Reference Sheet 1: Category Theory for Cross-Model Reasoning#

Target audience: Forge auditor who knows S5 modal logic, CEM, FOL, and basic game theory but needs category theory for formalizing cross-model structure.

1. Orientation#

Category theory is the mathematics of structure-preserving maps. Where S5 and CEM give you tools to reason within a single model, category theory gives you tools to reason between models — and to ask whether correspondences between models are structural (functorial) or accidental. The matheology system has two models (PET and JUB), 25 axioms split across them, alignment echoes that may or may not be systematic, and a compiler infrastructure (SISYF/PROMY) that moves content across a 5D label space. Category theory formalizes all of these as instances of a single framework: objects, arrows, and the laws they satisfy.

2. Key Concepts#

Category. A collection of objects and arrows (morphisms) between them, with identity arrows and associative composition. Notation: a category C has objects Ob(C) and for any A,B ∈ Ob(C), a hom-set C(A,B) of arrows A → B. Matheology use: PET is a category whose objects are axiom-groups (mereological core, modal, relational, divine nature, revelation bridge) and whose arrows are derivations (ax1–ax4 entail th1, etc.).

Functor. A structure-preserving map F: C → D sending objects to objects and arrows to arrows, respecting identity and composition: F(id_A) = id_{F(A)} and F(g ∘ f) = F(g) ∘ F(f). Matheology use: A functor PET → JUB would map each PET axiom group to a JUB axiom group and each PET derivation to a JUB derivation in a way that preserves the proof structure. If such a functor exists, the alignment echoes are structural.

Natural transformation. Given functors F,G: C → D, a natural transformation α: F ⇒ G is a family of arrows α_A: F(A) → G(A) for each object A, such that for every arrow f: A → B the diagram commutes: G(f) ∘ α_A = α_B ∘ F(f). Matheology use: If two different ways of mapping PET into JUB both work, a natural transformation between them measures how they differ — systematically, not ad hoc.

Limit and colimit. A limit is a universal construction that “combines” a diagram of objects by finding the most general object that maps into all of them (product, pullback, equalizer). A colimit does the dual: the most general object that all of them map into (coproduct, pushout, coequalizer). Matheology use: The compiled view of PET+JUB is a colimit — the smallest structure that contains both models with their overlap identified. Restricting to axioms shared by both is a limit.

Adjunction. Functors F: C → D and G: D → C form an adjunction F ⊣ G when there is a natural bijection D(F(A), B) ≅ C(A, G(B)) for all A, B. F is the left adjoint (free construction); G is the right adjoint (forgetful/underlying structure). Matheology use: If extending PET to JUB (adding axioms ax15–ax25) is left adjoint to forgetting the JUB-specific structure back to PET, then the extension is “optimally free” — it adds exactly what is needed and nothing more.

Comma category / slice category. For a fixed object C, the slice C/C has as objects all arrows into C, and as arrows the commuting triangles. Dually, C/C is the co-slice. Matheology use: The HELL findings rooted at a specific axiom form a slice category over that axiom — formalizing the “attack surface” of each axiom.

Equivalence of categories. Functors F: C → D and G: D → C with natural isomorphisms GF ≅ Id_C and FG ≅ Id_D. Strictly stronger than a functor existing in each direction; it means the categories have the same structure up to isomorphism. Matheology use: Are PET and JUB equivalent (same structure, different presentation) or merely related (functor exists but information is lost)? This is a precise question with a precise answer.

Diagram and cone. A diagram is a functor from an index category J to C. A cone over a diagram is an object with arrows to every object in the diagram that commute with the diagram’s arrows. The limit is the universal cone. Matheology use: The axiom dependency graph is a diagram; a proof that depends on multiple axioms is a cone over that sub-diagram.

3. Critical Theorems#

Yoneda Lemma. For any functor F: Cop → Set and object A of C, Nat(C(A,−), F) ≅ F(A). Informally: an object is completely determined by its relationships to all other objects. Why it matters: An axiom’s identity in the matheology system is fully determined by what it proves, what attacks it, and how it relates to every other axiom. Yoneda makes this precise: if two axioms have identical relational profiles, they are the same axiom up to isomorphism, regardless of their surface syntax.

