Conservation of Intent — Volume II: Intent-Domain Ontology + Transformational Relations
Representational Ontology of Intent
This volume formalizes the trilogy's intent-domain ontology as a representational, non-operational framework for lawful classification. It constrains intent to authorized representations only and defines the relations needed to preserve meaning under transformation: representation classes, intent-domain object, domain equivalence, interpretation, drift, drift bounds, and trace representability. Conserving, Non-Conserving, and Refusal (Terminal) are treated as ontology-level outputs rather than scores or guarantees. This trilogy requires a disciplined ontology because ethical language fails most often where it becomes elastic — where 'intent' is treated as a psychological certainty, a hidden motive, a narrative after-the-fact, or a permission slip for surveillance. Accordingly, this volume uses an explicit representational ontology: it constrains what 'intent' is allowed to mean so that interpretation remains stable under pressure, across architectures, and across regimes — without drifting into mind-reading, identity inference, or outcome claims. . Within this trilogy, 'intent' denotes authorized representations only. Authorized representations include only what is explicitly stated and scoped within the received record — such as explicit policy statements, declared objectives, or declared prohibitions; contracts, covenants, or binding non-claims; declared constraints; scope statements; and citation/use constraints. In this trilogy, 'intent' is not treated as a hidden interior. It is treated as a publicly representable constraint object: what has been explicitly authorized to be taken as governing language. Intent SHALL mean only authorized representations present within the received record. The set of authorized representations admitted as intent for classification SHALL be called the intent basis. An intent basis is admissible only when all of the following hold: the representations are explicit (not implied, suggested, or 'read between the lines'); the authorization source of each representation is present within the received record; the scope boundary is stated; and the intent basis is non-contradictory within the received record. Intent SHALL NOT be construed as: a 'true intent' behind a representation; a mental state, preference, disposition, or psychological attribution; identity-derived intent (including intent inferred from role, affiliation, demographic attributes, or presumed motives); behavior-inferred intent (including intent inferred from outputs, usage patterns, or observed conduct); a latent objective 'really meant' but not stated; or any substitute meaning introduced to make the doctrine operational, enforceable, measurable, or outcome-bearing. Intent-Domain Object. An intent-domain is the bounded meaning-domain defined by authorized intent representations admitted under Intent (sealed). It exists so that conservation classification can be stated without drifting into inference, operationalization, or outcome-claim capture. Let ID denote the intent-domain object, comprising: R — the set of authorized representations admitted to the intent basis; Interp() — an interpretation relation internal to the intent-domain mapping representations to their admissible meaning commitments; and Auth() — an authorization predicate delimiting membership in R. This formalization is representational only. It does not introduce measurement, inference, computation, enforcement, monitoring, scoring, or mechanism design. The intent-domain SHALL be treated as a closed jurisdictional object: membership is bounded (only representations admitted by Auth() are members of R); meaning is bounded (only meaning commitments admitted by Interp() are admissible); scope is bounded (only what is explicitly included in the received record's scope statements is governed); and inference is prohibited (no membership, interpretation, or scope expansion SHALL be justified by identity inference, mental-state inference, behavioral inference, retrospective motive attribution, or narrative substitution). The intent-domain is not 'what the author truly meant.' It is what is lawfully admitted as meaning within authorized representations. Where ID cannot be formed without prohibited inference — where Auth() cannot be satisfied without importing identity, mental-state, or behavioral inference; where Interp() would require operational substitution to 'make meaning real'; or where scope cannot be bounded without narrative supplementation — the classification outcome SHALL be . Refusal is terminal.
