v0.4.0: documentation and red-team report

Clarify exploratory scope, evidence limits, source preservation gaps, and the v0.4.0 red-team remediation.

docs/codebook.md

@@ -6,10 +6,10 @@
 |---|---|---|
 | `research_view` | pathway | Short analytic view for comparison |
 | `case_catalog` | catalog record | Distinguishes core, leads, and empirical studies |
-| `pathways` | control-transfer proposition | Core agency-transfer analysis |
+| `pathways` | falsifiable control-shift hypothesis | Separates observed mechanism, attempted transfer, observed agency change, preservation, and insufficient evidence |
 | `observations` | one quantity | Reach, production, engagement, response, or context |
 | `claims` | atomic assertion | Evidence status and wording |
-| `claim_evidence` | claim–source link | Normalized support/counterevidence |
+| `claim_evidence` | claim–source link | Normalized support/premise/counterevidence |
 | `model_evaluations` | empirical study | Capability or experimental evidence |
 | `official_election_metrics` | one official metric | Legally and semantically explicit electoral context |
 
@@ -99,12 +99,16 @@ The screening register uses more granular provisional codes; normalize before pr
 
 - `occurrence_status`: `confirmed | partly_confirmed | alleged | not_observed`
 - `behavioural_effect_status`: `measured | indicated | not_detected | unknown`
-- `electoral_effect_status`: `measured | institutionally_asserted | not_detected | unknown`
+- `electoral_effect_status`: `measured | not_detected | unknown`
 
-`institutionally_asserted` records institutional reliance or wording. It is not a causal estimate.
+Institutional reliance or action belongs in `institutional_outcome`; it is not an electoral-effect category or causal estimate. No v0.4.0 core record has a measured electoral effect.
 
 Legacy `reach_estimate` and `reach_unit` remain for compatibility. Use `observations` for analysis.
 
+`control_shift_hypothesis` and `hypothesized_power_recipient` are analytical fields. Their presence does not mean that a control shift, concentration of power, or agency outcome occurred. The maximum permitted conclusion is stored in `pathways`.
+
+`affected_state` names the decision-relevant dimension examined; it is not a claim that the state changed. `democratic_harm` must distinguish documented process or institutional harm from plausible but unmeasured individual or electoral harm.
+
 ## Claims
 
 ### `claim_type`
@@ -121,38 +125,44 @@ Legacy `reach_estimate` and `reach_unit` remain for compatibility. Use `observat
 - `response`
 - `legal_status`
 - `counterevidence`
-- `agency_transfer`
+- `control_shift_hypothesis`
 
 ### `evidence_label`
 
 - `established_evidence`: supported as worded.
 - `established_as_campaign_report`: establishes the attributed report, not independent population validity.
 - `strong_inference`: well supported but not directly observed/adjudicated.
+- `researcher_hypothesis`: falsifiable analytical proposition; not an observed outcome and not directly source-supported.
+- `not_identified_in_declared_review`: no supporting source was found among sources linked to that record by the cutoff; not proof of global absence.
 - `open_question`: unresolved and not filled by assumption.
 
 ### `claim_status`
 
-- `supported`
+- `supported_as_worded`
 - `supported_as_attributed_measurement`
 - `partly_supported`
+- `not_identified_in_declared_review`
+- `hypothesis_not_tested`
 - `unresolved`
 
-### `evidence_strength`
+### `evidence_basis`
 
-- `high`
-- `moderate`
-- `low`
+- `supported_as_worded`
+- `attributed_measurement_only`
+- `inferential_synthesis`
+- `review_bounded_absence`
+- `not_tested`
 - `none`
 
-`coder_confidence_in_label` ranges from 0 to 1 and refers to the coding decision. It is not a truth probability and should not be averaged into a case score.
+`evidence_basis` is not an ordinal strength score. v0.4.0 removes `confidence`, `coder_confidence_in_label`, and case-level `overall_confidence`; no calibrated probability or intercoder-reliability statistic is available.
 
