#Γ_epist — Knowledge-Specific Aggregation

B.1.3 · Stable · Part B - Trans-disciplinary Reasoning Cluster

#About this pattern

This is a generated FPF pattern page projected from the published FPF source. It is canonical FPF content for this ID; it is not a FPF Reference product feature page.

#How to use this pattern

Read the ID, status, type, and normativity first. Use the content for exact wording, the relations for adjacent concepts, and citations to keep active work grounded without pasting the whole specification.

At a glance. Use B.1.3 to compose exact U.Episteme inputs into one knowledge aggregate while preserving provenance, conceptual fit, context, and conservative assurance bounds.

Use this when. Use this pattern when a named synthesis or compilation use depends on how claims, models, datasets, or arguments are combined, and the aggregation must keep source, mapping, conflict, order, and temporal qualifications inspectable.

Not this pattern when. Use C.2.1 for episteme identity and edition continuity, A.14 for a proper temporal restriction of one unchanged episteme, A.15.1 for Work parts or occurrences, B.1.4 for a bounded aggregation of already recovered order or temporal relations, and B.3 for the assurance claim that consumes the aggregate.

What changes in practice. Identify every input episteme and mapping before folding; preserve provenance and conflicts; and return identity, edition, temporal restriction, Work, publication, and assurance questions to their direct owners.

► decided‑by: A.14 Advanced Mereology A.14/C.2.1 compliance — Use ConstituentOf for semantic parts and PortionOf only for quantitative splits of texts/data with declared μ. Use PhaseOf only for a proper interval of one unchanged C.2.1 episteme. When a MethodDescription or document episteme's claim content, EntityOfConcern, or effective ReferenceScheme changes, identify another episteme and assert EpistemeEditionRelation only when its historical-continuation predicate obtains. Work segmentation uses A.15.1; no ComponentOf is used here.

Plain‑English headline. Γ_epist composes epistemic holons (claims, models, datasets, arguments) into a single episteme while preserving provenance, applying conservative trust bounds (B.3 F/G/R), and penalizing poor conceptual fit via congruence levels (CL). It is not a physical sum; it is a semantic and evidential fold.

• Holonic foundation. In the FPF, a U.Episteme is a holon whose identity is knowledge‑bearing (A.1). It can be a statement/claim, a model, a theory, a specification, a dataset with semantics, or a compiled scholarly artifact.
• Strict Distinction (A.15). We separate: structure (what the episteme comprises), order (argument flow), identity and history (C.2.1 identities and edition relations), proper temporal restriction (A.14), work (what was spent to produce/validate it), and values (objectives/criteria). Γ_epist stays in the structure/semantics lane and calls out to Γ_ctx/Γ_time/Γ_work only after their direct inputs are recovered.
• Mereology (A.14). For knowledge composition we primarily use ConstituentOf (logical/semantic parts), UsageOf/ReferenceTo (external reliance), and MemberOf for collections (anthologies, corpora). We do not use ComponentOf (physical) in Γ_epist. PhaseOf may restrict the same unchanged episteme to a proper interval when its complete C.2.1 identity triple remains fixed. Distinct labelled versions or revisions require distinct C.2.1 identities when a discriminator changes and an independently obtaining EpistemeEditionRelation for any claimed historical continuation. RoleBearerOf is irrelevant here because knowledge does not play a role—it is used by a holon‑in‑role (Transformer) at run‑time (A.12).
• Assurance (B.3). Knowledge carries F, G, R (Formality, ClaimScope, Reliability). Integration edges carry CL (congruence level) that penalizes poor fit. Γ_epist must preserve provenance and apply conservative bounds: no “truth averaging,” no silent context hops. Obligations here are mode/assurance‑gated per C.2.1. # [M‑0]
• Order/time flavours. Argument sequences may need Γ_ctx (non‑commutative ordering of premises to conclusion). Knowledge evolution first uses C.2.1 to identify exact epistemes and any obtaining edition relations; B.1.4/Γ_time may then aggregate already recovered temporal restrictions, relation order, deprecation, or update windows for a bounded use. The aggregation creates neither identity nor continuity. When composition produces new closure or supervision (e.g., explanatory theory emerges), we declare MHT (B.2).

