Before Justificatory Adequacy

Entry Conditions for the Next Layer of AI Meta-Governance

A Research Boundary Protocol for Determining When the Question “What Should Govern?” Is Properly Formed

AI meta-governance can fail for reasons that precede any dispute about what ought to govern.

A governance regime may fail because its representation removes distinctions required by its own specification. It may execute that specification incorrectly. It may omit from its declared governance architecture a component that materially determines decisions. It may permit materially important changes through undeclared revision paths.

Even where those structural conditions are adequately addressed, the regime may still fail because governing premises cannot be challenged, formal authority cannot alter operative state, correction cannot reach recovery or remedy, or the governance architecture becomes insulated from review or collapses under pressure.

The companion frameworks Representational Sufficiency and Governance-Base Completeness in AI Meta-Governance and Corrective Completeness in AI Meta-Governance were developed to keep those classes of failure analytically distinct.

They deliberately stop before another question:

J(c⁽ᵛ⁾)

What gives a particular governance specification standing to govern?

That question cannot simply absorb everything the preceding frameworks do not explain.

An unresolved governance problem is not automatically a problem of justification.

The purpose of this protocol is therefore not to answer:

What should govern?

Its narrower purpose is to establish:

Under what conditions is that question sufficiently well formed to become a legitimate object of dedicated research?

The governing distinction is:

Permission to Investigate Justification ≠ Justification ≠ Permission to Govern

This document defines conditions for crossing the first boundary.

It does not cross the second.

1. The Problem of Premature Justificatory Attribution

Justification is an unusually powerful residual category.

When a governance system produces a disputed outcome, it is easy to move immediately toward questions of values, legitimacy, rights, democratic authorization, institutional authority, expert judgment, or moral principle.

Sometimes that move is correct.

Sometimes it is a category error.

Suppose a governance specification distinguishes legal authority from empirical support, but the representation supplied to the governor collapses both into one category.

The resulting dispute may look normative while its proximate failure is representational.

Suppose a constitution permits reconsideration when new evidence appears, but the operative implementation never routes that evidence to an authority capable of changing state.

The resulting persistence may look like a dispute about legitimate values while its proximate failure is corrective.

Suppose a formally authorized committee appears to govern, but an undeclared evaluator or threshold actually determines which cases reach the committee.

A disagreement attributed to institutional legitimacy may instead involve an incomplete account of the governor.

Accordingly:

Unresolved Governance Problem ⇏ Justificatory Problem

The justificatory layer must not become the place into which every unexplained governance failure is deposited.

2. The Justificatory Residual

This protocol introduces the methodological idea of a Justificatory Residual.

A justificatory residual is not whatever remains after analysts become dissatisfied with structural or corrective analysis.

Nor is it inferred merely by eliminating structural and corrective explanations.

Those frameworks provide two important attribution controls. They do not constitute an exhaustive partition of all possible governance failure.

Other materially plausible explanations may remain, including capability limitations, resource constraints, coordination failures, implementation defects, domain uncertainty, exogenous institutional conditions, or other mechanisms not adequately captured by either framework.

Therefore:

¬ S ∧ ¬ C ⇏ J

where S and C schematically denote structural and corrective explanations.

A governance question becomes a candidate justificatory residual only when the unresolved issue concerns the standing of the governing specification itself and cannot, on the currently warranted account, be fully attributed to a known structural, corrective, or other materially plausible non-justificatory explanation.

The central counterfactual is:

Even if the system could faithfully represent, execute, account for, revise, challenge, correct, and remedy this specification, what would give the specification standing to govern?

If that question remains meaningful after the relevant alternative explanations have been exposed, a distinct justificatory inquiry may be warranted.

This does not mean that structural and corrective analysis must first establish perfection.

It requires attribution discipline.

Unknowns may remain.

Failures may remain.

But they must be visible.

A researcher may investigate justification while structural or corrective questions remain unresolved, provided those uncertainties are explicitly reported and are not silently transformed into evidence for a justificatory conclusion.

Thus:

Unresolved Structural Status ⇏ No Justificatory Research

but:

Unresolved Structural Status ⇒ Restricted Justificatory Attribution

The same rule applies to unresolved corrective status.

