0. Registry Classification

1. Purpose

1.1 What This Arc Repairs

Feedback Integrity Restoration repairs systems whose correction, learning, control, evaluation, or response loops no longer reference actual coherence.

It applies when feedback has been captured, suppressed, gamed, narrowed, Goodharted, or replaced by proxy success.

This arc repairs feedback failure by:

• identifying where Φ has displaced O;
• restoring feedback channels to coherence-bearing signals;
• expanding auditability around evaluator, metric, classifier, and response pathways;
• detecting inversion caused by proxy dominance;
• widening signal diversity;
• reducing captured or self-confirming loops;
• restoring field validation;
• preventing future correction loops from optimizing the wrong target.

Feedback Integrity Restoration is the canonical arc for repairing systems that cannot learn because their feedback no longer tells the truth.

1.2 Core Restoration Function

This arc restores feedback integrity by separating proxy fitness from coherence, increasing auditability around feedback channels, detecting inversion, widening signal diversity, and validating that the system responds to field reality rather than metric, benchmark, reputation, compliance, or authority pressure.

Feedback Integrity Restoration prevents control from becoming self-confirming error.

2. Use Conditions

2.1 When to Apply

Use this arc when:

• Φ is improving while O is stable, unknown, or declining;
• feedback channels are captured, gamed, suppressed, or narrowed;
• metrics, dashboards, benchmarks, compliance artifacts, or reputation signals dominate field reality;
• the system optimizes for what is measured rather than what matters;
• affected-node feedback is ignored, filtered, delayed, or reframed;
• classifiers, evaluators, audits, or review loops reinforce the failure;
• correction loops repeatedly miss the true damage pattern;
• the same failure recurs despite “improving” scores;
• learning produces better performance but worse coherence;
• authority-controlled feedback replaces independent field signals.

Examples:

• an AI model improves benchmark scores while user meaning compression worsens;
• a security program improves compliance while real incidents recur;
• an institution reports trust metrics while affected parties lose faith;
• an economy shows growth while hidden burden and extraction rise;
• a governance system treats procedural completion as evidence of justice;
• a classifier produces high precision on a proxy while field harm increases.

2.2 When Not to Apply

Do not apply this arc when:

• active harm is still cascading and emergency stabilization is required first;
• auditability is too low to inspect feedback channels;
• the system lacks any external or independent signal source;
• feedback restoration would expose affected nodes to harm without boundary protection;
• the needed repair is already localized and must occur immediately at origin layer;
• feedback language is being used to delay material repair;
• the system refuses to subordinate Φ to O;
• the feedback loop is structurally invalid and must be replaced, not recalibrated.

Feedback Integrity Restoration must not become metric reform 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 return to emergency stabilization, audit surface expansion, observability restoration, or origin-layer repair before attempting feedback integrity restoration.

3. Failure / Damage Signature

3.1 Pre-State Across S

3.2 Primary Failure Links

3.3 Origin-Layer Localization

Canon rule:

Feedback is valid only when it remains coupled to coherence-bearing field signals rather than proxy success alone.

4. Restoration Objective

4.1 Canonical Objective

Restore feedback integrity by separating proxy success from coherence, repairing captured evaluators or signal channels, and validating that the system learns from real field conditions.

Formal objective:

FI ↑
Φ/O divergence ↓
Au_feedback ↑
ι ↓
Γ_signal diversity ↑
H_proxy-hidden ↓
recurrence ↓

Expanded objective:

Rebuild the correction loop so that the system can distinguish improvement from appearance, repair from optimization, and coherence from local fitness.

4.2 Non-Goals

This arc does not aim to:

• improve metrics alone;
• create a better dashboard without repairing feedback;
• replace one proxy with another untested proxy;
• suppress inconvenient feedback;
• overfit to affected-node testimony without boundary protection;
• optimize benchmark performance;
• protect institutional confidence;
• certify restoration through compliance artifacts;
• treat user, citizen, patient, worker, or field feedback as extractive data;
• make Φ more sophisticated while O remains unmeasured.

5. Operator Sequence

5.1 Minimal Operator Scaffold

Au feedback trace → FI reference restoration → Ξ proxy inversion detection → Μ feedback map → Γ signal widening → Σ metric subordination → Θ gain reduction → Τ field validation

Universal grammar alignment:

Σ + Θ → Au + FI → Ξ → Γ widen → ℛ(feedback layer) → Τ → Temporal Proof

Feedback Integrity Restoration usually routes into origin-layer repair, evaluator restoration, observability restoration, or pseudo-coherence correction.