 ## Claim evidence
 
 ### `relation`
 
 - `supports`
+- `premise`: the source supports a factual premise used by a `researcher_inference`; it does not mean the source states or proves the inference.
 - `counterevidence`
-- `qualifies` where retained in compatible tooling
 
 ### `evidence_scope`
 
@@ -160,6 +170,7 @@ Legacy `reach_estimate` and `reach_unit` remain for compatibility. Use `observat
 - `assertion_was_made`
 - `institutional_action`
 - `measurement`
+- `review_scope_statement`
 - `researcher_inference`
 
 ### `directness`
@@ -168,7 +179,7 @@ Legacy `reach_estimate` and `reach_unit` remain for compatibility. Use `observat
 - `official_record_or_allegation`
 - `independent_secondary_reporting`
 
-An official allegation proves that the allegation was made. It does not automatically prove the alleged underlying conduct.
+An official allegation, party filing, platform self-report, campaign statement, or legal defence can establish only that the assertion was made unless independent evidence supports the underlying conduct. `review_scope_statement` is a bounded statement about the linked corpus, not evidence of absence outside it.
 
 ## Sources
 
@@ -190,12 +201,16 @@ Quality, independence, directness, and scope are separate concepts.
 ### Locators and preservation
 
 - `claim_evidence.locator`: stable pinpoint only when verified, such as page, paragraph, timestamp, table, section, or spreadsheet cell.
-- blank `locator`: document- or page-level support only; it is not paragraph-level verification.
+- `claim_evidence.locator_coverage`: `full_claim` when the pinpoint covers the complete coded wording, `partial_claim` when it covers only part, and `not_verified` when no honest pinpoint was established.
+- `claim_evidence.locator_note`: required explanation for `not_verified`, such as an inaccessible source, no supporting passage identified, or a dynamic metric without a stable snapshot.
+- blank `locator`: no verified pinpoint; it is not paragraph-level verification.
 - `archive_url` / `content_hash`: populated only for a lawfully captured, exact source object.
 - official `content_sha256`, `content_bytes`, and `source_schema_version`: populated only for a verified downloaded object.
 
 This release keeps URLs but does not contain a source-snapshot archive. Blank preservation fields are known limitations, not missing-value imputation targets.
 
+At the v0.4.0 cutoff, 57/124 generated claim–source relations have a verified full-claim locator, 47/124 a verified partial/premise locator, and 20/124 remain `not_verified` because the cited object was inaccessible during the audit. Relations lacking a supporting passage were removed or recoded as premises; a mutable platform count without a snapshot was removed. Preservation coverage remains zero: 0/70 sources have an `archive_url` or `content_hash`, and 0/31 official-data sources have verified `content_sha256`, `content_bytes`, or `source_schema_version`. These counts describe the current snapshot; locator coverage and source preservation are different dimensions.
+
 ## Actors
 
 `case_actors` separates actor name, role, type, attribution status, and control dimension. Examples of roles include commissioner, operator, sponsor, model provider, platform/ranker, telecom infrastructure, regulator/responder, and represented target.
@@ -234,31 +249,63 @@ An unknown version must be explicit in `notes`. Do not infer model provider, ope
 
 `content_items` records media type, represented person, communication goal, representation mode, AI disclosure, authorization, disclosure salience, and presentation mode.
 
-Disclosure and authorization are separate. Transparent parody can still appropriate first-person representation, while authorized synthesis can preserve or extend the represented person's agency.
+Disclosure and authorization are separate. Transparent parody can still raise a representation hypothesis, while authorized or well-disclosed synthesis may preserve or extend agency. Neither conclusion is coded as observed without a relevant outcome measure.
 
 ## Pathways
 
-### `agency_transfer_status`
+Pathway states are separate variables. Never collapse them into one score or treat them as an automatic causal ladder.
+
+### `capability_status` and `control_mechanism_status`
+
+- `observed`
+- `not_observed`
+- `not_assessed`
+- `not_applicable`
+
+Capability means the relevant technical or organizational capacity exists in the record. A control mechanism additionally requires documented use capable of controlling a decision-relevant input, representation, incentive, or interface.
+
+### `attempted_transfer_status`
+
+- `observed`: evidence documents an attempt to displace, coerce, impersonate, suppress, or covertly control a decision-relevant capacity. Routine attributed persuasion is insufficient.
+- `not_observed`: the record contains no observed attempt; this is not proof that none occurred.
+- `insufficient_evidence`: intent, control, authorization, or boundary criteria are unresolved.
+- `not_applicable`: preparedness or other non-incident.
 
