a live unknown, with the rule for obtaining the needed measurement and the policy for handling that unknown in eligibility.
The TaskSignature makes eligibility reviewable; it does not select grinding, honing, polishing, or another method and does not establish that any part was finished.
Assignment. FinishingMethodEligibilityUse-E1 : U.Episteme states the exact receiving eligibility-comparison use. TaskSignatureAssignmentRelation(PartFamilyFinishingProblemCard-E2, SurfaceFinishingEligibilitySignature-E1, FinishingMethodEligibilityUse-E1) has exactly the problem-side episteme, signature, and receiving-use episteme as participants.
It obtains only while that exact signature is adopted as the task-typing declaration for that exact card and the use stated in FinishingMethodEligibilityUse-E1, under ShopFinishing-Scheme-A, the declared part-family scope, ShopInspectionScheme-E4, and the production window above. Withdrawal of that adoption or change of a participant or qualification ends this assignment occurrence; a shared row, carrier, or publication does not make it obtain.
F. Clinical rehabilitation method-family eligibility.
Problem-side episteme. CohortRehabilitationProblemCard-E3 : U.Episteme is the exact C.22.2 ProblemCard for a rehabilitation service with Cohort-2026-Q3 and a stated capability-change question under clinical safety constraints.
TaskSignature. RehabilitationFamilyComparisonSignature-E1 declares EntityOfConcernRef=RehabilitationCapabilityChangeTarget-4, effectiveReferenceScheme=ClinicalRehabilitation-Scheme-C, TaskKind=rehabilitation-method-family comparison, and ScopeSlice(G)=Cohort-2026-Q3 in the declared care setting during [2026-08-01T00:00Z, 2026-11-01T00:00Z). It also declares outcome characteristics with their scale kinds, contraindication and resource constraints, and unknown tolerance or comorbidity values preserved as unknown. Include the evidence relations and follow-up windows only because this comparison use relies on them.
C.22 makes the comparison inputs explicit. Diagnosis and treatment recommendations remain claims governed by the applicable clinical patterns. For a separately current evidence or benefit claim, apply its evidence or clinical evaluation pattern; for a named gate decision, use A.21; for a care-permission claim, use A.2.8.PER and its GrantedPermissionRelation@Context predicate; for decision authority, recover the applicable authority predicate. Return missing-governor under A.6.RCD only when no governing rule can state or test that exact needed relation claim.
If clinical Work occurs, establish its actual performer System and obtaining assignment under A.13, then independently admit the dated Work and enacted Method under A.15.1. Expose the assignment identifier and declared species, and apply F.6, when this account or its receiving use consumes precise assignment-bound attribution through that assignment.
Missing or failed F.6 attribution alone leaves independently admitted Work intact; recheck an admission premise and dependent conclusions if the same finding questions that premise. A local system-role classification or assignment supplies none of the separate permission or authority grounds above.
Assignment. RehabilitationInterventionFamilyComparisonUse-E1 : U.Episteme states the exact receiving comparison use. TaskSignatureAssignmentRelation(CohortRehabilitationProblemCard-E3, RehabilitationFamilyComparisonSignature-E1, RehabilitationInterventionFamilyComparisonUse-E1) has exactly the problem-side episteme, signature, and receiving-use episteme as participants.
It obtains only while that exact signature is adopted for that exact card and the use stated in RehabilitationInterventionFamilyComparisonUse-E1, under ClinicalRehabilitation-Scheme-C, the declared cohort and care ClaimScope, the qualification window above, and the stated evidence-use conditions. Withdrawal of that adoption or loss of a participant or qualification ends this assignment occurrence; cohort labels, records, carriers, and organizations add no signature field or fourth participant.
C.22:7 - Bias-Annotation (lexical and discipline guards)
- Selector and policy references. For selection behavior called a strategy, cite the A.19.SelectorMechanism relation and the G.5 criteria, policy or
SelectorOutcomeused in the claim. - Transdiscipline vs domain. Comparability uses the exact ReferenceScheme, ClaimScope, characteristic meanings, and
U.Disciplinerelation when current. A Domain label is only a catalog mark; it supplies no norm, Bridge, scope, or comparison rule. - Descriptions and specification uses. Use E.10.D2 when the wording must distinguish a description of an entity from the use of a description as a specification.
C.22:9 - Conformance Checklist (normative)
-
Minimal A.6.0 declaration.