#Keywords

• knowledge aggregation
• epistemic
• provenance
• trust
• KD-CAL.

#Relations

#Content

#Problem frame

• Holonic foundation. In the FPF, a U.Episteme is a holon whose identity is knowledge‑bearing (A.1). It can be a statement/claim, a model, a theory, a specification, a dataset with semantics, or a compiled scholarly artifact.
• Strict Distinction (A.15). We separate: structure (what the episteme comprises), order (argument flow), identity and history (C.2.1 identities and edition relations), proper temporal restriction (A.14), work (what was spent to produce/validate it), and values (objectives/criteria). Γ_epist stays in the structure/semantics lane and calls out to Γ_ctx/Γ_time/Γ_work only after their direct inputs are recovered.
• Mereology (A.14). For knowledge composition we primarily use ConstituentOf (logical/semantic parts), UsageOf/ReferenceTo (external reliance), and MemberOf for collections (anthologies, corpora). We do not use ComponentOf (physical) in Γ_epist. PhaseOf may restrict the same unchanged episteme to a proper interval when its complete C.2.1 identity triple remains fixed. Distinct labelled versions or revisions require distinct C.2.1 identities when a discriminator changes and an independently obtaining EpistemeEditionRelation for any claimed historical continuation. RoleBearerOf is irrelevant here because knowledge does not play a role—it is used by a holon‑in‑role (Transformer) at run‑time (A.12).
• Assurance (B.3). Knowledge carries F, G, R (Formality, ClaimScope, Reliability). Integration edges carry CL (congruence level) that penalizes poor fit. Γ_epist must preserve provenance and apply conservative bounds: no “truth averaging,” no silent context hops. Obligations here are mode/assurance‑gated per C.2.1. # [M‑0]
• Order/time flavours. Argument sequences may need Γ_ctx (non‑commutative ordering of premises to conclusion). Knowledge evolution first uses C.2.1 to identify exact epistemes and any obtaining edition relations; B.1.4/Γ_time may then aggregate already recovered temporal restrictions, relation order, deprecation, or update windows for a bounded use. The aggregation creates neither identity nor continuity. When composition produces new closure or supervision (e.g., explanatory theory emerges), we declare MHT (B.2).

#Problem

Naive aggregation of knowledge holons causes recurring failures:

1. Trust inflation by averaging. Averaging confidences of conflicting claims creates a falsely “reliable” whole; violates WLNK and B.3 conservatism.
2. Provenance erasure. Merges that drop sources, methods, or links break A.10 Evidence Graph Referring and make results unauditable.
3. Semantic drift. Folding across mismatched concepts without explicit mappings (and their CL) yields incoherent composites that look formal but mean nothing.
4. Order blindness. Arguments with essential dependency order (premise ⇒ lemma ⇒ conclusion) are treated as sets; non‑commutativity is lost and results become non‑reproducible.
5. Context chimeras. Combining items across bounded contexts (different vocabularies/units/policies) without a Context Reframe (B.2) silently corrupts claims and inflates R.
6. Category errors. Importing Γ_sys rules (e.g., “sum truth,” “avg formality”) into knowledge composition produces physically sounding but epistemically nonsensical models.