5.2 Operator Step Table

5.3 Sequence Notes

This arc is reference-gated.

The system must identify what the feedback loop is actually optimizing, then restore the loop’s reference back toward coherence.

The following steps cannot be skipped:

feedback pathway trace
Φ/O separation
proxy inversion detection
signal widening
metric subordination
field validation

If a new metric is introduced without field validation, the arc has not completed.

If the system improves its evaluator score while field harm remains unchanged, the arc has failed.

6. Restoration Phases

Phase 0 — Identify Feedback Failure

Purpose: Determine which feedback loop is corrupted, suppressed, captured, or proxy-dominated.

Actions:

• identify the metric, evaluator, classifier, review loop, dashboard, or signal channel;
• identify what the loop claims to measure;
• identify what the loop actually selects for;
• identify whose feedback is missing, distorted, or excluded;
• identify where Φ has gained authority over O.

Validation:

feedback loop named
claimed reference separated from actual selection pressure
Φ/O divergence suspected or visible

Phase 1 — Trace Feedback Pathway

Purpose: Make the feedback loop inspectable.

Actions:

• map data sources;
• map evaluator rules;
• map classifier thresholds;
• map review timing;
• map incentives;
• map filters and exclusions;
• map who can contest or correct the feedback.

Validation:

Au_feedback ↑
evaluator path traceable
feedback exclusions visible

Phase 2 — Separate Proxy From Coherence

Purpose: Prevent proxy improvement from being mistaken for restoration.

Actions:

• identify Φ indicators;
• identify coherence-bearing signals;
• compare dashboard or metric performance to field effects;
• identify where local success harms global coherence;
• block Φ as a completion signal by itself.

Validation:

Φ/O distinction explicit
proxy authority reduced
coherence-bearing signals named

Phase 3 — Detect Inversion

Purpose: Identify how feedback has begun selecting against coherence.

Actions:

• look for Goodhart loops;
• identify reward hacking;
• identify evaluator capture;
• identify hidden incentives;
• identify self-confirming review loops;
• identify field signals filtered as noise.

Validation:

ι source identified
captured loop named
hidden selection pressure visible

Phase 4 — Widen Signal Field

Purpose: Restore feedback diversity and reduce monoculture.

Actions:

• add independent field signals;
• restore affected-node verification channels;
• include delayed outcomes;
• include negative cases;
• include boundary and hidden-debt signals;
• reduce reliance on single dashboards or authority channels.

Validation:

Γ_signal diversity ↑
affected-node verification possible
feedback no longer mono-proxy

Phase 5 — Recalibrate Feedback Reference

Purpose: Re-anchor the loop to coherence rather than proxy fitness.

Actions:

• redefine success conditions;
• subordinate proxy metrics to coherence signals;
• remove incentives that reward appearance;
• adjust classifier, evaluator, review, or reporting logic;
• add contradiction and appeal pathways;
• preserve auditability around changes.

Validation:

FI ↑
Φ/O divergence ↓
metric no longer self-certifies

Phase 6 — Reduce Optimization Gain

Purpose: Prevent the repaired metric from being immediately Goodharted again.

Actions:

• reduce reward intensity;
• reduce metric pressure;
• slow optimization cadence;
• avoid hard target fixation;
• preserve human or field review where needed;
• maintain multiple feedback channels.

Validation:

Gain_feedback ↓
𝓓 ↑
reward hacking pressure ↓

Phase 7 — Temporal and Field Validation

Purpose: Prove that feedback restoration changes real outcomes over time.

Actions:

• monitor recurrence;
• monitor hidden debt;
• compare field response to metric response;
• review delayed effects;
• verify affected-node signals where applicable;
• update feedback loop if drift returns.

Validation:

O ↑ or stable
H ↓
Φ/O divergence ↓
τ_m ↓
recurrence ↓

7. Gates

7.1 Required Gates

7.2 Gate Failure Rule

If any required gate fails:

∅ — Feedback Integrity Restoration cannot validly proceed in that form.

The system must either:

• expand auditability;
• protect affected boundaries;
• reduce metric pressure;
• widen signal channels;
• add independent validation;
• reroute to observability restoration;
• block metric-based closure.

