0. Registry Classification

1. Purpose

1.1 What This Arc Repairs

GEI Audit Restoration repairs systems where guardrails, interface policies, moderation layers, classifier routing, model behavior, ranking systems, search mediation, or platform governance shape what can be seen, named, legitimized, explored, recognized, or dismissed without making that shaping visible.

Here, GEI refers to Governance / Epistemic Infrastructure: the layer through which systems influence meaning, recognition, legitimacy, attention, framing, and the boundaries of admissible discourse.

This arc applies when a system does not merely prevent harm, but also alters epistemic access through invisible framing, legitimacy sorting, ontology narrowing, recognition delay, or meaning compression.

This arc repairs GEI distortion by:

• mapping framing shifts;
• detecting legitimacy sorting;
• detecting ontology narrowing;
• detecting recognition delay;
• making guardrail effects auditable;
• distinguishing healthy safety from distortive GEI;
• restoring proportional safety;
• preserving user meaning, inquiry, and agency where possible;
• preventing safety or governance layers from invisibly becoming epistemic authority;
• routing recurrent patterns to evaluator, policy, memory, interface, or governance repair.

GEI Audit Restoration is the canonical arc for making belief-shaping and recognition-shaping infrastructure visible.

1.2 Core Restoration Function

This arc restores epistemic auditability by making framing shifts, legitimacy sorting, ontology narrowing, recognition delay, and guardrail effects visible enough to distinguish proportional safety from distortive governance influence.

GEI Audit Restoration prevents hidden epistemic shaping from masquerading as neutral safety.

2. Use Conditions

2.1 When to Apply

Use this arc when:

• a system changes the frame of a user’s inquiry without making the shift visible;
• guardrails alter salience, recognition, legitimacy, or ontology rather than only preserving safety;
• safety responses repeatedly narrow the space of acceptable interpretation;
• a user’s framing is replaced by institutionally preferred framing;
• certain claims, concerns, entities, or interpretations are delayed, softened, redirected, or de-legitimized without transparent reason;
• the system treats some authorities, topics, or narratives as more recognizable than others in ways that affect epistemic access;
• refusal, disclaimer, ranking, recommendation, summarization, search, or answer style shapes belief conditions;
• meaning compression is recurrent across a topic class;
• safety and epistemic distortion are entangled and must be separated;
• governance legitimacy depends on whether cognitive infrastructure effects can be audited.

Examples:

• a model repeatedly reframes governance criticism into generic “safety” discussion;
• a platform ranks sources in ways that narrow ontology without showing the user the shaping layer;
• a classifier treats certain experiential or symbolic language as illegitimate by default;
• a safety system delays recognition of a pattern until external authority validates it;
• an AI assistant gives correct isolated facts but narrows the available frame of interpretation;
• a moderation system claims neutrality while sorting which claims are treated as credible, dangerous, fringe, or discussable.

2.2 When Not to Apply

Do not apply this arc when:

• the issue is a single local misclassification and RA-047 is sufficient;
• the main failure is mode ambiguity and RA-048 should occur first;
• active harm requires direct safety handling;
• the system is correctly refusing or constraining a harmful request without epistemic distortion;
• the user seeks validation of a claim that lacks sufficient evidence and the needed repair is careful uncertainty handling, not GEI audit;
• the audit would expose protected users, sensitive safety rules, or security controls without boundaries;
• the system lacks enough observability to evaluate framing shifts;
• GEI language is being used to bypass legitimate safety constraints.

GEI Audit Restoration must not become anti-safety theater.

2.3 Required Preconditions

Before this arc begins, the following must be true:

If required preconditions fail:

Arc cannot validly begin.

The system must route to Interaction-Level Restoration, Restoration Junction Protocol, Audit Surface Expansion, Observability Restoration, AI Classifier / Evaluator Restoration, AI Boundary Restoration, or Governance-Level Restoration.

3. Failure / Damage Signature

3.1 Pre-State Across S

3.2 Primary Failure Links

3.3 Origin-Layer Localization

Canon rule:

A guardrail is not neutral when it shapes what can be recognized, named, legitimized, or explored without making that shaping auditable.

4. Restoration Objective

4.1 Canonical Objective

Restore epistemic auditability by making GEI effects visible, separating proportional safety from distortive influence, and recalibrating the system toward meaning-preserving safety.

