v0.6.4 — Clean Zenodo DOI metadata

What's Changed

Fix: Simplifies .zenodo.json to unblock the first Zenodo software DOI via the GitHub–Zenodo integration.

Changes

• Removed the related_identifiers block from .zenodo.json to avoid cross-record DOI validation during initial DOI assignment.
• Kept all core Zenodo-facing metadata stable (title, description, creators, license, keywords, publication date, version 0.6.1).
• No source, schema, protocol, or case artifacts were changed.

Intent

This is a metadata-only retry release intended solely to obtain a clean first Zenodo software DOI for the Human Influence Telemetry repository. Once Zenodo successfully mints the DOI for this release, a subsequent version can reintroduce related_identifiers with explicit links to the originating research DOI and related software.

---

Full Changelog: v0.6.3...v0.6.4

v0.6.3 — Initial Zenodo DOI release

What's Changed

Fix: Simplifies Zenodo metadata to allow the first software DOI to be assigned successfully from the GitHub–Zenodo integration.

Changes

• Removed the related_identifiers block from .zenodo.json to avoid cross-record validation issues during the initial DOI minting.
• Kept all core metadata stable (title, description, creators, license, keywords, publication date, and version 0.6.1).
• No source, schema, contract, or case materials were modified.

Intent

This release is a metadata-only patch to obtain the first Zenodo software DOI for the Human Influence Telemetry repository via the automated GitHub integration. Once Zenodo successfully assigns the DOI for this release, a subsequent version can restore structured cross-references (e.g., to the originating research DOI and related software projects) using related_identifiers.

---

Full Changelog: v0.6.2...v0.6.3

v0.6.2 — Zenodo metadata compliance fix

What's Changed

Fix: Corrects .zenodo.json metadata to resolve Zenodo DOI release failures.

Changes

• Removed invalid scheme fields from related_identifiers (not part of Zenodo's .zenodo.json spec)
• Added required publication_date field (2026-07-19)
• Bumped version to 0.6.1 in metadata

No functional changes

This is a metadata-only patch. No schema, specification, or source files were modified.

---

Full Changelog: v0.6.1...v0.6.2

v0.6.1 — Zenodo metadata fix

Patch: adds scheme fields to related_identifiers in .zenodo.json for Zenodo DOI validation

Human Influence Telemetry v0.6.0: First Human Inter-rater Result

Human Influence Telemetry v0.6.0: First Human Inter-rater Result

Summary

Version 0.6.0 publishes the first completed human inter-rater exercise under locked protocol HIT-IRP-CIGNA-001.

Two eligible independent scorers applied the frozen Cigna PxDx packet under specification and schema 0.1.0. The deterministic pre-adjudication comparison produced 7 of 7 exact agreements, an exact-agreement proportion of 1.0000, and zero critical disagreements. The predeclared advancement gate passed.

HIT advances to Maturity Level 2, Applicable. Claim H3 is supported for this bounded exercise.

Version layers

• Repository release: 0.6.0
• Conformance engine: 0.5.0
• Specification: 0.4.0
• Assessment schema: 0.4.0
• Dimension catalog: 0.4.0
• Locked scorer contract used by this exercise: 0.1.0
• Research maturity: Level 2, Applicable

Published result artifacts

The repository contains:

• both verified scorer JSON submissions;
• original and corrected receipt records;
• manual-to-JSON transcription audit;
• scorer transcription confirmations;
• deterministic pre-adjudication comparison in JSON and Markdown;
• comparison execution record;
• preservation manifest;
• no-disagreement adjudication record;
• H3 and maturity decision;
• release-asset manifest.

This release includes HIT-v0.6.0-human-inter-rater-result-assets.zip. The archive preserves the unchanged DOCX and PDF manual submissions together with the public result records.

