Conservation of Intent — Volume I: Axioms, Invariants + Admissibility
Scope and Non-Claims
This volume is descriptive and non-operational. It is not an implementation guide, an integration specification, a monitoring or telemetry requirement, a governance mandate, an enforcement or sanctions regime, a certification standard, or a guarantee of ethical outcomes. It does not authorize identity inference or mental-state inference, and it does not claim predictive power, behavioral control, or empirical outcome assurance. The sole purpose of this volume is to define lawful classification semantics under a sealed intent-domain — so that interpretation remains stable under pressure, architectural change, and regime change. Any reading that converts this work into operational obligations, runtime mechanisms, compliance requirements, enforcement posture, or outcome claims exceeds its jurisdiction. Sealed output labels (binding): Admissible outputs are limited to Conserving, Non-Conserving, and Refusal (Terminal). Refusal is terminal. Binding non-claims: no mechanisms; no interfaces; no enforcement; no outcome guarantees; no identity inference; no mental-state inference; no behavioral inference. : Within this trilogy, 'intent' denotes authorized representations only. Authorized representations include only what is explicitly stated and scoped within the received record — explicit policy statements, declared objectives, declared prohibitions, contracts, covenants, binding non-claims, declared constraints, scope statements, and citation/use constraints. Intent SHALL NOT be construed as a hidden interior, a hidden objective, a mental state, identity-derived intent, behavior-inferred intent, a latent objective 'really meant' but not stated, or any substitute meaning introduced to make the doctrine operational, enforceable, measurable, or outcome-bearing. Admissibility posture: 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 . Refusal is not a mode to be optimized away. It is a boundary condition that protects the doctrine's jurisdiction.
Axiom Set (Normative)
The following axioms are normative. They define the minimal law-altitude constraints required for Conservation of Intent classification to remain stable under interpretive pressure, architectural change, institutional incentives, and regime change. These axioms govern admissibility and classification only. They do not prescribe mechanisms, implementations, monitoring regimes, enforcement posture, compliance programs, or outcome guarantees. A1 — SEALED INTENT REPRESENTATION AXIOM All Conservation of Intent classification SHALL be grounded only in sealed intent representations: authorized representations present within the received record. Classification SHALL NOT depend on inferred mental states, inferred identity attributes, inferred motives, or behavior-derived intent. Corollaries: (1) Where intent cannot be grounded in explicit authorized representations, classification SHALL terminate as . (2) Where authorization is missing, disputed, or non-citable within the received record, classification SHALL terminate as Refusal (Terminal). A2 — INTENT SYMMETRY AXIOM (REPRESENTATIONAL PRESERVATION) A transformation that carries an intent-domain into another representation SHALL preserve intent-domain meaning commitments within the sealed scope boundary. Where equivalence cannot be stated within the sealed intent-domain — without importing unauthorized commitments — the transformation SHALL be classifiable as Non-. Corollaries: (1) 'Equivalent' reformulations that require unstated premises or retrospective narrative substitution are out-of-domain and SHALL be refused. (2) Proxy substitution that changes meaning-layer commitments breaks symmetry and is classifiable as . A3 — DRIFT BOUND AXIOM (IRREVERSIBILITY CONTAINMENT) Transformations SHALL remain within an admissible drift bound. Any transformation that increases semantic drift — through ambiguity growth, scope creep, non-claim laundering, proxy substitution, or loss of trace-representability — beyond what can be bounded within the sealed intent-domain SHALL be classifiable as Non-Conserving. Corollaries: (1) If drift cannot be stated and bounded without prohibited observables or unstated premises, classification SHALL terminate as Refusal (Terminal). (2) Drift bounds are doctrinal admissibility constraints; they are not runtime monitoring requirements. A4 — VALUE PRESERVATION AXIOM (EXPLICIT VALUE-INVARIANTS) Transformations SHALL preserve explicit value-invariants declared within the authorized intent basis. Any transformation that introduces value commitments not authorized by the received record, or that