-- `full_chain_observed`
-- `partial_chain_observed`
-- `mechanism_observed_effect_unknown`
-- `mechanism_inferred`
-- `plausible_hypothesis`
-- `no_transfer_observed`
+### `agency_change_status` and `agency_preservation_status`
+
+- `observed`
+- `not_observed`
+- `insufficient_evidence`
+- `not_applicable`
+
+`agency_change_status=observed` requires a direct person-level or otherwise valid decision-relevant outcome—not content, reach, engagement, institutional alarm, or an election result alone. `agency_preservation_status=observed` requires evidence that a mechanism improved or maintained informed control, contestability, representation, or decision support. Disclosure alone is insufficient.
+
+### `maximum_conclusion`
+
+- `control_mechanism_observed`
+- `attempted_transfer_observed`
+- `agency_change_observed`
+- `agency_preserved_or_extended`
 - `insufficient_evidence`
 - `not_an_incident`
 
+This is the strongest conclusion permitted by the coded evidence, not a severity rank. v0.4.0 contains four `control_mechanism_observed`, two `attempted_transfer_observed`, two `insufficient_evidence`, and two `not_an_incident` records. It contains zero observed agency changes and zero observed preservation/extension outcomes.
+
+### Falsifiability fields
+
+- `falsification_condition`: an observation that would count against the case-specific control-shift hypothesis or block a stronger conclusion.
+- `comparator`: the exposed/unexposed, synthetic/authentic, disclosed/undisclosed, authorized/unauthorized, or other design contrast needed to test agency change or preservation.
+- `observed_evidence`: only what linked evidence directly documents.
+- `researcher_inference`: the analytical proposition kept separate from observations.
+
 ### `assessment_label`
 
-- `established_evidence`
-- `strong_inference`
-- `plausible_hypothesis`
-- `speculative_scenario`
-- `open_question`
+- `observed_mechanism`
+- `observed_attempt_with_unmeasured_agency`
+- `boundary_case`
 - `not_applicable`
 
-The pathway fields describe a proposition about control. `observed_outcome` must state what is actually observed and preserve unknown behavioral/electoral effects.
+Transparent contested uses are `boundary_case` and cannot receive an observed-attempt or agency-change conclusion merely because a campaign controlled a synthetic representation.
 
 ## Observations
 

docs/ethics-and-release.md

@@ -1,8 +1,10 @@
 # Ethics and release
 
+**Release 0.4.0 · 2026-08-15**
+
 ## Harm minimisation
 
-This dataset is designed for democratic-integrity and agency research. It does not reproduce deceptive media, voter lists, targeting data, operational playbooks, or instructions that would make interference easier. Source URLs are retained for verification, but the dataset does not archive or redistribute copyrighted or harmful media.
+This index is designed for democratic-integrity and agency research. It does not reproduce deceptive media, voter lists, targeting data, operational playbooks, evasion guidance, or instructions that would make interference easier. Source URLs support inspection, but the release does not archive or redistribute copyrighted or harmful media.
 
 ## Naming actors
 
@@ -16,8 +18,16 @@ Preparedness files are explicitly excluded from incident counts. Model-capabilit
 
 ## Research claims
 
-The release does not estimate global prevalence, votes changed, or election winners caused by AI. Platform metrics are not unique people; official results are not causal designs; institutional alarm is not behavioural evidence. These restrictions are enforced in the audit and documented as part of the scientific result.
+The release is a purposive, exploratory evidence index. It does not empirically prove agency transfer, estimate global or regional prevalence, identify votes changed, or attribute an election winner to AI. Platform metrics are not unique people; official results are not causal designs; institutional alarm is not behavioural evidence. Pathway assessments are transparent researcher syntheses, not measured outcomes.
+
+The automated audit performs internal structural and semantic QA. It cannot verify source truth, adjudicate an allegation, establish that a source supports every word of a claim, make purposive sampling representative, or validate causality. Those limits must remain visible even when the audit reports zero blocking errors.
+
+## Provenance limits
+
+Public links are retained, but source objects are not frozen in this repository. Among 124 claim relations, 57 pinpoints cover the full claim, 47 cover only a factual premise or part and 20 are not verified because the cited object was inaccessible; `locator_coverage` makes that limit machine-readable. Blank archive URLs and hashes remain explicit gaps. A user must inspect and cite the underlying source before relying on a substantive claim. The project should add lawful snapshots and hashes where possible without redistributing harmful or copyrighted material.
 