Adjoint Functor Theorem (Freyd). A functor G: D → C between complete categories has a left adjoint if and only if it preserves all limits and satisfies a “solution-set condition.” Informally: the cheapest way to add structure is guaranteed to exist whenever the forgetful map respects all ways of combining structure. Why it matters: Before searching for a “minimal JUB extension” of PET, check whether the conditions of this theorem are met. If they are, the extension exists and is unique up to isomorphism. If not, no such canonical extension exists — and any construction will involve arbitrary choices.

Mac Lane Coherence Theorem. In a monoidal category, every diagram built from associators, unitors, and their inverses commutes. Informally: once you know the basic structural isomorphisms, you never need to worry about which sequence of re-bracketings you use. Why it matters: When composing multiple model transformations (e.g., PET → JUB → compiled view → audience-depth projection), coherence guarantees the result is independent of the order of composition — if the transformations form a monoidal category. If they do not, order matters and the compiler must track it explicitly.

4. Common Pitfalls#

Confusing “arrow exists” with “functor exists.” A functor must preserve all composition and identities, not just map objects to objects. Finding that PET axioms have JUB counterparts does not establish a functor — you must also show that every derivation in PET maps to a valid derivation in JUB.

Forgetting functors are not bijections. A functor PET → JUB need not be injective or surjective on objects. Multiple PET axiom-groups may map to the same JUB group (information loss), or some JUB groups may have no PET preimage (genuine novelty). This is expected and informative, not a failure.

Treating isomorphism as equality. In category theory, isomorphic objects are interchangeable but not identical. Two formulations of the same axiom that are provably equivalent are isomorphic in the proof category — but replacing one with the other changes the formal text. Keep track of which equivalences you are invoking.

Ignoring size issues. Categories of “all sets” or “all categories” lead to Russell-type paradoxes. For matheology purposes, both models are small (finite axiom sets, finite theorem sets), so size is not a practical concern — but if you ever quantify over “all possible models,” you need a universe hierarchy or Grothendieck universes.

Over-abstracting. Category theory can formalize anything as a category. The question is whether doing so reveals structure that was invisible before. If the categorical formulation merely restates what you already know in opaque notation, it is adding complexity without insight. Test each categorical claim against a concrete matheology question.

5. Bridge to Matheology#

Alignment echoes as natural transformations. The observed echoes between pet-ax1 (Containment) and jub-ax15 (Human Agency) can be tested categorically. Define functors F, G: PET → JUB by two different alignment proposals. If a natural transformation α: F ⇒ G exists, the echoes are structurally related. If no such α exists, the echoes are coincidental — valuable diagnostic information.

HELL findings as a slice category. For each axiom A, the con-findings targeting A form a slice over A. The pro-findings are arrows in the opposite direction. The question “is axiom A adequately defended?” becomes: is the diagram of con-arrows → A ← pro-arrows a limit diagram (every attack has a matching defense that commutes)?

Model compilation as colimit. SISYF’s compilation of PET+JUB into a unified view is a colimit over the diagram PET ← Shared → JUB, where Shared is the common sub-structure. The universal property of the colimit guarantees that any other system receiving both PET and JUB factors uniquely through this compiled view — formalizing SISYF’s claim to be the canonical assembly.

5D label space as a functor category. The BEST naming architecture (model × element × version × depth × view) defines an index category. Content pages are a functor from this index to the category of RST documents. The compiler is a natural transformation between functors (source → compiled output).

New questions category theory enables:

  • Is the PET → JUB extension a free construction (left adjoint to forgetting JUB axioms)?

  • Do the 6 scriptural tradition mappings form a natural family, or are some traditions categorically closer to the model than others?

  • Is there a terminal object in the category of matheology models — a model that every other model maps to uniquely?