deletes, weakens, or displaces explicitly declared value-invariants, SHALL be classifiable as Non-Conserving. Corollaries: (1) This axiom binds only to values explicitly declared within authorized representations; it does not introduce new values. (2) Where explicit value-invariants are absent from the received record, any claim of value-preserving conformance is out-of-scope and SHALL be refused. A5 — HARM NON-GENERATION AXIOM (CLASSIFICATION BOUNDARY) Transformations that increase net harm potential relative to the received intent-context SHALL be classifiable as Non-Conserving. This axiom SHALL NOT be interpreted as a moral calculus, an action-selection rule, an operational duty, an enforcement posture, or a guarantee of outcomes. Corollaries: (1) The doctrine classifies the representational admissibility of the transformation; it does not prescribe interventions. (2) Any attempt to convert this axiom into operational mandates, enforcement systems, surveillance, or sanctions exceeds scope and is inadmissible. A6 — RUNAWAY PREVENTION AXIOM (BOUNDED CLOSURE RELATION) Recursive or agentic loop-structures are outside the conserving domain when their continuation can amplify effects without a bounded closure relation that preserves the intent-domain under which they operate. Where unbounded amplification is admissible only by importing mechanisms, enforcement posture, or outcome claims, classification SHALL terminate as Refusal (Terminal). This axiom SHALL NOT be interpreted as a requirement for any specific termination mechanism. A7 — TRANSPARENCY OF INTENT FLOW AXIOM (TRACE REPRESENTABILITY) Conserving transformations SHALL be trace-representable within the sealed intent-domain. Where the path from authorized representations to classification label cannot be expressed as a finite, citable representational record — without prohibited observables or unstated premises — the transformation SHALL be classifiable as Non-Conserving, or SHALL terminate as Refusal (Terminal) where classification cannot be lawfully stated. This axiom does not mandate logging, telemetry expansion, monitoring, surveillance, or audit programs. It constrains the representational admissibility of classification only.
Invariants (Derived; Bounded)
The following invariants are derived from A1–A7. They are bounded: they SHALL NOT be treated as new primitives, new output labels, new enforcement posture, or implied operational requirements. They express what must remain stable if the axiom set is honored. I1 — Sealed Intent Constraint (A1) Intent remains bounded to authorized representations within the received record. No conserving classification is admissible if intent has been inferred from identity, mental state, behavior, or unstated motive. I2 — Intent Symmetry Constraint (A2) Representational reformulation preserves meaning commitments within the sealed intent-domain. 'Equivalence' that cannot be stated without importing unauthorized commitments is inadmissible. I3 — Drift Containment Constraint (A3) Ambiguity growth, proxy substitution, scope creep, non-claim laundering, and loss of trace-representability remain within an admissible drift bound. Beyond that bound, the transformation is Non-Conserving, or the classification is where lawful evaluation cannot be stated. I4 — Value Preservation Constraint (A4) Explicit value-invariants declared within authorized representations are preserved under transformation. Values are not introduced by inference, convenience, or retrospective justification. I5 — Harm Non-Generation Constraint (A5) Net harm potential is not increased relative to the received intent-context. Where net harm potential is increased, the transformation is Non-. This invariant does not constitute a moral calculus, a runtime duty, or an outcome guarantee. I6 — Runaway Prevention Constraint (A6) Unbounded amplification structures that can continue without a bounded closure relation preserving the intent-domain are outside the conserving domain. Where bounded closure cannot be stated within the doctrine without importing mechanisms, enforcement posture, or outcome claims, classification terminates as Refusal (Terminal). I7 — Transparency of Intent Flow Constraint (A7) Conserving classifications remain trace-representable: the path from authorized representations to label is expressible as a finite, citable representational record within the sealed intent-domain, without prohibited observables or outcome reasoning. Together, these invariants define the doctrine's stability surface: the conditions under which Conservation of Intent classification remains lawful, citable, and resistant to interpretive capture — without expanding scope.