#Forces

#Solution — Terms, operator family, invariant Standard, core rules

#Terms (didactic recap)

• U.Episteme — a claim-bearing knowledge holon. C.2.1 identifies it through the participant-determined EpistemeConstitutionRelation over <claim content, exact EntityOfConcern, effective ReferenceScheme>. ClaimGraphSlot, EntityOfConcernSlot, and ReferenceSchemeSlot name participant meanings only inside that relation's reusable declaration; they are not internal slots of the episteme. Empirical grounding uses the separate EpistemeEmpiricalGroundingRelation, while text, code, figures, datasets, SCR/RSCR references, publication forms, and presentation carriers remain separately governed provenance, representation, publication, or carrier material.
• Evidence/Provenance Graph — edges like evidences, derivesFrom, usesMethod, isMeasuredBy with anchors (A.10).
• Mapping edge — a typed relation between conceptual vocabularies (e.g., ontology alignment, unit conversion) with a CL score (0…3/4 per A.15/B.3 convention).
• SCR — a U.SCR that lists all symbol carriers included in the aggregate; never dropped.
• Bounded context — a modelling Standard (vocabulary/units/policy). Crossing it requires Context Reframe (B.2) or explicit mappings with CL.

Didactic reminders. • Knowledge does not “act.” Transformers (A.12) use knowledge. • MemberOf creates collections; it is not a semantic argument link. Use ConstituentOf for logical/evidential composition. • PhaseOf is only a proper temporal restriction of one unchanged episteme. Changed C.2.1 discriminators identify another episteme; test EpistemeEditionRelation separately. Use MHT only for a remaining whole-reidentification question, not as a substitute for C.2.1 identity.

#The operator family (companion flavours)

To keep design vs run clean (A.15), Γ_epist has two companion flavours that share the same algebra but serve different moments:

1. Synthesis (design‑time) — fold epistemes into a draft aggregate

Γ_epist^synth : ( D_know : DependencyGraph< U.Episteme >,
                  TA     : U.RoleAssignment[roleRef = TransformerRole@Context] ) → U.Episteme

• Domain. D_know uses ConstituentOf, UsageOf/ReferenceTo, evidences/derivesFrom, optional MemberOf for collections.
• Result. A composite episteme whose Object/Concept/Symbol components are assembled; provenance and SCR are preserved; F/G/R/CL are provisionally computed for later assurance. Gating: at M‑mode only tuple placeholders are required; numeric scoring may be omitted ([M‑0/M‑1]). At F‑mode the tuple MUST be computable in‑Context ([F‑*,L1+]). # [M/F]

2. Compile (run‑time) — produce the released artifact in a bounded context

Γ_epist^compile : ( E_synth : U.Episteme,
                    Ctx     : BoundedContext,
                    TA      : U.RoleAssignment[roleRef = TransformerRole@Context] ) → U.Episteme

• Domain. A synthesized episteme and a target context (journal, standard, program spec).
• Result. A context‑anchored episteme (e.g., published paper/spec) whose mappings to the context vocabulary are explicit and carry CL; assurance will reference this context baseline (B.3).

Relationship to Γ_ctx / Γ_time. If the knowledge fold explicitly depends on argument order (e.g., derivation), the internal fold uses Γ_ctx for the sequence. If a temporal storyline (updates, retractions) is important, first identify each exact episteme and any obtaining C.2.1 edition relation; then use B.1.4/Γ_time to aggregate the already recovered temporal restrictions, relation order, or applicability windows required by the use. Γ_epist composes the exact selected episteme inputs, not a label-defined “current slice”. If composition yields new explanatory closure beyond WLNK/CL, declare MHT (B.2).

#Invariant Standard (how the Quintet applies; math by level)

• IDEM (Idempotence). Folding a single episteme returns itself; no accidental “upgrade.”
• COMM/LOC (Local commutativity / locality). For independent subgraphs (no logical/evidential dependency), fold order/location is irrelevant; when dependencies exist, Γ_ctx controls order explicitly.
• WLNK (Weakest‑link bound). Aggregate Reliability (R) is bounded by the weakest supported link along any justification path, after considering the lowest CL on mappings used by that path.
• MONO (Monotonicity). Strengthening a part (raising R with valid evidence or raising CL on a needed mapping) cannot lower aggregate R. Adding contradictory evidence is not an improvement; it triggers conflict handling (below), not MONO.