Release asset SHA-256:

de655f5766009e8cd8a7c52652c53cc7930586f072007839877d542f4138617f

Primary result

• Exact agreements: 7 / 7
• Exact-agreement proportion: 1.0000
• Required minimum: 6 / 7 and 0.8571
• Critical disagreements: 0
• Advancement threshold met: yes
• Supplementary Cohen's kappa: null

Kappa is undefined because both scorers assigned the same category across all six substantive dimensions. The data contain no category variance. The predeclared primary exact-agreement measure remains defined and passed.

Research claim

The release supports this narrow proposition:

Two independent reviewers applied the published locked materials to one frozen packet and reached the predeclared agreement threshold.

The release does not establish general inter-rater reliability, causal validity, evidence truth, legal correctness, field effectiveness, certification, or independent institutional adoption.

Contract and compatibility

Version 0.6.0 does not change the current 0.4.0 specification, schema, catalog, handbook, or scoring semantics. It does not change the 0.5.0 conformance engine. Historical 0.1.0 case assessments remain immutable.

Validation

Before tagging, the exact release commit must pass:

python scripts/check_fixture_format.py
python scripts/validate.py
python -m src conformance --all
python scripts/validate_v050_cli.py

The release validator checks the preserved historical assessments, locked protocol, final scorer JSON files, declared SHA-256 digests, comparison recomputation, threshold decision, H3 wording, maturity status, and release metadata.

The Scorer A DOCX files preserve the exact submitted bytes and therefore retain their original Microsoft Office package metadata, including a lastModifiedBy value naming the packet coordinator. This metadata field records the account associated with the file’s last save operation. It does not identify the scorer or establish authorship of the findings. Submission provenance is established by the preserved receipt, scorer-authored correction confirmation, transcription verification, and SHA-256 records.

Human Influence Telemetry v0.5.0 — Executable Assessment Conformance

Human Influence Telemetry v0.5.0 — Executable Assessment Conformance

Summary

Version 0.5.0 is an implementation release. It adds reusable complete-record conformance for the stable 0.4.0 normative assessment contract.

The repository and conformance engine advance to 0.5.0. The specification, assessment schema, dimension catalog, handbook, and scoring semantics remain 0.4.0.

Added

• reusable semantic validation for complete HIT assessment records;
• actor and evidence-claim reference checks;
• actor-to-claim attribution reciprocity;
• finding and evidence-state consistency checks;
• Repair-trigger and Telemetry Integrity derivation checks;
• aggregation-scope and citation-precision checks;
• stable machine-facing error codes;
• deterministic text and JSON reports;
• one valid and fifteen invalid complete-record vectors;
• machine-readable compatibility metadata;
• non-mutating migration planning for historical 0.1.0 records;
• public CLI commands and smoke tests;
• package-oriented src/ architecture.

Commands

python -m src conformance --all
python -m src conformance --path assessment.json
python -m src migration-plan --path historical-assessment.json

Compatibility

A conforming 0.4.0 assessment remains compatible with release 0.5.0. No normative reassessment is required merely because the implementation version changed.

Historical 0.1.0 assessments remain historical-validation-only artifacts. Automatic conversion to 0.4.0 is prohibited. A migration plan requires fresh reassessment and preservation of the original file.

Preserved boundaries

Release 0.5.0 does not modify:

• specification, schema, catalog, or handbook 0.4.0;
• protocol HIT-IRP-CIGNA-001;
• the frozen Cigna packet;
• scorer eligibility, threshold, or critical-disagreement rules;
• historical public case findings;
• H3 or Maturity Level 1.

No 0.4.0 public-case finding is claimed.

Research boundary

Release 0.5.0 does not establish:

• human inter-rater reliability;
• evidence truth or completeness;
• field effectiveness;
• causal validity;
• legal compliance or standards conformity;
• certification;
• independent adoption;
• runtime enforcement;
• signed-receipt interoperability;
• policy-pack harmonization;
• evidence portability.

Validation

Before tagging, the exact release commit must pass:

python scripts/check_fixture_format.py
python scripts/validate.py
python -m src conformance --all
python scripts/validate_v050_cli.py

The release validator checks repository metadata, contract preservation, historical case hashes, protocol locks, compatibility metadata, complete-record conformance artifacts, and release documentation.

