# Evidence Attribution Human-Review Readiness and Translation-Quality Protocol

## Status

Current reviewed-with-limitations protocol for IARPG-OPS-2 2.0.30-wip. The protocol prepares twelve preserved evidence-attribution wrappers and seventy-two dispositions for later qualified review. It contains **zero real reviewer submissions**, **zero qualified translation approvals**, and **zero completed adjudications**.

## Purpose

Define a publication-safe, non-scoring sequence for human, legal, technical, translation, locally informed, affected-community, privacy, rights, remedy, and editorial review while preserving source bytes, prior dispositions, evidence axes, and correction history separately.

## Findings

1. Every source receives eight source-level review categories: legal status, current facts, translation, technical attribution, affected-community perspectives, privacy, rights, and remedy.
2. Every one of the seventy-two dispositions receives three separate questions covering evidence and wording, source and translation requirements, and rights, remedy, correction, and publication effect.
3. Every disposition has an individually addressable adjudication slot with null reviewer and adjudicator fields.
4. Claim status, analytic confidence, source reliability, attribution rung, source origin chain, translation state, and correction history remain separate fields.
5. Original bytes, filenames, byte counts, hashes, source-declared cutoffs, reviewed wrappers, and 2.0.29 dispositions remain individually addressable.
6. Original-language, locally authored, and affected-community evidence requirements are source-specific and unresolved.
7. Public wording distinguishes actor claims, platform findings, intelligence assessments, court allegations, formal judgments, reach, and causal effect; it also tests leaderless-role, symbol, source-chain, translation, and metadata-inference collapses.
8. The workflow creates no score, ranking, live endpoint, account, remote database, identity-verification service, or production collection mechanism.

## Required Review Sequence

1. Verify source custody and the exact prior disposition.
2. Recheck current facts and legal posture using dated primary or authoritative sources.
3. Review original-language material, transliteration, translation method, alternative interpretations, and translator qualification.
4. Trace source origins and independence; stop at the highest supported attribution rung.
5. Keep actor claims, platform findings, intelligence assessments, allegations, judgments, reach, and effect distinct.
6. Review privacy, due process, expression, association, source protection, affected communities, and non-discrimination.
7. Identify actual correction, appeal, complaint, deletion, notification, compensation, or other remedy routes.
8. Record a reviewer finding without overwriting the prior statement.
9. Conduct accountable editorial adjudication with rationale, publication effect, and appeal route.
10. Publish only after every applicable blocker has direct evidence.

## Translation-Quality Preparation

A translation record must retain original wording, script, direction, transliteration where useful, working translation, translator identity or consented anonymity, qualification, date, method, uncertainty, competing interpretations, and claim support. Model-generated or unreviewed translations remain drafts. English-only convenience wording cannot establish legal equivalence, local acceptance, or cultural fit.

## Public-Wording Test Classes

The machine tests cover actor claim, platform finding, intelligence assessment, court allegation, formal judgment, reach, causal effect, leaderless role, symbol and membership, metadata inference, source-chain independence, and translation uncertainty. They test structure, not the truth of any underlying case.

## Machine Records

- [Review-readiness index](../../../evidence-attribution/review-readiness/index.json)
- [Fairness register](../../../evidence-attribution/review-readiness/fairness.json)
- [Append-only history](../../../evidence-attribution/review-readiness/history.json)
- [Public methodology](/evidence-attribution/reviews)

## Safety and Non-Scoring Boundary

No country, reviewer, community, organization, source, susceptibility, threat, loyalty, morality, legitimacy, protected-characteristic, recruitability, or targetability score is permitted. No operational intrusion, surveillance, evasion, targeting, coercion, weapons, recruitment, or influence-execution instructions may be published. Private people must not be identified or profiled.

## Validation Performed

Local validation checks source hashes, 12 packets, 96 source questions, 216 claim questions, 72 empty adjudication slots, 12 source requirements, 12 wording tests, null reviewer-owned fields, schema conformance, reciprocal links, pointer records, discovery, shell continuity, and package equality. These checks do not perform human review.

## Memory References

- [Review-readiness hot memory](../../../.uai/evidence-attribution-review-readiness.uai#current-publication-contract)
- [Evidence-attribution method](../../../.uai/evidence-attribution.uai#current-publication-contract)
- [Source handling](../../../.uai/source-handling.uai#evidence-attribution-and-leaderless-network-collection)
- [International fairness](../../../.uai/international-fairness.uai#core-rule)
- [Durable pointer ledger](../../../.uai/long-term-memory.uai#evidence-attribution-review-readiness)

## Related Durable Documents

- [Cross-source fairness and unresolved approval](evidence-attribution-review-packets-cross-source-fairness-and-unresolved-approval.md#findings)
- [2.0.29 integration report](evidence-attribution-oversight-and-leaderless-network-research-integration-report.md#findings)
- [OSINT verification wrapper](osint-verification-and-attribution-under-uncertainty.md#reviewed-synthesis)
- [Comparative oversight wrapper](comparative-intelligence-oversight-transparency-and-remedy.md#reviewed-synthesis)

## Supersession Status

Current protocol until replaced by a later protocol that preserves every source, prior disposition, packet, slot, and append-only event.