TaskSignatureexposes exactEntityOfConcernRef, effective reference scheme,SubjectKind,RangedValueKind, optionalResultKind,SliceSet, andExtentRule, plus Vocabulary, Laws, and Applicability. AddSignatureManifestonly when dependency replay needs imports and provided names; it does not supply signature identity. -
Signature and assignment present. Every exported selector-facing case names one TaskSignature identity and edition plus one
TaskSignatureAssignmentRelationwhose exact problem-side episteme, TaskSignature, receiving-use episteme, obtaining conditions, and occurrence extent are recoverable. No setting, carrier, or organization is added as a fourth participant. Current characteristic bindings are CHR-typed; a live unknown preservesunknown, while a non-current optional vocabulary item remains absent.1a. Publication and designators do not define identity. Two E.17 publications or serialized records that resolve to the same
<declaration content, EntityOfConcernRef, effectiveReferenceScheme>identify the same TaskSignature episteme. Carrier, layout, serialization,SignatureId, or edition label alone does not create a new identity component. -
CHR admissibility proven. Any numeric comparison or aggregation cites CG-Spec by Characteristic id and proves CSLC admissibility; no mean on ordinals; no unit mixing.
-
Unknowns remain typed. A live unknown remains
unknown, cites the direct downstream policy, and is not coerced. The acceptance, eligibility, or selector pattern records its own governed result. -
Evidence and assurance are conditional. For relied-on evidence or provenance, retain the supporting claim and its direct relation, with only the source edition, currentness and freshness conditions the use needs. Apply A.2.4 to an episteme’s evidence-use or status-use question and A.10 when replay of evidence, provenance or bounded reliance needs its account. When a named assurance claim about an exact target claim is current for this receiving use, including when a material-reliance rule requires it, cite its separate B.3 result, applicable assurance lanes, and declared fold. Otherwise the TaskSignature needs neither an evidence dossier nor an assurance fold.
-
Cross-semantic use is separated. Declare
ReferencePlanefor a value or objective head when its interpretation or comparison needs it. A scheme or plane difference alone creates neither a Bridge nor a penalty. Resolve two exact F.17 local senses, test the F.9 predicate, and state any proposed use separately. -
Acceptance thresholds live in CAL. G.4 AcceptanceClauses govern acceptance-gate thresholds. Signatures and implementations reference those clauses instead of redefining them as CHR properties or hidden constants.
-
Selector-use support. The TaskSignature exposes the scales, units, polarities, and admitted order relations needed by
G.5; it carries no mixed-scale scalarization or local selector verdict.G.5governs the emitted outcome and its source under G.5:4.4b–4.4e, including retained alternatives drawn from a non-dominated front. -
Bridge and use claims remain distinct. For current cross-semantic reuse, cite the two exact local senses and an F.9 Bridge only when its predicate is true. State the action, direction, correspondence rule, tolerated loss, and polarity in a separate bounded-use claim; if the predicate is false, keep the local values separate.
-
Packaging is conditional. Create an F.9 card, terminology row, Name Card, or reusable package only when a named receiving pattern independently needs that artifact. Packaging describes an already tested relation and separate use claim; it establishes neither.
-
Structure and gates are conditional. Apply E.18 crossing checks only when a selected transformation-flow structure and exact
GateCrossingare current. Apply A.21 only when a named gate decision is current. Each pattern supplies its own result; neither a structure nor a gate makes F.9 obtain, and C.22 requires no crossing or gate package otherwise. -
QD fields (when QD is in scope). A TaskSignature for an active QD search is complete only when it carries CHR-typed CharacteristicSpaceRef with the dimension required by its declared descriptor and comparison rules, the separate characteristicSpaceDescriptionRef to its exact defining episteme, ArchiveConfig (topology, resolution, K,
InsertionPolicyRef,DistanceDefRef.edition) and EmitterPolicyRef; each characteristic declares its ReferencePlane when interpretation or the receiving contract requires it.PortfolioMode=Archiveor a set-valued selector outcome alone activates none of these fields. -
DominanceRegime default.
DominanceRegimedefaults toParetoOnly. Illumination enters dominance only through a cited CAL.Acceptance policy enabling that relation; the SCR records the policy id. -
Telemetry (when used). The active telemetry record carries PathSliceId when an E.18 path slice is current, the applicable decay and refresh policy ids, and the exact space-description episteme and the editions of the distance and emitter definitions on which that telemetry depends. An illumination increase is traceable to its governing policy id. Ordinary evidence retention does not require an illumination record.