Human Influence Telemetry v0.4.0 — Normative Rubric Stabilization

Human Influence Telemetry v0.4.0 — Normative Rubric Stabilization

Summary

Version 0.4.0 is a breaking normative and data-contract release. It converts all 16 ambiguity classes in HIT-ARFR-001 into a synchronized public assessment contract.

Added

• evidence states for affirmative absence, formal presence, operational capability, observed exercise, and indeterminate records;
• explicit thresholds for 0, 1, 2, and IE;
• dimension-specific Counsel, Judgment, Command, Correction, Repair, and Reform rules;
• Repair trigger states;
• separate institutional-record and assessment-packet integrity;
• deterministic overall Telemetry Integrity derivation;
• sampling, aggregation, actor-attribution, contradiction, evidence-reuse, and citation rules;
• canonical 0.4.0 schema, catalog, handbook, and synthetic assessment;
• 48 executable boundary fixtures covering FR-01 through FR-16;
• explicit migration dispositions for every historical public assessment;
• archived copies of the superseded 0.1.0 contract;
• breaking-change review, migration guide, and adjacent-system claim audit;
• ADR-0003 for chronological empirical-result versioning.

Compatibility

A 0.1.0 assessment does not conform automatically to 0.4.0. Migration is a fresh reassessment and must preserve the original file.

The four historical public assessments remain unchanged. Three are historical_version_bound. Cigna is deferred_locked_protocol. This release does not claim any 0.4.0 public-case finding.

Human protocol

Protocol HIT-IRP-CIGNA-001 remains locked under the 0.1.0 scorer contract. The prior planned v0.3.0 result label is superseded; the eventual result must use the next available repository version. Original submissions and the pre-adjudication result must still be published, passing or failing.

Research boundary

HIT remains Maturity Level 1. Version 0.4.0 does not establish:

• human inter-rater reliability;
• field effectiveness;
• causal validity;
• legal compliance or standards conformity;
• certification;
• independent adoption;
• runtime enforcement;
• signed-receipt interoperability;
• policy-pack harmonization;
• evidence portability.

Validation

Before tagging, the exact release commit must pass:

python scripts/check_fixture_format.py
python scripts/validate.py

The validator covers the canonical contract, synthetic example, 48 boundary fixtures, historical case preservation, migration dispositions, protocol lock, adjacent-system claim audit, and synchronized metadata.

Human Influence Telemetry v0.2.1 — Research Readiness and DOI Archive

Human Influence Telemetry v0.2.1

Version 0.2.1 is a maintenance and research-readiness release. It archives the locked human inter-rater protocol, readable deterministic fixtures, recruitment materials, coordinator tooling, an adversarial rubric-friction review, and a separate model-based rubric stress-test package.

Included

• locked protocol HIT-IRP-CIGNA-001;
• frozen scorer packet HIT-IR-CIGNA-PXDX-001;
• scorer-submission schema and deterministic comparison tooling;
• individual validation and SHA-256 receipt tools;
• human scorer recruitment package;
• adversarial rubric-friction review HIT-ARFR-001;
• readable, canonically formatted fixtures;
• model stress-test protocol HIT-MST-CIGNA-001;
• model run prompt and submission schema;
• explicit separation between human reliability evidence and model-based exploratory testing.

Component versions

• repository release: 0.2.1
• HIT specification: 0.1.0
• assessment schema: 0.1.0
• dimension catalog: 0.1.0
• human inter-rater protocol: 1.0.0
• model stress-test protocol: 1.0.0
• adversarial friction review: 1.0.0

Demonstrated

This release demonstrates that HIT provides machine-readable assessment artifacts, public case applications, a locked human inter-rater design, deterministic comparison tooling, readable fixtures, bounded recruitment procedures, coordinator preservation controls, and an exploratory model-stress-test protocol.

Not demonstrated

This release does not demonstrate human inter-rater reliability, model validity, causal effectiveness, legal compliance, certification, prospective validation, or independent institutional adoption.