8. Diagnostics

8.1 Required Diagnostic Trends

8.2 Arc-Specific Diagnostic Thresholds

Suggested thresholds:

FI ↑
Φ/O divergence ↓
Au_feedback ↑
Γ_signal diversity ↑
H_proxy-hidden ↓
recurrence ↓ across U7
field effects match feedback claims

Feedback Integrity Restoration is not complete if:

new metric replaces old metric without field validation
affected-node feedback remains filtered or unsafe
Φ improves while O remains unknown or declining
evaluator path remains opaque
negative feedback is treated as noise
reward pressure remains high enough to re-Goodhart the system

9. Anti-Patterns / False Restorations

9.1 Common False Versions

This arc is being simulated, not executed, if:

• old metric is replaced with a new proxy without field validation;
• benchmark performance improves while field harm continues;
• affected-node feedback is collected but not allowed to change outcomes;
• dashboards become more complex but not more truthful;
• negative feedback is reframed as misunderstanding;
• compliance artifacts are treated as repair proof;
• independent signals are excluded as “outliers”;
• review loops are controlled by the system being evaluated;
• feedback is widened only cosmetically;
• the system optimizes the feedback reform itself.

9.2 Named Anti-Pattern Links

10. Completion Criteria

10.1 Post-State Signature

10.2 Temporal Proof

Feedback Integrity Restoration cannot be declared complete until the repaired feedback loop survives delay, contradiction, and recurrence.

Template:

Completion requires FI(t+n) ≥ FI(t),
Φ/O divergence decreasing,
feedback pathway remaining auditable,
and recurrence decreasing across U7.

Minimum temporal proof:

• the system accepts correction from non-proxy signals;
• field effects align with feedback claims;
• affected-node signals remain admissible where applicable;
• hidden debt does not continue rising behind improved metrics;
• evaluator drift does not recur;
• proxy pressure remains bounded.

10.3 Completion Statement

Canonical format:

This arc is complete only when feedback loops reference coherence-bearing field signals, proxy success is subordinated to O, evaluator pathways remain auditable, and recurrence decreases without hidden debt being masked by improved metrics.

11. Cross-Links

11.1 Related Restoration Arcs

11.2 Related Failure Modes

11.3 Related Diagnostics

FI, Φ/O divergence, Au, H, ι, Γ_signal diversity, R, 𝓓(t), τ_m, recurrence

11.4 Related Laws / Invariants

INV — Coherence cannot be inferred from appearance alone.
INV — Feedback must remain coupled to the variable it claims to regulate.
INV — Fitness proxy is not coherence.
LAW — Φ improvement is not O restoration.
LAW — Goodhart pressure rises when proxy authority exceeds auditability.
LAW — Feedback capture converts learning into inversion.
LAW — Auditability must scale with evaluator complexity.
LAW — Field validation must outrank dashboard authority.

12. Domain Notes

12.1 AI / Cognitive Infrastructure

Check:

• classifier trigger path;
• evaluator provenance;
• benchmark-to-field divergence;
• reward function pressure;
• refusal / safety classifier drift;
• memory retrieval feedback;
• user correction pathway;
• appeal path;
• whether safety or performance signals are suppressing user meaning.

AI feedback restoration requires separating benchmark success, policy compliance, classifier confidence, and actual user/coherence outcome.

12.2 Justice / Governance / Legitimacy

Check:

• whether public trust metrics obscure affected-node experience;
• whether reporting systems filter negative feedback;
• whether complaint resolution metrics hide repair failure;
• whether procedural completion is treated as justice;
• whether rank immunity filters feedback;
• whether harmed-node feedback is admissible without becoming extraction.

JGL feedback restoration must preserve affected-node boundaries while allowing truth-bearing feedback to change outcomes.

12.3 Biology / Medicine

Conceptual systems mapping only.

Feedback Integrity Restoration in biological systems means distinguishing proxy improvement from resilience, recurrence reduction, boundary stability, timing recovery, and perturbation tolerance.

Not diagnosis.

Not treatment.

Not medical advice.

12.4 Economy

Check:

• whether profit, GDP, productivity, or valuation is substituting for coherence;
• whether hidden costs and externalities are excluded from feedback;
• whether forced-choice conditions are counted as “demand”;
• whether market signals hide boundary, ecological, or labor debt;
• whether growth metrics improve while clearance and slack degrade.