-
GeneratorIntent (when OEE is in scope). A TaskSignature supports the claimed OEE generator-family use only when
GeneratorIntentcitesEnvironmentValidityRegionandTransferRulesRefwith ids resolvable in G.5 and C.23. Any downstream abstention is their result, not a C.22 output. -
Budgets. When
Budgetingis live, its evaluation, time, and batch values carry declared units and the applicable E/E-LOG exploration-budget id. -
Archive-comparison support. The TaskSignature supplies the measurement and comparison basis required by the claimed comparison. When that comparison uses distance or novelty, its distance definition and novelty measures must be CSLC-admissible and editioned, including
DistanceDefRef.editionwhen used. Retaining the alternatives in an archive does not by itself require distance. The archive or selector pattern defines or constrains any downstream abstention or returned-set result. -
Planes. QD heads and characteristics declare a
ReferencePlanewhen their meaning or comparison depends on it. A plane difference alone creates neither correspondence nor an assurance adjustment. Use item 8 for a current cross-semantic claim and item 4 for any separately current B.3 assurance result. -
Unknown QD values. A live unknown QD field remains
unknown, cites the policy governing its downstream use, and is not coerced or mapped by C.22 itself. -
Specialization claims referenced. Name the holder or scoped portfolio, task family, work target, performance criterion and supporting basis. For an acquisition claim, support the relevant change; for a C.22.1 adaptation comparison, retain its budget, prior-exposure and threshold-crossing conditions. Keep the TaskSignature edition and assignment, freshness and provenance required by the receiving use.
C.22:8 - Common Anti-Patterns and How to Avoid Them
| Countercase | Repair |
|---|---|
| A preferred method or strategy name is inserted into the TaskSignature before eligibility is tested. | Remove the method value, restore the problem traits, and let A.19, C.23, G.4, and G.5 govern later comparison and selection. |
A live unknown is encoded as false, 0, or an empty value. | Restore unknown, name the direct basis relation and receiving-use policy, and let the downstream pattern produce the result governed by that policy. |
| Ordinal values or values with unlike units are averaged into one score. | Recover scale, unit, polarity, reference plane, and admitted order for every head; use only a directly governed admissible comparison or leave the candidate set partially ordered. |
| One TaskSignature mixes design-time traits, later run observations, and incompatible DesignRunTag positions. | Split the claims by their actual Work and relation positions; retain only the traits current in this signature edition. Use E.18 only when a selected transformation-flow relation is current, F.9 only when exact local meanings differ and its Bridge predicate is true, and the Work pattern for the actual run. |
| A new file, card, or database row is treated as a new TaskSignature. | Compare the declaration content, exact task or work target, and effective reference scheme. Reuse the same identity when only publication, carrier, serialization, or designator changed; identify another episteme when an identity component changed, then apply C.22:5.2’s signature-membership and edition-continuity checks. |
| A broad domain, organization, or location label is used as if it supplied scope, measurement, evidence, or selection rules. | Recover the exact EntityOfConcern, effective ReferenceScheme, A.2.6 ClaimScope relation, U.Discipline when current, characteristic rules, and direct selector or policy relations that the use needs. |
Data shift is assumed away because the old profile used iid. | State the current ShiftClass or unknown, cite its evidence and currentness relation, and state the changed-use result under the exact acceptance or selector predicate. |
| A vendor, tool, or fashionable method label is treated as a normative selector input. | Keep it only as a Plain example or recover the exact method-description, capability, evidence, and selector relations on which comparison relies. |
C.22:10 - Selector Fields And Evidence Relations
Inputs.
- One stabilized problem-side episteme and CG-Spec ids;
- a
ProblemProfile(…Description) only when C.22:5.0b’s replay need is current; - the supporting evidence-use or provenance relation when relied on, with an A.10 account only when replay of that basis needs it;
- D.CTX only when that separate context relation is current;
- the QD values or references named by CharacteristicSpaceRef, ArchiveConfig, and EmitterPolicyRef only when QD is live;
- GeneratorIntent only when OEE is live.
Produces. Declare one TaskSignature episteme as the U.Signature species specified in C.22:5.2. When that signature is adopted for the problem-side episteme and receiving use, one separate TaskSignatureAssignmentRelation obtains among that signature, the exact problem-side episteme, and exact receiving-use episteme under C.22:5.2.
TaskSignature is neither a new root U-kind nor a record kind: its A.6.0/C.2.1 identity tuple and declaration content determine episteme identity; signature membership and edition continuity follow C.22:5.2, while publication, carrier, and serialization remain outside identity. Optional QD, archive, generator, PortfolioMode, and telemetry vocabulary appears only when current.
Used by. G.5 (Eligibility and Selection kernel), G.4 (Acceptance and Evidence), C.23 (admit, degrade, and abstain rules and method-family maturity checks).
C.22:11 - Consequences (informative)
- Admissible selection. Selection is explainable and inspectable: every admission or rejection reason cites the TaskSignature positions and CHR rules on which it relied. When a separate B.3 assurance result is current, that result cites its own evidence relations, applicable assurance lanes, and declared fold.
- Use-bounded first, cross-semantic use explicit. The exact EntityOfConcern, effective ReferenceScheme, ClaimScope, and receiving use are primary. When local meanings differ, an obtaining F.9 Bridge and a separate bounded-use claim make the proposed reuse inspectable; evidence reliance, assurance, gates, and packaging remain optional neighboring questions.
- Common downstream inputs. G.1-G.5 use one typed TaskSignature; acceptance-gate thresholds are separated into Acceptance; unknowns are not guessed.