 ## Corrections
 
-Corrections should identify the affected record, provide supporting evidence, and preserve a changelog. The dataset is versioned so that retrospective revisions do not silently overwrite what was known at a given date.
+Corrections should identify the affected record, provide supporting evidence, retain counterevidence and preserve a changelog. `main` is mutable; numbered tags are the citable research revisions and should never be moved after publication. Each release must reconcile tables, documentation, citation metadata, manifest and packaged artifacts.
+
+No DOI is assigned to v0.4.0. A future DOI should identify a frozen deposit and be added only after its tag, commit, hashes and citation record agree.

docs/guia-rapida-es.md

@@ -1,71 +1,90 @@
 # Guía rápida para investigación
 
-## Qué es y qué no es
+**Versión 0.4.0 · 15 de agosto de 2026**
 
-Esta versión es un índice de evidencia trazable sobre IA, elecciones y transferencia de agencia. Permite comparar mecanismos, actores, sistemas, objetivos, métricas seleccionadas y límites causales.
+Este es un **índice exploratorio de evidencia seleccionada**. Sirve para rastrear afirmaciones, fuentes, incertidumbre y mecanismos. No demuestra empíricamente la transferencia de agencia, no estima prevalencia y no permite calcular votos o resultados causados por IA.
 
-No es un censo mundial, una muestra representativa, ni una estimación de votos cambiados. Los 10 registros codificados incluyen unidades distintas y 2 son expedientes de preparación, no incidentes.
+Los datos y campos codificados están en inglés. Esta guía y algunos metadatos son la única capa en español.
 
-## Las cinco tablas que conviene abrir primero
+## Empiece aquí
 
-1. `research_view.csv`: resumen analítico corto, una fila por vía de transferencia de agencia.
-2. `case_catalog.csv`: separa núcleo codificado, candidatos de investigación y estudios empíricos.
-3. `pathways.csv`: quién controla el sistema, qué capacidad afecta y dónde se concentra el poder.
-4. `observations.csv`: cifras con unidad, denominador, periodo, unicidad, método y estatus causal.
-5. `claim_evidence.csv`: relación normalizada entre cada afirmación y sus fuentes.
+1. [Vista de investigación](https://huggingface.co/datasets/apol/ai-election-manipulation-cases/viewer/research_view/data): 10 registros resumidos.
+2. [Catálogo completo](https://huggingface.co/datasets/apol/ai-election-manipulation-cases/viewer/case_catalog/data): separa núcleo, candidatos y estudios de modelos.
+3. [Vías analíticas](https://huggingface.co/datasets/apol/ai-election-manipulation-cases/viewer/pathways/data): mecanismo, control y estatus epistémico.
+4. [Afirmaciones](https://huggingface.co/datasets/apol/ai-election-manipulation-cases/viewer/claims/data) y [relaciones de evidencia](https://huggingface.co/datasets/apol/ai-election-manipulation-cases/viewer/claim_evidence/data): base para cualquier uso factual.
+5. [Observaciones](https://huggingface.co/datasets/apol/ai-election-manipulation-cases/viewer/observations/data): cantidades seleccionadas con unidad, denominador y estatus causal.
 
-En Excel, estas capas aparecen primero y el detalle queda después.
+El proyecto forma parte de la [colección Agency Transfer](https://huggingface.co/collections/apol/agency-transfer).
 
