Glossary: canon, authority, non-response
This family groups the notions that bound the legitimacy of a response produced by an AI system. It addresses a central question: what can a model infer from partial signals, and under what conditions does non-response become the correct output?
Each entry links to: a canonical definition (if it exists), a framework (if applicable), and related pages for deeper understanding.
Quick access
- Canonical definitions
- Frameworks and applicable standards
- Clarifications (scope, limits, ambiguous cases)
- Doctrine (principles and positions)
Terms in the “canon, authority, non-response” family
Authority boundary
Explicit limit between what a source declares and what a model deduces. A clear boundary prevents normative extrapolations and makes the response enforceable.
- Definition: Authority boundary
Interpretability perimeter
The exact zone where a source authorizes inference, and where it does not. The perimeter reduces misinterpretations produced by over-interpretation.
- Definition: Interpretability perimeter
Canonical silence
Status where the absence of response is not a gap, but the normal effect of a canon that does not authorize inference or that imposes reservation.
- Definition: Canonical silence
Legitimate non-response
Non-response produced because no response can be formulated without violating a legitimacy limit (absence of evidence, unresolved ambiguity, authority conflict).
- Definition: Legitimate non-response
- Clarification: Plausible hypotheses, ungoverned inference, and forbidden plausibility
Authority conflict
Collision between two sources or truth frameworks that claim the same semantic zone. Without arbitration, the model “averages” or invents a compromise.
- Definition: Authority conflict
- Doctrine: Governed negation
Governed negation
Controlled mechanism for negating, refuting, or correcting a proposition without triggering a compensatory hallucination. Governed negation protects correction.
- Definition: Governed negation
- Doctrine: Governed negation: managing conflicts without hallucinating
Response conditions
Explicit constraints that determine when a response is authorized (and under which evidence, limits, and formats). They transform a “plausible response” into an enforceable response.
- Definition: Response conditions
- Framework: Enforceable response conditions for AI agents
Related frameworks and pages (recommended)
- Interpretive governance (definition)
- The framing role of interpretive legitimacy
- Relations: links, dependencies, and internal coherence
Previous page: Glossary: drifts and inertia
Next page: Glossary: evidence, audit, and observability
Phase 1 authority surfaces
This family now includes three priority canonical surfaces for authority, admissibility and refusal logic:
- Source hierarchy for source ordering.
- Answer legitimacy for response authorization.
- Silent delegation of authority for uncontrolled transfer of interpretive authority.
Phase 2 authority, refusal, and coherence-control terms
Interpretive authority
Interpretive authority identifies the locus that governs meaning. It prevents a retrieved or generated fragment from becoming governing authority by default.
- Definition: Interpretive authority
Authority ordering
Authority ordering decides which admissible authority prevails when several sources could govern the same claim. Without ordering, the model may choose the most retrievable or most coherent source rather than the governing source.
- Definition: Authority ordering
Interpretive perimeter
The interpretive perimeter is the boundary of authorized interpretation. It determines what may be inferred, asserted, summarized, refused or left silent under a declared authority.
- Definition: Interpretive perimeter
Mandatory silence
Mandatory silence is the required non-response when answering would cross authority, evidence, perimeter, version or commitment conditions. It is not missing content. It is a governed output.
- Definition: Mandatory silence
Inference prohibition
Inference prohibition prevents the model from deducing claims from silence, proximity, similarity, absence, examples or incomplete evidence. It is the operational form of “do not infer X from Y”.
- Definition: Inference prohibition
Unauthorized synthesis
Unauthorized synthesis appears when a response combines real fragments into a conclusion no authority authorized. It is one of the most common ways a plausible response becomes indefensible.
- Definition: Unauthorized synthesis
Manufactured coherence and surface coherence
Manufactured coherence is the process that smooths gaps, conflicts or missing authorities. Surface coherence is the visible appearance of order that results. Both must be tested against proof of fidelity.
- Definition: Manufactured coherence
- Definition: Surface coherence
Phase 10 routing layer: inference, arbitration, indeterminacy and fidelity
This page now routes inference-control questions toward the phase 10 canonical layer: interpretive error space, free inference, default inference, arbitration, indeterminacy, and interpretive fidelity.
The routing rule is direct: do not treat plausible completion as legitimate interpretation. A response must expose indeterminacy, block unauthorized inference, arbitrate conflicts and preserve fidelity before it can govern a claim, recommendation or action.
How to read this lexical family
This family defines the discipline of not answering beyond authority. It is not a vocabulary of refusal for its own sake. It is a vocabulary for deciding when the canon is strong enough, when the hierarchy of sources is sufficient, when a response condition is met and when silence is the more legitimate output.
The terms belong together because authority is rarely a single attribute. A source can be canonical for identity but not for pricing, current status, legal interpretation, service availability or technical execution. A response can therefore be supported by a source while still exceeding the source’s authority.
Typical misreadings
The most frequent error is to treat canonical material as unlimited permission to answer. A canonical source governs what it states, not every implication a model can generate from it. Governed negation, mandatory silence and inference prohibition exist to prevent that expansion.
Another error is to treat non-response as weakness. In this doctrine, a legitimate non-response may be the strongest possible answer when the source hierarchy is incomplete, the commitment boundary is crossed or the requested inference would create an unsupported claim.
Use in audit and routing
Use this family when evaluating whether a system should answer, qualify, defer, refuse or remain silent. It is especially useful for regulated claims, institutional statements, legal-adjacent interpretations, financial consequences, medical-adjacent content, HR contexts, contractual contexts and tool-mediated execution.
For routing, this family should support canonical definitions around interpretive authority, source hierarchy, authority boundary, response conditions and legitimate non-response. It should not be confused with general content moderation.