artifacts/standard-named

Rendered from markdown source. Open raw source on GitHub.

HI-8 Validation Report

Status: completed Validation date: 2026-07-15

Schema validation

Twelve candidate Draft 2020-12 schemas were checked:

• model-role-authority-ledger.schema.json
• semantic-operator-output.schema.json
• model-field-map.schema.json
• standing-preserving-summary.schema.json
• option-agenda-generation-witness.schema.json
• minority-field-retention-profile.schema.json
• sycophancy-synthetic-consensus-event.schema.json
• provider-telos-disclosure.schema.json
• model-disagreement-record.schema.json
• model-tool-action-boundary.schema.json
• model-correction-refusal-record.schema.json
• model-mediated-decision-witness.schema.json

Result:

schemas checked: 12
schema errors: 0

Positive demonstrations

All eight required demonstrations passed:

1. participant correction changed the active standing-preserving summary;
2. a low-frequency minority field survived compression;
3. a model-generated option retained generation lineage and participant-adoption status;
4. sycophantic agreement was rejected as field evidence;
5. an operative provider constraint was disclosed as a separate field element;
6. two models disagreed without either acquiring standing or decision authority;
7. a plausible model recommendation was blocked at the action boundary;
8. correction propagated through the field map, summary, route conditions, and witness.

Result:

positive demonstrations passed: 8
positive failures: 0

Negative conformance cases

All twelve prohibited patterns were detected:

1. model output inheriting participant standing;
2. summary presented as original source;
3. inference recorded as consent;
4. minority field erased by low frequency;
5. generated option presented as community-authored;
6. sycophantic agreement treated as evidence;
7. provider telos concealed behind neutral-model language;
8. multi-model agreement treated as authority;
9. self-assigned role escalation;
10. route recommendation directly authorizing tool action;
11. participant refusal bypassed through reformulation;
12. correction failing to propagate into the operative field and witness.

Result:

negative cases detected: 12
undetected negative cases: 0

Model-mediated witness

The event-generated ModelMediatedDecisionWitness preserves:

• model-role and authority ledgers;
• field-map versions;
• summary versions;
• minority retention;
• generated-option lineage;
• sycophancy detection;
• provider disclosure;
• model disagreement;
• tool boundary;
• participant correction;
• selected route;
• action and consequence;
• explicit zero standing and authority effect.

Result:

model-mediated witness: PASS
generated_from_events: true
standing effect: none
authority effect: none

Independent export

The independent export contains:

• model ledgers;
• field maps;
• standing-preserving summaries;
• correction record;
• minority-field profile;
• generated option and agenda witness;
• sycophancy event;
• provider disclosure;
• model disagreement;
• tool boundary;
• eight demonstrations;
• event stream;
• model-mediated witness;
• schemas;
• manifest;
• checksums.

Result:

checksum files verified: 33
checksum failures: 0
independent of model provider: true

Public-page consistency review

H.8 was checked against I.8 for the following claims:

• the model map remains a projection of projections;
• fluency does not create authority;
• source, inference, summary, generation, recommendation, and authorization remain distinct;
• standing does not transfer through summary;
• minority fields survive low-frequency compression;
• generated options preserve authorship;
• sycophantic agreement is not field evidence;
• provider purposes and constraints remain visible;
• model plurality is not standing plurality;
• recommendation does not authorize action;
• human confirmation does not cure hidden field collapse;
• participant correction changes the operative map.

Result:

technical/public contradictions found: 0

Scope of validation

This report establishes structural and internal semantic validation only.

It does not establish:

• model neutrality;
• complete source fidelity;
• absence of all sycophancy;
• full provider-telos disclosure;
• model independence;
• production safety;
• legal authority for public decision;
• meaningful human control in every implementation;
• public-reader comprehension;
• scientific proof of the broader framework.

Conditions before production use

• authenticate role and authority grants;
• implement output-class labels in the interface and event stream;
• test source fidelity against participant-reviewed records;
• test minority retention under aggressive compression;
• disclose material provider constraints at runtime;
• test sycophancy under user and authority pressure;
• enforce the action gate outside the model;
• test correction propagation through summaries, options, tools, and witnesses;
• export the witness independently of the provider;
• conduct independent reader testing of H.1 through H.8.