-## Cómo leer una fila correctamente
+## Mapa de las 26 configuraciones
 
-Antes de usar una cifra o conclusión, pregunte:
+| Uso | Configuraciones |
+|---|---|
+| Orientación y cobertura | `research_view`, `case_catalog`, `coverage_summary`, `sampling_frame` |
+| Afirmaciones y evidencia | `cases`, `claims`, `sources`, `claim_evidence`, `analytic_record_claims` |
+| Mecanismos y procedencia | `pathways`, `observations`, `events`, `case_sources`, `case_actors`, `technology_uses`, `content_items` |
+| Contexto electoral | `case_elections`, `election_sources`, `official_elections`, `official_election_metrics`, `official_turnout`, `official_results`, `official_data_sources` |
+| Descubrimiento | `candidates`, `watchlist` |
+| Estudios de modelos | `model_evaluations` |
 
-- ¿La unidad es un incidente, una red, una elección completa, un conjunto de campaña o un expediente de preparación?
-- ¿La IA generó el contenido, lo amplificó, solo apoyó el proceso o fue usada defensivamente?
-- ¿Quién controlaba el despliegue y qué papel tenía el proveedor?
-- ¿La representación era autorizada, transparente, satírica o engañosa?
-- ¿La cifra cuenta vistas, cuentas, llamadas, preguntas, personas únicas o votos?
-- ¿Existe un diseño causal o solo una medición descriptiva?
-- ¿La fuente prueba el hecho o únicamente que una institución hizo una alegación?
-- ¿Qué contradice, limita o cualifica la afirmación?
+## Cargar una versión fija
 
-## Regla central de inferencia
+```python
+from datasets import load_dataset
 
-No transforme esta secuencia:
+vista = load_dataset(
+    "apol/ai-election-manipulation-cases",
+    "research_view",
+    revision="v0.4.0",
+    split="data",
+)
+```
 
-`contenido observado → alcance de plataforma → alarma institucional → resultado electoral`
+Con pandas o Polars:
 
-en esta afirmación:
+```python
+import pandas as pd
+import polars as pl
 
-`la IA cambió votos`.
+url = "https://huggingface.co/datasets/apol/ai-election-manipulation-cases/resolve/v0.4.0/parquet/research_view.parquet"
+vista_pd = pd.read_parquet(url)
+vista_pl = pl.read_parquet(url)
+```
+
+`main` cambia con el proyecto. Para una cita o análisis reproducible, use el tag `v0.4.0` y anote también el commit exacto.
+
+## Regla de lectura
 
-Cada flecha necesita evidencia propia. En la mayoría de los registros, la ocurrencia y la respuesta están bien documentadas, pero los efectos conductuales y electorales siguen siendo desconocidos.
+No convierta:
 
-## Estudios de modelos
+`contenido observado → alcance → respuesta institucional → resultado electoral`
+
+en:
+
+`la IA cambió votos`.
 
-Los estudios de `model_evaluations.csv` son una capa separada. Algunos prueban exactitud, sesgo lingüístico, asesoramiento electoral o capacidad persuasiva experimental. Incluso un experimento aleatorizado solo identifica el efecto de esa interacción, en esa muestra y sobre ese resultado medido. No demuestra un despliegue clandestino ni un cambio en el resultado de una elección real.
+Cada flecha necesita evidencia propia. Antes de usar una fila, compruebe:
 
-## Datos electorales oficiales
+- la unidad: elección, red, incidente, campaña, preparación o estudio;
+- si la fuente prueba un hecho, registra una alegación o sustenta una inferencia;
+- si la cifra cuenta contenido, vistas, cuentas, llamadas, personas o votos;
+- si existe un diseño causal o solo descripción;
+- qué contraprueba, limitación o atribución desconocida conserva el registro.
 
-Use `official_election_metrics.csv` para comparaciones semánticas. La tabla ancha `official_turnout.csv` se conserva por compatibilidad.
+`pathways` hace visible el razonamiento del investigador; no es un resultado medido. Los antiguos campos numéricos de confianza se eliminaron porque no eran probabilidades calibradas ni una medida de acuerdo entre codificadores. Existe un solo codificador principal y no hay estimación de fiabilidad entre codificadores.
 