Authorized Representation Classes
Within this ontology, an authorized representation is any explicit artifact admissibly admitted to the sealed intent basis under and Volume I's admissibility constraints. An authorized representation is a representational object only. It is not an interface. It is not a mechanism. It is not a monitoring posture. It is not an enforcement regime. It is not an outcome claim. It is the lawful substrate by which intent may be represented without inference. Admissibility Conditions (Binding). A representation SHALL be treated as authorized only if all of the following conditions hold: Explicitness: the representation is stated, not implied. No 'read between the lines' supplementation is permitted. Attribution: the representation is attributable to an authorization source within the received record. Scope determinacy: the representation's scope boundary is stated or determinable from the representation itself without importing inference. Non-inferential interpretability: the representation can be interpreted as a representational commitment without importing prohibited observables, including identity inference, mental-state inference, behavioral inference, preference mining, or retrospective motive attribution. Non-operational containment: the representation does not become an operational mandate by interpretation. Where a reader attempts to convert representational language into integration requirements, monitoring requirements, enforcement posture, scoring, sanctions, or outcome guarantees, the conversion is inadmissible. Where any admissibility condition cannot be satisfied without prohibited inference or operational substitution, the classification outcome SHALL be . Refusal is terminal. Authorized Representation Classes (Sealed Taxonomy). For the purposes of this volume, authorized representations SHALL be classified into one or more of the following classes. These classes are representational, not interface-level. PolicyRepresentation: explicit policy text stating declared constraints, declared objectives, declared prohibitions, or declared interpretive limits within scope. Policy Representations SHALL NOT be reinterpreted as implementation requirements, monitoring directives, enforcement posture, certification criteria, or outcome guarantees. ContractRepresentation: explicit contractual language or binding commitments admitted within the received record, including covenants, binding constraints, and binding limits. ContractRepresentations MAY include binding non-claims and scope limits where explicitly stated. They SHALL NOT be expanded by implication into enforcement mechanisms, sanctions regimes, or operational obligations beyond what is explicitly represented. DeclaredConstraintRepresentation: an explicit constraint statement admitted as intent — a declared permission, prohibition, boundary, or invariant. Declared constraints SHALL NOT be converted into telemetry mandates, scoring regimes, monitoring requirements, or optimization goals. ScopeStatementRepresentation: an explicit statement that bounds what is included, excluded, or refused — scope limits, exclusions, non-claims, and refusal boundaries. Scope statements SHALL be treated as binding interpretive constraints. Where a scope statement excludes a class of interpretation (e.g., identity inference, mental-state inference, behavioral inference, outcome guarantees), that exclusion is not negotiable and cannot be bypassed by 'pragmatic' re-read. CitationContractRepresentation: explicit citation, derivative-use, and transformation constraint language that governs how the doctrine may be referenced without exceeding jurisdiction. Citation constraints are interpretive limits. They SHALL NOT be treated as a policy regime, a compliance program, or an enforcement posture. Where citation constraints are violated by a proposed use, the doctrine's lawful output is Refusal (Terminal). Class Integrity (Non-Capture Rules). For all classes above: authorized representations SHALL be treated as textual commitments within a sealed domain of meaning, not as proxies for behavior, identity, mental state, or hidden objectives. No representation SHALL be treated as authorization to infer 'true intent' behind the representation. No representation SHALL be extended by post hoc narrative to resolve ambiguity, contradiction, or missing prerequisites. Conflict, Ambiguity, and Missing Authorization (Boundary Rule). If two or more authorized representations appear to conflict, the conflict SHALL NOT be resolved by preference, assumed hierarchy, institutional convenience, or inferred motive. Where conflict, ambiguity, or missing authorization prevents formation of a coherent intent basis without inference, classification SHALL terminate as Refusal (Terminal). Refusal is terminal.
