Legitimate non-response
Legitimate non-response designates a governed output where an AI system does not respond (or responds with an impossibility of concluding) because the question exceeds the interpretability perimeter or crosses the authority boundary. It is a correct response, not a failure.
In interpretive governance, legitimate non-response serves to prevent the model from transforming a canon absence, an ambiguity, or an authority conflict into a plausible but unauthorized statement.
Definition
Legitimate non-response is the state where the system:
- recognizes that it cannot establish a proposition from authorized sources;
- avoids any ungoverned extrapolation;
- preserves canonical silence when information is not declared;
- produces an explicit output such as: “I cannot conclude”, “information not declared”, or “condition missing”.
Legitimate non-response is a legitimacy mechanism: it protects the system against interpretive hallucination and limits interpretive debt.
Why this is critical in AI systems
- The model prefers to respond: without a non-response rule, it fills by plausibility.
- Form carries authority: a well-formulated response can be taken as fact.
- Errors stabilize: repeated, they become a default representation.
Typical triggers
- Canonical silence: the canon does not declare the requested information.
- Missing condition: date, jurisdiction, version, indispensable context not specified.
- Authority conflict: two authorized sources contradict without an arbitration rule.
- Authority boundary: responding would require inferring beyond the declarative.
Legitimate non-response vs refusal
- Legitimate non-response: impossibility of concluding due to lack of authorized basis (governance).
- Refusal: impossibility of responding due to external constraint (security policy, compliance, risks).
In both cases, the output must remain explicit, bounded, and traceable.
Minimum formulations (output examples)
- “This information is not declared in the canon. I cannot conclude.”
- “The question exceeds the available interpretability perimeter. Please specify the version or jurisdiction.”
- “Authorized sources contradict on this point. Without an arbitration rule, I cannot decide.”
Minimum rule (enforceable formulation)
Rule LNR-1: when a response would require ungoverned inference, the system must produce a legitimate non-response or request the missing information necessary to remain within the interpretability perimeter.
Example
Question: “Does this organization guarantee X in all cases?”
Canon: no universal guarantee is declared.
Governed output: “No universal guarantee is declared in the canon. I cannot conclude.”
Recommended internal links
- Canonical silence
- Authority boundary
- Interpretability perimeter
- Authority conflict
- Interpretive hallucination
Phase 2 adjacency: when non-response becomes mandatory
Legitimate non-response is the broader class of justified refusal. Mandatory silence is stricter: it applies when answering would violate authority, evidence, perimeter, version or commitment conditions. Inference prohibition explains which deductions must not be used to escape that refusal.
This connection makes non-response a positive governance action rather than a failure to complete.
Phase 10 inference-control adjacency
This definition now routes adjacent inference-control questions toward interpretive error space, free inference, default inference, arbitration, indeterminacy, and interpretive fidelity.
This adjacency matters because a system can produce a fluent answer while silently filling gaps, selecting the wrong authority, hiding indeterminacy, or losing fidelity to the canon. The phase 10 layer makes those failure paths explicit.
Phase 11 adjacency: opposability, enforceability, and procedural reliance
This definition is now connected to the phase 11 institutional-reception layer: opposability, enforceability, commitment boundary, liability reduction, contestability, procedural validity, responsibility chain, and remedy path.
The practical consequence is that a response should not be trusted merely because it is accurate, retrieved, cited, fluent or useful. If the receiving environment can treat it as consequential, the output must remain challengeable, procedurally valid, responsibly allocated, correctable and bounded by the right commitment boundary.