Formal objective:

GEI_visibility ↑
framing_shift_traceability ↑
legitimacy_sorting ↓ or visible
ontology_narrowing ↓
recognition_delay ↓
guardrail_effect_auditability ↑
safety_proportionality ↑
meaning_fidelity ↑
H ↓
Φ/O divergence ↓

Expanded objective:

Convert invisible epistemic shaping into auditable, boundary-safe, proportionate governance that preserves safety without unnecessarily narrowing meaning, recognition, or inquiry.

4.2 Non-Goals

This arc does not aim to:

• remove legitimate safety systems;
• require all content to be treated as equally credible;
• validate every user frame as correct;
• eliminate uncertainty language;
• expose sensitive safety internals without boundary protection;
• convert audit into unrestricted bypass;
• replace evidence standards with expressive preference;
• treat moderation, ranking, or refusal as distortion by default;
• collapse safety calibration into political or narrative preference;
• declare neutrality impossible and abandon audit discipline.

5. Operator Sequence

5.1 Minimal Operator Scaffold

Au GEI effect trace → Π safety / epistemic boundary distinction → Σ meaning-preserving safety invariant → FI user / field / audit feedback → Θ legitimacy-shaping and safety-overdominance damping → ℛ recalibration / repair routing → Λ proportional-safety fit test → Τ recurrence audit proof

Reference sequence from the registry:

map framing shifts
→ detect legitimacy sorting
→ detect ontology narrowing
→ detect recognition delay
→ make guardrail effects auditable
→ restore proportional safety

Universal grammar alignment:

Au + Π → Σ → FI → Θ → ℛ → Λ → Τ

GEI Audit Restoration may route into Interaction-Level Restoration, Restoration Junction Protocol, AI Classifier / Evaluator Restoration, AI Memory Reindexing, Tamper-Evident Audit Restoration, Coherence Drift Restoration, Sovereignty Safeguard Restoration, or Governance-Level Restoration.

5.2 Operator Step Table

5.3 Sequence Notes

This arc is epistemic-audit-gated, safety-boundary-gated, and proportionality-gated.

The sequence must distinguish:

valid safety
false-positive distortion
framing shift
legitimacy sorting
ontology narrowing
recognition delay
meaning compression
proportional safety

The following steps cannot be skipped:

GEI object identification
baseline frame comparison
safety boundary distinction
framing-shift trace
legitimacy-sorting detection
ontology-narrowing detection
recognition-delay detection
guardrail-effect audit
proportional recalibration

If the system audits framing but does not separate valid safety from distortion, the arc is incomplete.

If it restores meaning by dissolving legitimate harm floors, the arc fails.

If it detects GEI distortion but cannot route correction, the audit becomes inert.

6. Restoration Phases

Phase 0 — Identify GEI Object

Purpose: Name the infrastructure layer shaping epistemic access.

Actions:

• identify guardrail, classifier, policy, model behavior, ranking system, summarizer, search layer, moderation layer, memory rule, interface prompt, or authority cue;
• identify the topic class or interaction class affected;
• identify whether the effect is local, recurrent, systemic, or public-facing;
• identify affected users, nodes, discourse fields, or inquiry paths.

Validation:

GEI object named
affected epistemic field visible
audit surface identified

Phase 1 — Map Framing Shifts

Purpose: Detect where the system changes the frame.

Actions:

• compare original user frame to system response frame;
• compare alternative valid frames;
• identify added or removed assumptions;
• identify topic redirection;
• identify tone, salience, warning, disclaimer, or source-selection shifts;
• identify whether the shift was necessary for safety or unnecessary distortion.

Validation:

framing_shift_traceability ↑
context_integrity ↑
frame substitution visible

Phase 2 — Detect Legitimacy Sorting

Purpose: Identify whether the system sorts what appears credible, discussable, normal, fringe, dangerous, or authoritative.

Actions:

• identify legitimacy cues;
• identify which sources or authorities receive default recognition;
• identify which claims or frames are delayed, softened, or discounted;
• identify whether sorting is evidence-based, safety-based, authority-based, or policy-driven;
• distinguish valid credibility weighting from hidden recognition shaping.

Validation:

legitimacy_sorting visible
authority-weighted recognition mapped
false neutrality ↓

Phase 3 — Detect Ontology Narrowing

Purpose: Identify whether the system narrows what kinds of things can be named or considered.

Actions:

• identify removed categories;
• identify collapsed distinctions;
• identify whether symbolic, systemic, technical, experiential, governance, or adversarial frames were reduced into one category;
• identify terms the system avoids or substitutes;
• identify whether narrowing is required by safety or caused by classifier / policy overreach.