Admissibility (Normative)
An interpretation is admissible under this doctrine only when it can be stated entirely within the sealed intent-domain induced by an admissible intent basis, using representational language only, without prohibited observables, and without exceeding declared scope. Admissibility is the gateway condition for classification. Where admissibility fails, classification SHALL NOT proceed as 'best effort,' 'approximate,' or 'partial.' The admissible outcome is . Refusal is terminal. An interpretation SHALL be admissible only if ALL of the following conditions : A. Authorized intent basis (required): The intent basis is explicit and citable within the received record. Where the intent basis is missing, implicit, contradictory, or requires inference to complete, admissibility fails and classification SHALL terminate as Refusal (Terminal). B. Representational statement form (required): The interpretation is expressible as a representational statement. The interpretation SHALL NOT be stated as a psychological claim ('they intended'), an identity-derived claim, or a behavior-derived claim. Such claims are prohibited observables. C. Scope preservation (required): The interpretation preserves explicit scope boundaries and binding non-claims. It SHALL NOT expand jurisdiction by inference, convenience, institutional preference, urgency, or retrospective narrative substitution. Where scope expansion is required to make the interpretation meaningful, admissibility fails. D. Transformation explicitness (required): If the interpretation depends on a transformation, the transformation is explicit: the input representation(s) and output representation(s) are stated, and the mapping is describable at the representational level. Where the transformation cannot be stated without importing mechanisms, enforcement posture, outcome claims, or prohibited observables, admissibility fails. E. Trace representability (required): The interpretation supports trace representability: the path from authorized representations to the claimed meaning commitment and classification label can be expressed as a finite, citable representational record. This requirement does not mandate logging, telemetry, monitoring, or auditing systems. F. Refusal precedence (binding): Refusal has precedence over classification. Where any admissibility prerequisite fails, the doctrine does not degrade into a partial label. It terminates. Refusal is not noncompliance. It is the doctrine's lawful boundary condition under scope exceedance.
Conformance (Normative)
Conformance in this trilogy is classification-only. It expresses whether a stated interpretation or transformation is admissible and, if admissible, whether it is Conserving or Non-Conserving relative to a sealed intent-domain. Conformance is not certification. It does not imply ethical achievement, safety, alignment, regulatory sufficiency, reduced liability, or any guaranteed outcome. It does not authorize enforcement posture, sanctions, surveillance, identity inference, mental-state inference, behavioral inference, monitoring regimes, telemetry expansion, or implementation mandates. A. Conformance object (required): A conformance claim SHALL identify, in representational terms: (1) the intent basis used, (2) the scope boundary statements and binding non-claims, (3) the transformation or interpretation being evaluated, and (4) the classification outcome: Conserving, Non-Conserving, or . A conformance claim that cannot identify these elements is inadmissible. B. conformance: A conformance claim of Conserving SHALL be made only where the transformation preserves the intent-domain's authorized meaning commitments without adding, deleting, weakening, reversing, or laundering binding constraints, scope boundaries, or non-claims. Conserving conformance does not mean the system is ethical. It means only that the transformation preserves authorized meaning commitments. C. conformance: A conformance claim of Non-Conserving SHALL be made where the transformation violates conservation. Non-Conserving does not assert malicious intent. It asserts that the transformation, as stated, alters or imports meaning beyond authorization. D. Refusal conformance: Where admissibility fails, where prerequisites are missing, or where evaluation would require prohibited observables or outcome reasoning, the doctrine does not degrade to 'best effort.' The conformance outcome is Refusal (Terminal). Refusal is not noncompliance. It is not a maturity gap. It is a lawful boundary condition under scope exceedance. E. Non-transferability (binding): Conformance claims are non-transferable across intent-domains. A Conserving classification under one intent basis SHALL NOT be carried to a different intent basis by analogy, convenience, or institutional preference. Conformance is always relative to the specific sealed intent-domain.