- QD and OEE-ready. Typed QD and GeneratorIntent fields make declared returned-set structure and open-ended generation contexts explicit, with admissible dominance, editioned distances, and policy-aware illumination.
C.22:12 - Rationale
Selecting a method before the relevant eligibility, acceptance, unknown-handling, and admissible-comparison relations are explicit invalidates the selector-facing problem record. When the selection use relies on evidence, provenance, currentness, or assurance, those separate relations and results must also be explicit; C.22 does not require them otherwise.
C.22:12.1 - SoTA-Echoing
Wolpert and Macready’s “No Free Lunch Theorems for Optimization”, 1997, remains historical lineage for the warning that method superiority is distribution-dependent. It does not by itself supply the current C.22 field set, a selector policy, or evidence that one TaskSignature is adequate. The following sources inform different parts of this task-typing use.
| Source and scope | Contribution | Use boundary |
|---|---|---|
| Roger Jiao, “Towards rigorous problem formulation for engineering design research: from motivations to measurable claims via metric-measure-method”, Journal of Engineering Design 37, 2026; Szajnfarber, Lifshitz and Tushman, “Beyond translation: how context work during problem formulation enables effective solving by outsiders”, 2026. | Formulate the problem before selecting a method; connect operational characteristics and measures to the decision or action that will use them. In C.22 this supports the task/work target, declaration content and distinct problem-side and receiving-use participants. | Engineering research formulation and outsider problem solving do not establish FPF kinds or a universal signature schema. An available method or dataset does not determine the task declaration. |
| Cenikj, Kudela, Tuba and Eftimov, “Evaluating Real-World Generalizability of Algorithm Selection Models”, June 2026. | Problem characteristics inform selection, while transfer from benchmark to real-world landscapes may fail. Preserve scope, unknowns and the evidence or currentness basis consumed by the use; reopen a prior selector result only when its dependency changed. | The optimization-landscape study supplies neither a universal transfer-failure claim nor a complete TaskSignature field list for every sector. |
| Qin et al., “A survey on Quality-Diversity optimization: Approaches, applications, and challenges”, Swarm and Evolutionary Computation 100, 2026; Lin et al., “Quality-Diversity Optimization as Multi-Objective Optimization”, 2026 preprint. | QD uses collection-valued results, declared behavior/characteristic spaces, containers and policies. Its optional signature fields preserve the declared dominance rule, archive basis and separate telemetry; the MOO treatment is one approach. | The sources do not define all QD through MOO or authorize every diversity measure to enter dominance. A different container, metric or scalarization needs its own applicable comparison and policy basis. |
| SciML, “Problem Interface” and “Common Solver Options”. | Separate the constructed problem value from solver dispatch and explicitly remake it when fields change. For C.22, apply identity/membership/continuity checks to semantic declaration changes and recheck an assignment whose participant changed. | This software interface supplies no transdomain ontology and does not require every project to use an immutable software record. A material interface change reopens only the adaptation that relies on it. |
C.22:13 - Relations
Builds on: C.16 MM-CHR, G.0 CG-Spec. Coordinates with: G.4 Acceptance, G.5 Selector, C.18 NQD-CAL, C.19 E/E-LOG, C.23 Method-SoS-LOG, and C.32.P2S when typed problem pressure continues into architecture selected structures and synthesis. Constrained by: E.10 for selected EntityOfConcern, Description-episteme, specification-use, and publication-lane wording. Coordinate with E.18 only when a named receiving use depends on a selected transformation-flow crossing or gate position; E.18 supplies neither the F.9 Bridge nor a mandatory publication package.
C.22:14 - Practical Use Checks
- If two candidate approaches are answering different
TaskKinds or differentScopeSlice(G)cuts, a direct comparison is not admissible yet. - For a specialization claim, recover the holder or portfolio, task family, performance criterion and grounds. Recover acquisition history and the C.22.1 adaptation budget and threshold-crossing basis when those are the claim or comparison being made.
- If crossing, normalization or missingness changes the comparison, state the relevant meaning and rule in the signature or its cited basis. Ordinary explanatory prose is sufficient when it supplies that declared content; an undocumented assumption is not.
- If
QDorOEEheads are in scope, archive and generator fields belong in the same typed signature rather than in a detached explanatory appendix.
C.22:15 - Generator-specific difficulty targets
When a declared search or learning policy seeks feasible but challenging tasks, state the feasibility and difficulty criteria for that task family in the policy. Use C.19 when the question is treatment of a still-live pool, and C.18 for the generation or archive account. Difficulty alone does not establish that a TaskSignature is informative. An ordinary signature includes only traits needed by its receiving use, with no general requirement to make the task harder. G.1 applies when authoring a reusable generator/selector kit; it supplies no universal task-difficulty target.