2. Reliability fold. Along any support spine, R_raw = min_i R_i; apply congruence penalty Φ(CL_min) → R_eff = max(0, R_raw − Φ(CL_min)). No averaging; weakest‑link. Math by level: – [M‑0/M‑1] allow ordinal comparisons only (no arithmetic on R); Φ may be stated qualitatively (“low/med/high”). – [M‑2/L1] require numeric Φ table (default in §4.4) and reproducibility tag on empirical edges. – [F‑*,L1/L2] require formal derivability of the fold rules from LOG‑CAL; constructive mode annotates proof.kind=constructive. # [M/F]

#Core rules for epistemic aggregation (design‑time synthesis)

When computing Γ_epist^synth(D_know, T):

1. Provenance preservation. The provenance/evidence graph is unioned with de‑duplication; every claim in the aggregate remains traceable to its sources and methods. No source, method, or dataset that supports a retained claim may be dropped.

2. SCR construction. Build a U.SCR that lists all symbol carriers (texts, code, figures, datasets) that materially participate in the aggregate. Provenance nodes must be mappable to SCR entries.

3. Object alignment. Determine a common Object via domain taxonomy (e.g., least common ancestor) or create a U.CompositeEntity with explicit mappings. Record CL for each mapping; do not silently merge homonyms.

4. Concept integration with CL penalty. Compute provisional F/G/R of the aggregate:

   • F_eff = min(F_i) (formality is as strong as the least formal constituent actually used).
   • G_eff = function of coverage; typically monotone in included scope, capped by weakest definitional fit.
   • R_eff = min over justification paths of { R_i along the path } penalized by the lowest CL used by that path: R_eff := max(0, min_path( min_claimR(path) − Φ(CL_min(path)) )), where Φ is the normative penalty function defined below. If a mapping with CL < threshold is essential to a path, mark the claim provisional.

5. Normative Penalty Function Φ (v1.0) The penalty function Φ quantifies the loss of reliability due to poor conceptual alignment between parts.

A domain profile MAY provide an alternative table but MUST preserve monotonic decrease (a lower CL cannot have a smaller penalty). The default values are derived from empirical fits in KD-CAL Bench 0.3.

6. Conflict detection (no averaging). Detect contradictions (e.g., p and ¬p with overlapping scope). Do not average. Either (i) separate by context or scope (bounded contexts; Γ_time slices), (ii) mark provisional with explicit conflict edges, or (iii) if resolution yields new closure, consider MHT.

7. Handling of Axiomatic vs. Postulative Epistemes In alignment with ADR-028, the computation of R_eff depends on the episteme's declared mode.

• For an input episteme E_i with mode: axiomatic, empirical R is N/A; take R_i_eff = F_i. Tag: line=formal. # [F‑*]
• For mode: postulative, use declared R_i with decay; Tag: line=empirical. # [M‑1/M‑2/F]
• The aggregate E_eff MUST also declare a mode. If all inputs are axiomatic, the output is axiomatic. If any input is postulative, the output MUST be postulative.
• Constructive note. Under F‑constructive, equivalence claims use isomorphism/equivalence in the chosen UF library; CL=2 means proof‑reconstructed alignment, not mere model‑theoretic appeal. # [F‑constructive]

7. Order‑aware arguments (optional). If the argument requires premise ordering, embed a Γ_ctx fold inside Γ_epist; record the OrderSpec for reproducibility (NC‑1..3). Gating: OrderSpec is recommended at M‑1 and required at M‑2/F. # [M‑1→F]

8. No costs here. Any compute/collection effort is Γ_work; attach references but do not mix costs into epistemic aggregation.

#Core rules for compilation (run‑time context anchoring)

When computing Γ_epist^compile(E_synth, Ctx, T):

1. Context bindings. # [M‑1+] Map all operative concepts/units/claims into Ctx; record mappings and their CL. If the rebase changes boundary/objective of the episteme (e.g., from descriptive compendium to explanatory theory with commitments), declare Context Reframe (MHT) per B.2.