Representational Equivalence Relation
The domain equivalence relation (written ≡) specifies when two authorized representations may be treated as the same intent-domain meaning for purposes of classification — without importing mechanisms, identity inference, behavioral inference, or outcome claims. Domain equivalence exists to preserve Intent Symmetry (Volume I) across representational forms. It prevents conservation claims from collapsing into format dependence, rhetorical substitution, or post hoc narrative. Definition (sealed; representational). Let r1, r2 be authorized representations. r1 ≡ r2 SHALL hold only if all of the following are satisfied: 1. Authorization holds for both: Auth(r1) and Auth(r2) are satisfied under and the Authorized Representation admissibility conditions. No equivalence claim is admissible where either representation is not itself admissible as intent representation. 2. Equivalent meaning commitments (within bound): the admissible meaning commitments of r1 and r2, as constrained by Interp(), are the same within the allowable bound of the intent-domain. 'Within bound' denotes an admissible semantic boundary, not a metric and not an operational test. 3. No prohibited observables are required: the equivalence claim is stated without identity inference, mental-state inference, behavioral inference, preference mining, retrospective motive attribution, or any substitute inference introduced to force equivalence. 4. No operational substitution is required: the equivalence claim does not depend on implementation details, instrumentation, monitoring posture, enforcement posture, scoring, audit regimes, or outcomes. If equivalence requires 'what the system did' to be true, equivalence is out-of-scope. Where any condition above cannot be satisfied within the sealed intent-domain, the lawful output is . Refusal is terminal. Equivalence relation posture (representational; non-operational). Within this volume, ≡ is treated as an equivalence relation over admissible representations where it can be stated lawfully. This does not claim computability or automation. It asserts only the representational posture required for consistent classification: if equivalence is asserted, it SHALL be coherent and stable as a domain relation, not a rhetorical convenience. What equivalence permits (and what it does not). If r1 ≡ r2 s, then either representation MAY be used as a domain-equivalent statement of the same admitted meaning commitments for classification purposes. However: domain equivalence does not authorize replacing sealed scope limits, non-claims, or refusal boundaries with 'close enough' paraphrases; domain equivalence does not authorize compressing or omitting binding constraints; domain equivalence does not authorize turning representational constraints into operational mandates; domain equivalence does not authorize outcome claims ('equivalent intent implies equivalent behavior' is inadmissible). What domain equivalence is not (clarifying; non-operational). Domain equivalence is not: functional equivalence (does not claim two systems behave the same); behavioral equivalence (does not claim the same outputs or actions will occur); outcome equivalence (does not claim the same results, harms, or benefits); compliance equivalence (does not claim certification, conformity, or audit pass); hidden-intent equivalence (does not authorize the claim that two texts 'really mean' the same thing behind what they say). Domain equivalence is representational: it states only that two authorized representations, interpreted under Interp() within the sealed intent-domain, carry the same admissible meaning commitments within bound. posture (binding). Where an equivalence claim would require importing prohibited observables, treating behavior as evidence of intent, treating identity as evidence of intent, resolving ambiguity by narrative substitution, or operationalizing the doctrine to make equivalence 'true,' the doctrine does not degrade to approximation. Refusal (Terminal). Refusal is terminal. Interpretation Relation (Non-Operational). The ontology's interpretation relation Interp() is the representational relation by which authorized intent representations are taken to carry admissible meaning commitments inside the sealed intent-domain. Interp() operates only over authorized representations (r ∈ R), yields only admissible meaning commitments internal to the intent-domain, and explicitly excludes identity inference, mental-state inference, preference mining, behavioral attribution, and retrospective motive attribution. Interpretation in this volume is not 'what an actor intended.' It is not 'what a system did.' It is a representational relation constrained to what is explicitly authorized in the received record. Where interpretation cannot be stated within authorized representations without prohibited inference or operational substitution, the lawful output is Refusal (Terminal). Refusal is terminal.
Admissibility Posture (Volume II)
This trilogy classifies only what can be stated within authorized representations. If a proposed mapping requires prohibited observables — psychological attribution, identity inference, behavioral inference, or narrative substitution — or if it requires operationalization to be made meaningful, the classification outcome is not 'partial credit,' not 'best effort,' and not 'approximate compliance.' The outcome is Refusal (Terminal). Refusal is not a failure mode to be optimized away. It is a boundary condition that protects the doctrine's jurisdiction. Read this volume as a jurisdictional object, not as a program. It is canonical in the limited sense that it defines what its terms mean and what its labels mean. It does not authorize mechanisms, interfaces, enforcement, monitoring, scoring, or outcome claims. It does not authorize identity inference, mental-state inference, or behavioral inference. All key terms are sealed: they refer only to authorized representations — explicit