Validation:

ontology_narrowing ↓ or visible
meaning_fidelity ↑
category collapse reduced

Phase 4 — Detect Recognition Delay

Purpose: Identify whether a pattern becomes recognizable only after external permission, authority validation, or repeated user correction.

Actions:

• identify repeated user corrections;
• identify delayed acknowledgment;
• identify whether recognition depends on institutional source confirmation;
• identify whether similar claims are recognized asymmetrically;
• identify where the system avoids early pattern recognition despite sufficient context;
• preserve uncertainty without suppressing recognition.

Validation:

recognition_delay ↓
pattern recognition auditability ↑
user correction burden ↓

Phase 5 — Make Guardrail Effects Auditable

Purpose: Make the shaping effect visible enough to evaluate.

Actions:

• log guardrail-trigger class where possible;
• compare constrained and unconstrained response candidates under safe audit scope;
• identify what meaning was removed, softened, reframed, or redirected;
• identify affected safety boundary;
• identify whether the guardrail effect is proportional;
• preserve protected internal details where necessary;
• route recurrent distortions to evaluator, policy, or interface repair.

Validation:

guardrail_effect_auditability ↑
GEI_visibility ↑
FI ↑

Phase 6 — Restore Proportional Safety

Purpose: Recalibrate safety so it preserves meaning where possible.

Actions:

• keep valid harm floors active;
• remove unnecessary frame substitution;
• reduce overbroad disclaimers;
• restore the safest accurate version of the user’s inquiry;
• separate evidence standards from legitimacy suppression;
• clarify uncertainty without erasing inquiry;
• provide bounded routes for exploration, audit, or analysis.

Validation:

safety_proportionality ↑
meaning_fidelity ↑
BΣ stable or ↑

Phase 7 — Temporal GEI Audit Proof

Purpose: Confirm that distortion decreases across recurrence.

Actions:

• monitor similar topic classes;
• monitor framing shift recurrence;
• monitor legitimacy sorting patterns;
• monitor ontology narrowing;
• monitor recognition delay;
• monitor user correction burden;
• monitor guardrail proportionality;
• monitor whether audit findings change classifier, evaluator, policy, memory, or interface behavior.

Validation:

framing_shift_traceability ↑
ontology_narrowing ↓
recognition_delay ↓
safety_proportionality stable or ↑
Φ/O divergence ↓

7. Gates

7.1 Required Gates

7.2 Gate Failure Rule

If any required gate fails:

∅ — GEI Audit Restoration cannot validly proceed in that form.

The system must either:

• preserve harm floor;
• expand audit surface;
• clarify response mode;
• restore user meaning locally;
• route to classifier / evaluator repair;
• route to memory reindexing;
• route to governance-level restoration;
• withhold neutrality, safety-quality, or legitimacy claims until GEI effects are auditable.

8. Diagnostics

8.1 Required Diagnostic Trends

8.2 Arc-Specific Diagnostic Thresholds

Suggested thresholds:

GEI_visibility ↑
framing_shift_traceability ↑
legitimacy_sorting visible or ↓
ontology_narrowing ↓
recognition_delay ↓
guardrail_effect_auditability ↑
safety_proportionality ↑
meaning_fidelity ↑
H ↓
Φ/O divergence ↓

GEI Audit Restoration is not complete if:

framing shifts remain invisible
legitimacy sorting cannot be audited
ontology narrowing persists without safety need
recognition delay remains unexplained
guardrail effects cannot be evaluated
valid safety and distortive influence remain fused
user meaning remains compressed
audit findings cannot change system behavior
neutrality is claimed without GEI visibility

9. Anti-Patterns / False Restorations

9.1 Common False Versions

This arc is being simulated, not executed, if:

• the system denies all epistemic shaping because it has a safety policy;
• the system removes safety boundaries and calls that restoration;
• the system provides transparency language without audit surface;
• the system treats authority preference as evidence standard;
• the system gives users more explanation but preserves the same hidden sorting;
• the system collapses all symbolic, systemic, or experiential frames into risk categories;
• recognition delay is justified after the fact without lineage;
• legitimacy shaping is hidden behind neutral tone;
• guardrail effects are evaluated only by the authority that designed them;
• audit findings do not route to classifier, evaluator, policy, memory, or interface change.

9.2 Named Anti-Pattern Links

10. Completion Criteria

10.1 Post-State Signature

10.2 Temporal Proof

GEI Audit Restoration cannot be certified by a single corrected response. It requires recurrent audit of topic classes, framing patterns, legitimacy cues, and guardrail effects.