2. Assurance baseline (gated). Recalculate the assurance tuple (B.3) in Ctx: F and R may change with formalization and mapping penalties; G is re‑expressed in Ctx’s scope. Gating:

• [M‑0] narrative justification only;
• [M‑1] qualitative tuples allowed;
• [M‑2/L1] numeric tuple required;
• [F‑*/L2] tuple and proof obligations on weight/penalty model selection. # [M/F]

3. Release SCR. Produce RSCR with carrier hashes; at L2 require independent re‑hash verification. # [M‑1/L2]

4. Order/time hooks. If the compiled artifact includes an internal derivation, carry the OrderSpec. If it selects knowledge for a time-bounded use, name the exact C.2.1 episteme identity and link to the already recovered proper temporal restriction, edition relation order, applicability window, or B.1.4/Γ_time aggregation actually used.

#Archetypal grounding (worked, didactic)

#Episteme — Meta‑analysis into a guidance statement

• Inputs (U.Episteme): E₁ randomized trial (R=0.84, F=3, G=medium), E₂ observational study (R=0.55, F=2, G=wide), E₃ mechanistic model (R=0.60, F=3, G=narrow). Mappings: dosage units (mg ↔ IU), outcome definitions (pain scale variants), each with declared CL (e.g., unit mapping CL=3, outcome alignment CL=2).

• Γ_epist^synth:

  • Provenance preservation: all study protocols, datasets, analysis scripts listed in the SCR.
  • Object alignment: “acute low‑back pain within 6 weeks” via taxonomy LCA; non‑aligned chronic cohorts excluded or mapped with low CL and flagged.
  • Concept integration: compute provisional R_eff along each justification path, penalized by **Φ(CL_min(path)); aggregate R_eff` = min over paths.
  • Conflict handling: E₂ contradicts E₁ in a subgroup; kept as provisional with explicit conflict edge and scope note (different baseline severity).

• Γ_epist^compile (journal context): Map outcomes to journal’s required measure, recalc F/G/R with mapping penalties; produce release SCR (hashes, versions) and context baseline. Result: “Guidance Statement v1.0” with conservative R.

• Why not averaging? Averaging would inflate R and hide low‑CL outcome mappings; Γ_epist enforces pathwise min + CL penalty.

#Episteme — Safety case from heterogeneous evidence

• Inputs: requirement spec (F=3, R=0.7), hazard analysis (F=2, R=0.6), test logs (F=1, R=0.8), formal proof of controller property (F=3, R=0.9).

• Γ_epist^synth:

  • Provenance union; SCR includes requirements, proof carrier, test datasets.
  • Concept integration: controller proof applies only under assumptions A; test logs violate A in edge case → CL low for mapping “test scenario ≡ proof assumption.”
  • R_eff bounded by the weakest justification path after Φ(CL_min); claim on “system‑level safety” marked provisional until assumption alignment is demonstrated.

• Γ_epist^compile (certification context): Context re‑base to regulatory vocabulary; if the re‑base changes objective/boundary (e.g., from internal assurance to public certification), consider MHT (Context Reframe) per B.2.

#Contrast (didactic)

#Proof obligations (normative)

At synthesis (Γ_epist^synth):

1. PO‑SYN‑PROV. The provenance/evidence graph MUST be preserved (union with de‑duplication); every retained claim is traceable to sources/methods in the SCR.
2. PO‑SYN‑OBJ. The Object MUST be identified (single subject via LCA or explicit U.CompositeEntity) with declared mappings and their CL.
3. PO‑SYN‑CL. All mapping edges that bridge semantics/units MUST carry CL; the chosen penalty Φ MUST be monotone in CL (lower CL ⇒ higher penalty). Thresholds for marking provisional MUST be stated.
4. PO‑SYN‑R. R_eff MUST be computed as min over justification paths of (claim reliabilities along the path minus Φ(CL_min(path))). No arithmetic mean is allowed for reliability.
5. PO‑SYN‑CONFLICT. Contradictions MUST be either (i) separated by context/scope, (ii) marked as provisional with explicit conflict edges, or (iii) escalated to MHT if resolution yields new explanatory closure.
6. PO‑SYN‑ORDER. If order matters, the OrderSpec MUST be recorded and Γ_ctx NC‑1..3 (determinism, context hash, partial‑order soundness) MUST hold.
7. PO‑SYN‑NOWORK. Resource spending, yields, and dissipation MUST NOT be computed here; instead, attach references to the aligned Γ_work composition.