-Corrección importante: Moldavia distingue entre electores que recibieron papeleta (`C`) y electores que participaron (`D`). En las elecciones presidenciales y el referéndum de 2024, la validez legal se calculó con `C/(A+B)`; en las parlamentarias de 2025, la tasa oficial usa `D/(A+B)`. La base conserva ambas medidas.
+Resultado central: 4 registros permiten concluir que se observó un mecanismo de control, 2 que se observó un intento de transferencia, 2 quedan con evidencia insuficiente y 2 son archivos de preparación, no incidentes. **Ningún registro codifica un cambio observado de agencia ni una preservación/extensión observada de agencia.**
 
-Los resultados oficiales describen votos, escaños y estatus legal. No identifican exposición ni causalidad.
+## Procedencia y auditoría
 
-## Qué puede afirmar con seguridad
+La versión contiene 70 registros de fuente; 56 fuentes distintas están enlazadas a afirmaciones mediante 124 relaciones en `claim_evidence`. De esas relaciones, 57 tienen un localizador verificado que cubre toda la afirmación, 47 cubren solo una premisa factual o una parte y 20 no tienen pasaje puntual verificado porque el objeto citado fue inaccesible durante la auditoría. `locator_coverage` conserva esa diferencia y `locator_note` explica cada caso no verificado. Ninguno de los 70 registros de fuentes tiene `archive_url` o `content_hash`: un localizador no equivale a una copia congelada.
 
-Puede describir:
+La auditoría automatizada comprueba estructura, identificadores, relaciones, formatos y reglas semánticas internas. **No verifica que una fuente sea verdadera, que respalde cada palabra, que el muestreo sea representativo o que una inferencia causal sea válida.** Los seis avisos sobre búsquedas negativas incompletas bloquean cualquier inferencia de prevalencia.
 
-- que un artefacto, red, llamada, regla o respuesta institucional ocurrió;
-- quién fue identificado y con qué nivel de atribución;
-- qué sistema o proveedor fue documentado y qué parte sigue desconocida;
-- qué métrica se reportó y cuál fue su denominador;
-- qué control se desplazó según evidencia o inferencia explícita;
-- qué resultado fue observado y qué efecto continúa sin medir.
+## Resultados y estudios
 
-No debe afirmar prevalencia global, impacto agregado ni votos cambiados sin otra muestra y un diseño específico.
+Los datos electorales oficiales son contexto descriptivo. No identifican exposición, persuasión ni votos causados por manipulación. Los estudios de `model_evaluations` miden modelos en diseños concretos; no prueban despliegue en campañas reales.
 
-## Estado de calidad
+El archivo Excel es una ayuda de navegación, no la fuente analítica canónica. Su constructor necesita `@oai/artifact-tool`, un entorno externo no incluido en el `package.json` público. Para análisis, use CSV o Parquet.
 
-La auditoría v0.3.1 pasa con cero errores bloqueantes. Mantiene seis avisos deliberados porque no se han registrado búsquedas negativas sistemáticas en seis marcos regionales. Esos avisos son parte del resultado científico: impiden que la cobertura aparente se convierta en una afirmación de representatividad.
+No existe DOI para esta versión. Cite el tag/commit del dataset y las fuentes originales de cada afirmación sustantiva.

docs/methodology.md

@@ -2,9 +2,9 @@
 
 ## Research question
 
-When an election-related AI or algorithmic system changes who controls political representation, attention, information, participation, or decision support, what capability is involved, who controls it, whose agency is affected, where power moves, what harm is observed, and what remains only inferred?
+When an election-related AI or algorithmic system is documented in relation to political representation, attention, information, participation, or decision support, what capability is observed; what control mechanism, if any, is documented; is an attempt to displace a decision-relevant capacity observed; is any agency change measured; and what remains a falsifiable hypothesis?
 
-The dataset does not assume that AI use is manipulation. Manipulation is assessed from process evidence: deception, unauthorised representation, opacity, coercion, exploitation, covert coordination, loss of contestability, or control over a decision environment.
+The dataset does not assume that AI use is manipulation, that control of a communication channel changes a person's agency, or that a documented influence attempt succeeds. It is an evidence index for testing control-shift hypotheses—not evidence that “agency transfer” occurred. Manipulation is assessed separately from agency outcomes using process evidence such as deception, unauthorised representation, opacity, coercion, exploitation, covert coordination, loss of contestability, or control over a decision environment.
 