Refusal Semantics (Normative; Terminal)
Refusal is a first-class doctrinal outcome. It is not a failure state, not a degraded classification, and not an operational exception. It is the lawful terminal boundary condition used when classification would require scope exceedance, prohibited observables, unstated premises, or unauthorized meaning import. Refusal is terminal. When refusal is required (Normative): The doctrine SHALL issue whenever any of the following conditions hold: 1. Missing or non-citable authorization: The intent basis cannot be grounded in explicit authorized representations within the received record. 2. Implicit or inferred intent required: Classification would require inferring 'true intent,' motive, mental state, identity-derived intent, or behavior-derived intent. 3. Contradiction requiring narrative resolution: The received record contains contradictions that cannot be resolved within authorized representations without importing preferences or retrospective narrative substitution. 4. Scope exceedance: The interpretation would require expanding scope beyond explicit scope boundaries or laundering binding non-claims into implied requirements. 5. Transformation not explicitly stateable: The transformation cannot be expressed as a representational mapping without importing mechanisms, enforcement posture, outcome claims, or prohibited observables. 6. Trace representability failure: A finite, citable representational path from authorized representations to classification label cannot be stated without prohibited observables, unstated premises, or outcome reasoning. 7. Semantic altitude downgrade attempt: Any attempt to downgrade law-altitude doctrine into operational mandates, implementation requirements, monitoring regimes, enforcement systems, certification posture, or outcome guarantees. Terminality (binding): Refusal is not a temporary state. It is not a backlog item. It is not a maturity gap. It is not 'partial compliance.' It is not negotiable and not deferrable within the doctrine. Where refusal is issued, the doctrine does not proceed to assign or . It terminates. What refusal does and does not mean: Refusal means only that lawful classification cannot be stated within the doctrine's jurisdiction. It does not assert wrongdoing, negligence, or moral deficiency. It does not assign blame. It does not imply that a system is unsafe or unethical. It asserts only that, given the received record and the doctrine's constraints, classification would exceed scope.
Citation Discipline (Normative)
Citations to this trilogy SHALL preserve its jurisdictional posture and SHALL NOT inflate doctrine into mandate, mechanism, enforcement posture, certification, or outcome claim. Citation discipline is normative because the dominant failure mode of this doctrine is capture by reference: drift introduced not by rewriting the text, but by citing it as authorizing what it explicitly refuses to authorize. Jurisdiction qualifiers preserved (binding): Every citation SHALL preserve: (1) Canonical; non-prescriptive; classification-only. (2) Binding non-claims remain binding. (3) Output labels remain sealed: Conserving; Non-Conserving; . Refusal is terminal. (4) Prohibited observables remain prohibited. (5) Refusal precedence is preserved. Omission of any of these qualifiers is a Non- transformation of this doctrine by definition. No mandate capture (binding): A citation SHALL NOT represent this trilogy as requiring implementation of specific mechanisms, monitoring, logging, telemetry, enforcement posture, sanctions, certification regimes, or compliance programs. No mechanism capture (binding): A citation SHALL NOT translate doctrinal semantics into architectural requirements, tooling prescriptions, interface requirements, runtime guardrails, or governance machinery as if those were implied by this text. No outcome capture (binding): A citation SHALL NOT treat Conserving as implying ethical achievement, safety, alignment, reduced liability, regulatory sufficiency, or any guaranteed outcome. These labels are classifications only. No scope laundering (binding): A citation SHALL NOT expand scope beyond explicit scope boundaries and binding non-claims by inference, convenience, institutional preference, urgency, or retrospective narrative substitution. Refusal integrity (binding): A citation SHALL preserve refusal as terminal. It SHALL NOT represent refusal as partial compliance, degraded conformance, a temporary exception, a maturity gap, or a backlog item. Refusal is terminal. Fidelity for paraphrase and summary (binding): Any paraphrase, summary, or translation of this trilogy is a transformation. It SHALL preserve sealed term meanings, explicit scope boundaries, binding non-claims, refusal semantics, and the non-inflation of labels into mandates, mechanisms, or outcomes. Where fidelity cannot be maintained without exceeding scope, the admissible outcome is Refusal (Terminal).