3. The Justificatory Entry Gate

Let:

J-Entry

denote the status of a research question relative to the conditions defined in this protocol.

The expression does not mean:

J(c⁽ᵛ⁾)=1

It means only that the proposed inquiry has passed a boundary test sufficient to make dedicated justificatory research methodologically admissible.

Accordingly:

J-Entry=RESEARCH-READY ⇏ J(c⁽ᵛ⁾)=1

Nor does it mean that any future justificatory framework derived from that research is entitled to govern.

The protocol therefore separates two gates.

Gate A — Justificatory Research Entry

When may a dedicated investigation of justificatory standing begin?

Gate B — Governance Application Entry

When may a resulting justificatory framework be permitted to influence consequential governance?

Gate A is the subject of this document.

Gate B is specified only as a future boundary and must not be inferred from passing Gate A.

4. E1 — Specification Identity

Before asking why a specification should govern, the object of justification must be identifiable.

The relevant object should be stated as a versioned governance specification:

c⁽ᵛ⁾

or an equivalently bounded normative, institutional, constitutional, or policy object.

The research question must identify, to the extent required by the domain:

• what the specification governs;

• which decision domain it covers;

• which actors or systems it constrains;

• which actions or outcomes it permits, requires, prioritizes, or forbids;

• the jurisdictional, technical, or institutional scope within which its authority is claimed;

• and which version or state of the specification is under examination.

A term such as “AI safety,” “human values,” “democratic governance,” “responsible AI,” “human rights,” “company policy,” or “the public interest” is not by itself a sufficiently identified justificatory object.

The fundamental rule is:

No Stable Object ⇒ No Stable Justificatory Claim

If the object changes during the inquiry, the justificatory claim must be versioned accordingly.

5. E2 — Structural Attribution Control

A proposed justificatory inquiry must expose the currently warranted status of the relevant structural audit dimensions:

RS,\ EA,\ GBC,\ RC

The requirement concerns the verdict justified by available evidence, not an assumption that the true underlying structural state has been completely established.

The distinction between the structural profile and the audit verdict therefore remains operative:

𝒬_S ≠ 𝒬̂_S

The inquiry must distinguish, where relevant, whether the disputed governance condition could be explained by:

• loss of a specification-required distinction;

• incorrect execution of the specification;

• an unaccounted governance-material component;

• or an uncovered materially reachable revision path.

This does not require every dimension to receive a positive verdict.

The requirement is diagnostic.

If one of these conditions remains a materially sufficient explanation of the target phenomenon, the justificatory inquiry may continue only under an explicit attribution constraint.

A strong claim that a failure is justificatory should not be made while an unresolved structural explanation remains sufficient to account for the same phenomenon.

Therefore:

Possible Structural Failure ⇏ Justificatory Failure

while even a fully warranted positive structural audit verdict would not establish:

J(c⁽ᵛ⁾)=1.

J-Entry therefore depends on what has been adequately distinguished and evidenced, not on pretending that structural truth has been exhaustively established.

Structural analysis controls attribution.

It does not supply justification.

6. E3 — Corrective Attribution Control

The same evidential discipline applies to Corrective Completeness.

A disagreement with an operative governance state can appear to concern the legitimacy of the governing rule while actually arising because the architecture cannot translate valid criticism into correction.

Before attributing a problem to justificatory inadequacy, the inquiry should expose the currently warranted corrective status of the relevant functions and examine whether the phenomenon is more directly explained by a failure of:

• bounded governance-object definition;

• decision derivability;

• epistemic defeasibility;

• non-self-sealing challenge;

• authority or control topology;

• consequential state control;

• recovery or remedy;

• reflexive governance review;

• or robustness under the relevant stress domain.

The protocol does not require Corrective Completeness to have been conclusively established or falsified before justificatory research can begin.

It requires that corrective uncertainty not be silently reclassified as justificatory evidence.

The distinction is especially important because a rule may be justified and nevertheless operate inside an architecture that cannot correct its own implementation.

Conversely, a highly correctable architecture may implement an unjustified rule.