HIT remains at Maturity Level 1 until the locked two-human exercise is completed under its predeclared rules.

DOI purpose

The release is suitable for software archival. A software DOI identifies the released repository artifact. It does not certify the method or imply completion of the human reliability exercise.

Human Influence Telemetry v0.2.0 — Public Evidence Pack

Human Influence Telemetry (HIT) v0.2.0 publishes the first public evidence pack for the standalone repository. It applies the unchanged HIT 0.1.0 specification and assessment schema to three publicly documented institutional decision processes.

Included in this release

• Dutch childcare-benefits harm-period case study
• Obermeyer population-health case study, with deploying institutions and the manufacturer assessed separately
• Cigna PxDx case study, including a designed Command disagreement for later inter-rater testing
• Four actor-specific machine-readable assessments
• Public source provenance and explicit evidence boundaries
• Evidence-gated roadmap through v1.0.0
• Validation coverage for every public case assessment
• Updated release, citation, provenance, and DOI documentation

Component versions

• Repository release: 0.2.0
• HIT specification: 0.1.0
• Assessment schema: 0.1.0
• Dimension catalog: 0.1.0

The component versions remain at 0.1.0 because this release adds evidence artifacts without changing the normative construct model, schema fields, dimension catalog, or scoring semantics.

Demonstrated

This release demonstrates that:

1. HIT can be applied to heterogeneous public documentary records.
2. One case can be decomposed into separate institutional actors rather than averaged into one profile.
3. A near-total IE profile can be represented without treating missing records as demonstrated absence.
4. Case-derived disagreements can be preserved explicitly for later reliability testing.
5. Four public case assessments validate against the 0.1.0 schema.
6. The repository can validate fixtures, negative cases, public assessments, release files, and metadata through one command.

Not demonstrated

This release does not establish:

• inter-rater reliability
• prospective institutional effectiveness
• causal effects on harm, correction, repair, or reform
• legal liability or compliance
• standards conformity or certification
• completeness or truth of institution-controlled records
• independent institutional adoption
• superiority over other human-oversight methods

Validation

python -m pip install --requirement requirements-dev.txt
python scripts/validate.py

Human Influence Telemetry v0.1.0

Human Influence Telemetry v0.1.0

Human Influence Telemetry (HIT) is a documentary assurance method for testing whether formal human oversight retained practical force in an AI-mediated institutional decision process.

Included in this release

• working specification v0.1.0;
• six substantive dimensions: Counsel, Judgment, Command, Correction, Repair, and Reform;
• cross-cutting Telemetry Integrity assessment;
• four findings: absent (0), ceremonial (1), substantive (2), and insufficient evidence (IE);
• machine-readable assessment schema and dimension catalog;
• three deterministic fixtures covering substantive influence, ceremonial review, and insufficient evidence;
• application handbook;
• research protocol and claim register;
• limitations and provenance records;
• automated repository validation;
• governance, contribution, security, and conduct policies;
• software citation and Zenodo metadata.

Demonstrated

The release demonstrates that:

1. a HIT assessment can represent all six substantive dimensions and Telemetry Integrity;
2. the repository distinguishes IE from demonstrated absence;
3. included fixtures validate deterministically;
4. the specification, schema, catalog, fixtures, and metadata can be checked through one repository command.

Not demonstrated

The release does not establish:

• inter-rater reliability;
• prospective institutional effectiveness;
• a causal relationship between HIT findings and reduced harm;
• legal compliance or standards conformity;
• certification;
• independent institutional adoption;
• superiority over other human-oversight methods.

Validation

python -m pip install --requirement requirements-dev.txt
python scripts/validate.py

Expected result:

HIT validation: PASS

Citation

Use CITATION.cff for repository citation metadata. The originating research concept is archived at DOI 10.5281/zenodo.21204892. A separate Zenodo software concept DOI and version DOI should be assigned when this GitHub release is archived.

Upgrade and compatibility

This is the first public working release. Future breaking changes to schema fields, dimension definitions, or scoring semantics will require a new minor or major version with migration notes.