At compilation (Γ_epist^compile):

1. PO‑COMP‑CTX. The target bounded context MUST be declared; all active concepts MUST be mapped with CL; context vocabulary/units recorded.
2. PO‑COMP‑ASSUR. The assurance tuple (F/G/R) MUST be recomputed in the target context with the applied CL penalties.
3. PO‑COMP‑REL. A release‑grade SCR (hashes, versions, dates) MUST be produced.
4. PO‑COMP‑MHT. If the compilation re‑anchors boundary, objective, or identity (e.g., from compendium to explanatory theory), an MHT (Context Reframe) MUST be declared with a Promotion Record (B.2).
5. PO‑COMP‑ORDER/TIME. If derivational order is essential, the OrderSpec MUST be referenced. If temporal selection is essential, name the exact C.2.1 episteme identity and reference the already recovered proper restriction, edition-relation order, applicability window, and B.1.4/Γ_time aggregation actually consumed.

#Conformance Checklist (normative)

#Anti‑patterns & repairs

#Consequences

Benefits

• Auditability by construction. Every retained claim remains tied to its sources; SCR guarantees reconstructability.
• Safe synthesis. R cannot be inflated; CL penalties make conceptual misfit explicit.
• Context‑aware releases. Compiled epistemes or publications are aligned with a declared context; cross‑context reuse is principled.
• Didactic clarity. Separates semantic folding (Γ_epist) from order (Γ_ctx), time (Γ_time), spend (Γ_work), and emergence (B.2).

Trade‑offs

• Mapping overhead. Declaring mappings and CL costs time; it prevents silent incoherence.
• Conservative stance. Results may look pessimistic; this is deliberate (WLNK). Use MHT only for genuine explanatory closure.

#Rationale (informative)

• Epistemic composition is not physical addition. Reliability must be bounded by the weakest justified path, not averaged; conceptual misalignment must reduce confidence, not be ignored.
• Provenance is part of meaning. Dropping sources/methods changes what the episteme is; Γ_epist treats provenance and SCR as first‑class.
• Context matters. Bounded contexts structure practice; formal Context Reframe (MHT) prevents quiet re‑interpretations of claims.
• Parsimony with power. A small set of rules (provenance preservation, CL‑penalized pathwise min, order/time hooks, context discipline) is enough to model scientific and engineering knowledge without importing domain‑specific tool jargon.

#Relations

• Builds on: A.12 (Transformer Role—compilers/editors enact), A.14 (ConstituentOf/MemberOf and proper temporal restriction of one unchanged carrier), C.2.1 (episteme identity and independently obtaining edition relations), and A.15/A.15.1 (Strict Distinction and Work-temporal law).
• Coordinates with: B.1.1 dependency-structure and relation-grounding checks, B.1.4 (Γ_ctx/Γ_time inside knowledge folds), B.1.6 (Γ_work for compute/collection spend).
• Triggers/Complements: B.2 (MHT) when explanatory closure or context re‑base creates a new whole (theory, standard).
• Used by: B.3 assurance uses F/G/R and CL baselines computed here as inputs to trust calculations.

One‑sentence takeaway. Γ_epist preserves provenance, penalizes poor conceptual fit, forbids reliability averaging, and makes context explicit—so that knowledge aggregates are conservative, auditable, and genuinely coherent.

#B.1.3:End

Last Updated: 2026-08-04 — upstream FPF commit 67092138 (github.com/ailev/FPF)