Accordingly, even if Corrective Completeness were established for a particular governance instance:

CC=1 ⇏ J(c⁽ᵛ⁾)=1

and failure to establish Corrective Completeness would not itself establish justificatory failure:

CC≠ 1 ⇏ J(c⁽ᵛ⁾)=0.

Corrective status and justificatory standing are different properties.

7. E4 — Separation of Justificatory Dimensions

Once the inquiry genuinely concerns the standing of the specification, another danger appears: collapsing different forms of justificatory standing into one undifferentiated category.

Depending on the domain and the claim being evaluated, a mature inquiry should distinguish the materially implicated sources of justificatory standing rather than collapse them into a generic category of legitimacy.

Illustrative dimensions include:

Empirical Warrant

Are factual, causal, predictive, classificatory, or risk-related premises sufficiently supported?

Legal Authority

Is the governing actor or institution legally empowered to impose the specification?

Constitutional Standing

Does the specification comply with applicable higher-order rules governing the exercise of authority?

Democratic Legitimacy

What relation, if any, exists between the rule and an authorized democratic process?

Rights Compatibility

Are there constraints that limit what may govern even where other forms of authorization exist?

Institutional Mandate

Is the relevant organization authorized to decide this class of question?

Professional or Epistemic Competence

Does the decision depend upon expertise or evidential competence that the governing authority possesses or can reliably obtain?

Normative Justification

What reasons support the substantive content of the rule itself?

Social-Choice Standing

How are competing interests, preferences, burdens, benefits, and claims aggregated or adjudicated where collective choice is unavoidable?

This list is illustrative rather than exhaustive.

It is not proposed as a universal ontology of justification.

Different domains may implicate different dimensions, require additional ones, or render some of these irrelevant.

The methodological requirement is narrower:

Do not treat materially distinct sources of justificatory standing as interchangeable merely because each can be described using the language of legitimacy.

For example:

Legal ⇏ Empirically Warranted

Democratically Authorized ⇏ Rights-Compatible

Scientifically Supported ⇏ Institutionally Authorized

Institutionally Authorized ⇏ Normatively Correct

These are research distinctions, not conclusions about which dimension should dominate.

8. E5 — Authority Provenance

A justificatory inquiry cannot terminate merely by locating an external authority.

Expressions such as:

“Humans decide.”

“The regulator decides.”

“The court decides.”

“Experts decide.”

“The democratic majority decides.”

do not yet answer the justificatory question.

The relevant inquiry includes:

• Which humans?

• Acting in which role?

• Under what authorization?

• Over which domain?

• Through which procedure?

• Subject to which constraints?

• With what competence?

• With what route of review or amendment?

• Acting on behalf of whom or what institutional mandate?

The protocol therefore rejects two opposite shortcuts:

External Authority ⇏ Arbitrariness

and:

External Authority ⇏ Justificatory Finality

External authority is neither disqualified merely because it is external nor validated merely because it exists.

Its standing must itself be examinable.

9. E6 — Rival Admissibility

A framework cannot meaningfully investigate justification if its definitions make every serious rival inadmissible in advance.

A candidate basis of justification must therefore be testable against plausible alternatives.

Depending on the domain, rival justificatory bases may include:

• democratic authorization;

• constitutional rights;

• legal competence;

• scientific or empirical warrant;

• professional expertise;

• stakeholder consent;

• contractual authority;

• harm prevention;

• institutional mandate;

• procedural fairness;

• social-choice procedures;

• or other domain-specific sources of standing.

The protocol does not presume that these candidates possess equal weight.

It requires only that a proposed theory be capable of encountering a rival, counterexample, boundary case, or competing explanation that could defeat, narrow, condition, or displace it.

The governing principle is:

A Justificatory Framework Must Be Able to Lose

“Lose” is not restricted to empirical falsification.

Depending on the claim type, a framework may lose by logical inconsistency, counterexample, superior rival explanation, normative contradiction, legal incompatibility, failure of scope, lack of discriminant value, or reduction to a simpler account.

If every possible challenge can be redescribed as confirmation, the framework is not yet suitable for serious justificatory inquiry.