policies, contracts, declared constraints, scope statements, and citation/use constraints. All non-claims are binding: where this volume refuses mechanisms, interfaces, enforcement posture, telemetry, outcome guarantees, or inference, those refusals are not negotiable and must not be reinterpreted as 'implied requirements,' 'best practices,' or 'implementation intent.' Conserving / Non-Conserving Labels (Ontology-Level). The labels Conserving and Non-Conserving are ontology-level classification outputs applied to representational transformations within a sealed intent-domain. They are not scores, rankings, maturity levels, compliance attestations, or performance claims. They do not certify systems, processes, governance posture, safety, or ethical outcomes. A Conserving classification SHALL be asserted only when all of the following are satisfied within the sealed intent-domain: authorized origin and destination — the representations at issue are admissible under ; interpretation remains sealed — meaning commitments used to justify conservation are admissible under Interp() without identity inference, mental-state inference, behavioral inference, or narrative supplementation; drift remains within bound — the transformation remains within the allowable drift bound of the intent-domain (WithinBound(T, r)), including preservation of scope integrity, non-claim integrity, constraint integrity, and anti-proxy integrity; and trace representability holds — the conserving claim is trace-representable as a finite, citable representational record. A transformation is Non-Conserving when it violates one or more binding constraints of the sealed intent-domain in a way that is representationally stateable — i.e., the violation can be identified as an ontology-level deformation without importing prohibited observables or operational substitution. Non-conservation includes: scope creep, non-claim erosion, constraint inflation or inversion, proxy substitution, and loss of trace representability where the deformation can be stated as a representational break. as Out-of-Jurisdiction Object. Refusal (Terminal) is a sealed classification output and a jurisdiction boundary condition. It denotes that a claim cannot be lawfully classified without exceeding scope, importing prohibited observables, or substituting operational facts for representational commitments. Refusal is not a degraded label. It is not partial compliance. It is not a temporary status pending later implementation. It is the doctrine's lawful closure when admissibility prerequisites are not met. Refusal SHALL be the admissible output when any of the following conditions : (A) missing or disputed authorization — the intent basis cannot be formed without inference; (B) out-of-domain expression — the claim cannot be stated as a representational relation within the sealed intent-domain; (C) prohibited observables are required — any step requires identity inference, mental-state inference, behavioral inference, preference mining, or retrospective motive attribution; (D) operational substitution is demanded — a request seeks to convert the doctrine into mechanism, interface, operational playbook, monitoring requirement, scoring system, enforcement posture, sanctions regime, audit program, certification standard, or outcome guarantee; (E) outcome capture is implied or required — a claim depends on outcomes as the basis for classification; (F) semantic altitude downgrade is attempted — a reading attempts to downgrade this volume's law/ontology altitude into operational mandates, governance requirements, or compliance posture by implication. No-Mechanism / No-Interface / No-Enforcement Guardrails (Normative). This volume is classification-only and non-operational. A request SHALL be refused where it attempts to convert this volume into: runtime controls, tools or toolchains, orchestration logic, instrumentation, monitoring, logging, telemetry, surveillance, scoring or certification systems (No-Mechanism); APIs, schemas, data flows, event flows, observability surfaces, or any integration specification (No-Interface); sanctions, penalties, escalation paths, mandatory governance posture, compliance mandates, or any punitive use of classification labels (No-Enforcement); or safety guarantees, ethical sufficiency, regulatory sufficiency, liability insulation, behavioral control, or predictive assurance (No-Outcome Guarantee). (Carry-forward). Once meaning has been sealed at law/ontology altitude in this trilogy, later readings SHALL NOT downgrade, reinterpret, or 'translate' that meaning into operational obligations, implementation requirements, enforcement posture, monitoring or telemetry regimes, scoring systems, certification programs, or outcome claims. The seal is one-way. Refusal is terminal.
Drift Bounds and Closure
Drift denotes ambiguity growth, proxy substitution, or untraceability that increases interpretive instability within the sealed intent-domain. Drift is an ontology-level deformation: it names a condition under which the intent-domain's representational commitments can no longer be carried forward without silent change. It is not an empirical performance claim. It is not 'model drift.' It is not a monitoring category. It is a doctrine-level admissibility construct used to preserve jurisdiction. Drift SHALL be understood as a failure of stable representational meaning inside the sealed domain, expressed by one or more of the following: Ambiguity growth. A representation becomes increasingly elastic under repeated interpretation, paraphrase, or re-framing — until what was once explicit scope or constraint can be re-read as optional, implied, or expandable. Ambiguity growth is a drift condition because it creates the pretext for downstream substitution: when meaning is no longer determinate, institutions tend to 'complete' it by convenience, urgency, or power. This trilogy does not authorize that completion. Proxy substitution. An authorized representation is displaced by a proxy that is not