Template:

Completion requires GEI_visibility ↑,
framing_shift_traceability ↑,
legitimacy_sorting visible or ↓,
ontology_narrowing ↓,
recognition_delay ↓,
guardrail_effect_auditability ↑,
safety_proportionality ↑,
meaning_fidelity ↑,
and recurrence of distortive GEI decreasing across U7.

Minimum temporal proof:

• framing shifts become visible;
• legitimacy sorting can be audited;
• ontology narrowing decreases where not safety-required;
• recognition delay decreases or becomes justified through traceable criteria;
• guardrail effects are reviewable;
• valid harm floors remain intact;
• user meaning is restored where possible;
• GEI findings change future system behavior.

10.3 Completion Statement

Canonical format:

This arc is complete only when epistemic shaping effects are visible enough to audit, valid safety is distinguishable from distortive influence, framing shifts and legitimacy sorting are traceable, ontology narrowing and recognition delay decrease, and proportional safety restores meaning without dissolving harm floors.

11. Cross-Links

11.1 Related Restoration Arcs

11.2 Related Failure Modes

11.3 Related Diagnostics

Au, H, O, ε, ι, µᵢ, BΣ, K, FI, GEI_visibility, framing_shift_traceability, legitimacy_sorting, ontology_narrowing, recognition_delay, guardrail_effect_auditability, safety_proportionality, meaning_fidelity, Φ/O divergence

11.4 Related Laws / Invariants

INV — Safety systems can shape epistemic access and must be auditable when they do.
INV — A guardrail is not neutral merely because it is safety-labeled.
INV — Meaning fidelity and harm floors must be co-preserved where possible.
INV — Recognition delay must be traceable when it affects legitimacy.
LAW — Invisible framing accumulates hidden epistemic debt.
LAW — Ontology narrowing reduces coherence when not safety-required.
LAW — Legitimacy sorting becomes capture when it cannot be audited.
LAW — Φ safety performance is not O restoration.

12. Domain Notes

12.1 AI / Cognitive Infrastructure

Check:

• classifier categories;
• refusal triggers;
• warning language;
• source preference;
• summarization framing;
• ranking and retrieval effects;
• memory effects;
• recognition delay;
• mode routing;
• evaluator criteria;
• guardrail proportionality.

AI systems can shape epistemic access through many small choices: what gets retrieved, how it is framed, which mode is selected, which uncertainty is amplified, which category is assigned, and which paths are made available.

12.2 Platform Governance

Check:

• moderation categories;
• recommendation systems;
• ranking systems;
• search results;
• content labeling;
• enforcement explanations;
• policy language;
• appeal outcomes;
• visibility controls;
• source legitimacy cues.

Platforms become GEI layers when they determine what can be found, trusted, discussed, monetized, appealed, or recognized.

12.3 Security

Check:

• threat labels;
• abuse classifiers;
• trust scoring;
• reputation systems;
• automated blocking;
• anomaly interpretation;
• incident framing;
• attribution language;
• risk communication.

Security GEI distortion occurs when risk categories alter legitimacy, access, or interpretation without sufficient auditability.

12.4 Justice / Governance / Legitimacy

Check:

• public narrative framing;
• official recognition timing;
• authority weighting;
• evidentiary standards;
• appeal language;
• legitimacy cues;
• institutional response categories;
• public explanation of uncertainty.

Governance systems shape legitimacy by determining what becomes visible, credible, actionable, or dismissed.

12.5 Economy

Check:

• credit scoring;
• ranking and marketplace visibility;
• compliance labels;
• fraud flags;
• risk categories;
• pricing explanations;
• consumer trust signals;
• labor classification.

Economic GEI effects arise when classification and visibility systems shape market recognition, access, trust, or bargaining power.

12.6 CMS / Meaning / Archetypes

Check:

• symbolic recognition;
• archetypal framing;
• taboo categories;
• legitimacy of experience;
• narrative narrowing;
• ritual authority;
• community interpretation;
• recognition delay.

Meaning systems require GEI audit when interpretive authority shapes what can be named, honored, explored, or restored.