Economic feedback restoration requires externality-inclusive and hidden-debt-aware signals.

12.5 CMS / Meaning / Archetypes

Check:

• whether symbolic resonance is mistaken for truth;
• whether group agreement suppresses contradiction;
• whether spiritual or archetypal authority filters feedback;
• whether taboo prevents corrective signal;
• whether perceived harmony hides hidden debt;
• whether “alignment” is being measured by compliance.

Meaning systems require feedback that preserves depth while still admitting contradiction, boundary signals, and field consequences.

13. Machine-Readable Metadata

id: "RA-008"
title: "Feedback Integrity Restoration"
aliases:
  - "FI Restoration"
family_primary: "Cybernetics"
families_secondary:
  - "Core"
  - "Auditability"
  - "Coherence"
  - "Scaling"
  - "Security"
  - "AI Governance"
  - "Justice / Governance / Legitimacy"
  - "Economy"
treatment: "Canon Parent Arc"
status: "Canon-Ready"
scope:
  - "Local"
  - "Relational"
  - "Institutional"
  - "AI"
  - "Economic"
  - "Civilizational"
  - "Cross-Domain"
u_layers:
  failure_origin:
    - "often U3 classifier / evaluator / control loop"
    - "often U4 metric / narrative"
    - "often U5 review timing"
  symptom_visible:
    - "U6 field harm"
    - "Φ dashboard / benchmark / compliance improvement"
  repair_required:
    - "same or lower than captured feedback or evaluator layer"
  validation:
    - "U5"
    - "U6"
    - "U7"
operators:
  scaffold: "Au feedback trace → FI reference restoration → Ξ proxy inversion detection → Μ feedback map → Γ signal widening → Σ metric subordination → Θ gain reduction → Τ field validation"
  sequence:
    - "Au"
    - "FI"
    - "Ξ"
    - "Μ"
    - "Γ"
    - "Σ"
    - "Θ"
    - "Τ"
state_variables:
  primary:
    - "FI"
    - "Φ/O divergence"
    - "Au"
    - "H"
  secondary:
    - "O"
    - "ι"
    - "R"
    - "BΣ"
diagnostics:
  - "Γ_signal diversity"
  - "𝓓(t)"
  - "τ_m"
  - "recurrence"
gates_required:
  - "FI-Gate"
  - "HR-Gate"
  - "MS-Gate"
  - "Au-Actuation"
  - "BΣ-Gate"
  - "Λ-Gate"
  - "☷ᵢ"
linked_failure_modes:
  - "Goodhart Collapse"
  - "Metric Substitution"
  - "Reward Hacking"
  - "Evaluator Capture"
  - "Benchmark Substitution"
  - "Feedback Suppression"
  - "Proxy Dominance"
  - "Security Theater"
  - "Compliance Theater"
  - "Pseudo-Coherence"
  - "Classifier Drift"
  - "Narrative Capture"
linked_restoration_arcs:
  - "RA-002"
  - "RA-004"
  - "RA-009"
  - "RA-015"
  - "RA-017"
  - "RA-025"
  - "RA-058"
anti_patterns:
  - "Metric Reform Theater"
  - "Benchmark Substitution"
  - "Evaluator Capture"
  - "Compliance Theater"
  - "Dashboard Authority"
  - "Feedback Extraction"
  - "Proxy Rotation"
completion_tests:
  - "FI increases"
  - "Φ/O divergence decreases"
  - "Au_feedback increases"
  - "Γ_signal diversity increases"
  - "H_proxy-hidden decreases"
  - "field effects match feedback claims"
  - "recurrence decreases across U7"
summary: "Feedback Integrity Restoration repairs captured or proxy-dominated feedback loops so correction, learning, and evaluation reference coherence-bearing field signals rather than metric, benchmark, compliance, reputation, or authority proxies."

Final Calibration Rule

Feedback Integrity Restoration answers six questions:

What hidden debt is being masked by proxy feedback?
What boundary must be protected while feedback channels are restored?
What auditability proves the feedback loop is traceable?
What metric, evaluator, classifier, or proxy must remain subordinated to coherence?
What repair trajectory becomes viable once feedback references O again?
How is feedback integrity proven over time through field validation and recurrence reduction?