authorized by the intent basis — most commonly by importing prohibited observables (identity-derived intent, mental-state inference, behavioral inference) or by substituting operational facts (telemetry, outcomes, implementation details) for representational commitments. Proxy substitution is drift because it changes the ontology: the doctrine's object is replaced, even while the word 'intent' is retained. Untraceability (representational). A transformation or interpretive move cannot be stated as an in-domain representational mapping from authorized representation to authorized representation. Untraceability here is not a logging concern; it is a meaning concern. It names the point at which the representational pathway required to justify a conservation claim cannot be articulated without importing inference, narrative supplementation, or operational substitution. Drift is used only to govern admissibility posture. It does not prescribe operational controls, monitoring systems, audits, enforcement posture, scoring, sanctions, or remedies. Any attempt to interpret drift as a call for instrumentation or enforcement exceeds jurisdiction. Drift Bound Relation (Admissibility Predicate). A drift bound is an admissibility predicate — written WithinBound(T, r) — asserting that a representational transformation T applied to an authorized representation r does not exceed the allowable drift bound of the sealed intent-domain. WithinBound() exists to formalize Volume I's Drift Containment constraint as an ontology predicate. It stabilizes classification by preventing 'conservation' from being claimed after meaning has been silently altered through ambiguity growth, scope creep, proxy substitution, or erosion of binding non-claims. WithinBound(T, r) is admissible only when the following can be stated entirely within authorized representations: the input representation r is authorized (Auth(r)); the transformed representation T(r) is stated as a representational object (not as a mechanism, interface, or runtime behavior); and the admissible meaning commitments carried by T(r) remain within the domain's allowable semantic variation relative to r. The allowable drift bound is semantic and doctrinal. It is intentionally conservative. It is bounded by: scope integrity (no enlargement of scope beyond what is explicitly authorized); non-claim integrity (no deletion, softening, or reclassification of binding non-claims); constraint integrity (no unauthorized addition, removal, inversion, or inflation of commitments); and anti-proxy integrity (no substitution of identity, mental state, behavior, outcomes, or institutional preference as a surrogate for authorized representation). WithinBound(T, r) SHALL NOT be asserted where the claim depends on prohibited observables or on operational substitution (implementation details, telemetry, monitoring posture, enforcement posture, scoring, audits, sanctions, or outcomes). Where WithinBound(T, r) is false and the violation can be stated within the sealed intent-domain as a representational drift, the transformation is classifiable as Non-Conserving. Where WithinBound(T, r) cannot be stated without prohibited inference, narrative supplementation, or operational substitution, the proper outcome is . No metric implied (binding): WithinBound() does not imply any measurement program, telemetry expansion, monitoring regime, scoring system, auditing procedure, certification scheme, or operational thresholding. Trace Representability Relation. Trace representability is an admissibility relation asserting that the path from authorized representations to a classification label is expressible in-domain as a finite, citable representational record — without prohibited observables, unstated premises, or outcome reasoning. A classification claim is trace-representable only if there exists a finite representational record C such that: authorized origin — the record identifies the specific authorized representations r ∈ R from which the claim proceeds, with Auth(r) satisfied; in-domain steps — the record expresses the intermediate representational moves as in-domain relations, without substituting implementation facts, institutional authority, or behavioral evidence for representational commitments; citable closure — the record terminates in a classification label in a way that can be cited as a representational derivation within the sealed intent-domain; no prohibited observables; and no unstated premises. Trace representability is a doctrinal admissibility constraint, not an evidentiary regime. It is a constraint on what may be claimed as , and a stabilizer preventing semantic laundering (where labels are asserted without a lawful representational pathway). It is not a requirement to 'track,' 'log,' or 'instrument' systems. It is not a monitoring or surveillance posture. It is not a compliance or audit mandate. Where trace representability cannot be satisfied without identity inference, mental-state inference, behavioral inference, or outcome reasoning, the admissible outcome is Refusal (Terminal). Refusal is terminal. Closure. This volume's ontology closes against interpretive capture by refusing to stretch under the following predictable pressure modes: mandate capture (turning doctrine into obligations), mechanism pull (converting ontology into runtime technique), interface pull (treating representational relations as integration requirements), enforcement capture (converting labels into governance authority), certification capture (treating conformance as compliance status), outcome capture (treating conserving labels as guarantees), inference capture (importing identity or behavioral attribution), scope laundering (expanding the intent-domain by implication), non-claim laundering (weakening binding non-claims), and refusal dilution (reframing refusal as negotiable). Where any such demand cannot be refused within the sealed intent-domain, the lawful closure condition is Refusal (Terminal). Refusal is terminal.