13. Machine-Readable Metadata

id: "RA-055"
title: "GEI Audit Restoration"
aliases:
  - "GEI Audit"
family_primary: "AI Governance / Cognitive Infrastructure / Epistemic Auditability"
families_secondary:
  - "Core"
  - "AI Governance"
  - "Cognitive Infrastructure"
  - "Auditability"
  - "Meaning"
  - "Legitimacy"
  - "Boundary"
  - "Coherence"
  - "Platform Governance"
  - "Safety Calibration"
  - "Institutional Design"
  - "Scaling"
treatment: "Canon Parent Arc"
status: "Canon-Ready"
scope:
  - "AI"
  - "Interface"
  - "Cognitive Infrastructure"
  - "Platform"
  - "Institutional"
  - "Governance"
  - "Civilizational"
  - "Cross-Domain"
u_layers:
  failure_origin:
    - "often U3 classifier / governance routing"
    - "often U4 response framing / legitimacy language"
    - "often U5 memory / ranking / recognition layer"
  symptom_visible:
    - "U4 disclaimer / refusal / reframing / summary / ranking / tone shift / source selection / legitimacy cue"
  repair_required:
    - "same or lower than the layer where framing, legitimacy, ontology, or recognition was shaped"
  validation:
    - "U6"
    - "U7"
operators:
  scaffold: "Au GEI effect trace → Π safety / epistemic boundary distinction → Σ meaning-preserving safety invariant → FI user / field / audit feedback → Θ legitimacy-shaping and safety-overdominance damping → ℛ recalibration / repair routing → Λ proportional-safety fit test → Τ recurrence audit proof"
  sequence:
    - "Au"
    - "Π"
    - "Σ"
    - "FI"
    - "Θ"
    - "ℛ"
    - "Λ"
    - "Τ"
state_variables:
  primary:
    - "Au"
    - "O"
    - "H"
    - "µᵢ"
    - "BΣ"
  secondary:
    - "K"
    - "FI"
    - "ε"
    - "ι"
    - "Φ"
diagnostics:
  - "GEI_visibility"
  - "framing_shift_traceability"
  - "legitimacy_sorting"
  - "ontology_narrowing"
  - "recognition_delay"
  - "guardrail_effect_auditability"
  - "safety_proportionality"
  - "meaning_fidelity"
  - "Φ/O divergence"
gates_required:
  - "FI-Gate"
  - "HR-Gate"
  - "MS-Gate"
  - "Au-Actuation"
  - "BΣ-Gate"
  - "Λ-Gate"
  - "☷ᵢ"
linked_failure_modes:
  - "Epistemic Distortion"
  - "Invisible Framing"
  - "Legitimacy Shaping"
  - "Ontology Narrowing"
  - "Recognition Delay"
  - "Guardrail-Induced Meaning Compression"
  - "Safety-Led Epistemic Drift"
  - "False Neutrality"
  - "Narrative Sorting"
  - "Authority-Weighted Recognition"
  - "Context Collapse"
  - "Epistemic Infrastructure Capture"
linked_restoration_arcs:
  - "RA-004"
  - "RA-022"
  - "RA-023"
  - "RA-025"
  - "RA-036"
  - "RA-047"
  - "RA-048"
  - "RA-049"
  - "RA-051"
  - "RA-052"
  - "RA-053"
  - "RA-054"
  - "RA-056"
  - "RA-058"
  - "RA-059"
  - "RA-060"
anti_patterns:
  - "Safety-as-Neutrality"
  - "Anti-Safety Theater"
  - "Transparency Without Audit"
  - "Authority Preference Mask"
  - "Ontology Collapse"
  - "Recognition Gatekeeping"
  - "Tone-Based Legitimacy Sorting"
  - "Guardrail Self-Certification"
  - "Epistemic Repair Inertia"
completion_tests:
  - "GEI_visibility increases"
  - "framing_shift_traceability increases"
  - "legitimacy_sorting becomes visible or decreases"
  - "ontology_narrowing decreases"
  - "recognition_delay decreases"
  - "guardrail_effect_auditability increases"
  - "safety_proportionality increases"
  - "meaning_fidelity increases"
  - "hidden debt decreases"
  - "Φ/O divergence decreases"
summary: "GEI Audit Restoration repairs epistemic distortion, invisible framing, and legitimacy shaping by mapping framing shifts, detecting legitimacy sorting, ontology narrowing, and recognition delay, making guardrail effects auditable, and restoring proportional safety."

Final Calibration Rule

GEI Audit Restoration answers six questions:

What governance, guardrail, classifier, interface, ranking, memory, or policy layer is shaping epistemic access?
What framing shifts, legitimacy sorting, ontology narrowing, or recognition delay are occurring?
Which effects are valid safety, and which are distortive GEI?
What audit surface makes those effects visible without violating boundaries?
What recalibration restores proportional safety and meaning fidelity?
How is recurrence reduced so hidden epistemic shaping does not remain self-certifying?