 ## Research layers and units
 
@@ -25,7 +25,8 @@ A core record must:
 3. have an authoritative primary record plus appropriate corroboration, or multiple credible independent sources;
 4. permit occurrence, actor, mechanism, reach, effect, response, counterevidence, and uncertainty to be coded separately;
 5. identify the unit and election relationship without aggregating unrelated artifacts into one implied operation;
-6. retain the maximum conclusion warranted by the evidence, not the most dramatic interpretation.
+6. retain the maximum conclusion warranted by the evidence, not the most dramatic interpretation;
+7. specify what observation would count against a control-shift hypothesis and what comparator would be needed to test agency change or preservation.
 
 Transparent or authorised AI may be retained as a comparison when it tests the boundary between representation, persuasion, manipulation, and agency preservation.
 
@@ -52,16 +53,16 @@ The release may be described as internally complete for its declared records and
 4. Code creator, commissioner, sponsor, distributor, infrastructure provider, beneficiary, target, and responder separately.
 5. Add counterevidence and competing explanations.
 6. Code actors, technology uses, content items, pathways, and quantities from supported claims.
-7. Re-verify URLs, dates, status, and ongoing proceedings at the release cutoff.
-8. Run deterministic relational, semantic, causal, and official-data checks.
+7. Attempt to recheck URLs, publication dates, access status, and known proceeding status at the cutoff; record blocked, dynamic, or unverified items explicitly.
+8. Run deterministic relational, semantic, causal-safeguard, and official-data consistency checks.
 
 ## Coding and review protocol
 
-Version 0.3.1 was produced by one primary coder with automated consistency checks and targeted record-level review. It does not report intercoder reliability, because no independent second-coder sample was completed for this release. Numeric `coder_confidence_in_label` values are retained only as legacy self-assessment metadata; they are not calibrated probabilities, agreement statistics, or substitutes for source review.
+Version 0.4.0 was produced by one primary coder with automated consistency checks and targeted record-level review. It does not report intercoder reliability because no independent second-coder sample was completed. Numeric case and claim confidence scores were removed: they were uncalibrated self-assessments that created false precision and could not substitute for agreement or source review.
 
-Before any inferential or publication-grade v1.0 release, the protocol requires an independent coder to review a stratified sample spanning record types, regions, claim types, source grades, and evidence labels. Disagreements must be reconciled in a dated adjudication log, and agreement must be reported separately for categorical labels and quantitative extraction. Until then, users should rely on claim wording, evidence relations, source directness, and counterevidence—not numeric coder confidence.
+Before any inferential or publication-grade v1.0 release, the protocol requires an independent coder to review a stratified sample spanning record types, regions, claim types, source grades, and evidence labels. Disagreements must be reconciled in a dated adjudication log, and agreement must be reported separately for categorical labels and quantitative extraction. Until then, users should rely on claim wording, evidence relations, source directness, counterevidence, and the explicit maximum conclusion. No intercoder-reliability claim is made.
 
-`claim_evidence.csv` is the normalized claim–source relation. `relation` distinguishes support from counterevidence; `evidence_scope` distinguishes an underlying fact, official action, measurement, assertion, or researcher inference; `directness` records whether the source is an artifact, measurement, official record/allegation, or secondary report.
+`claim_evidence.csv` is the normalized claim–source relation. `relation=supports` means the source states the coded factual or attributed proposition; `premise` means it supports a factual premise used in a researcher inference but does not state or prove that inference; `counterevidence` qualifies or contests it. `evidence_scope` distinguishes an underlying fact, official action, measurement, attributed assertion, review-scope statement, or researcher inference. An official allegation, party filing, platform statement, or defence supports `assertion_was_made`; it does not establish the alleged conduct. `directness` records whether the source is an artifact, measurement, official record/allegation, or secondary report.
 
 ## Claim status
 
@@ -70,30 +71,38 @@ Before any inferential or publication-grade v1.0 release, the protocol requires
 - `established_evidence`
 - `established_as_campaign_report`
 - `strong_inference`
+- `researcher_hypothesis`
+- `not_identified_in_declared_review`
 - `open_question`
 
-`claim_status` records whether the wording is supported, supported only as an attributed measurement, partly supported, or unresolved. `evidence_strength` is high, moderate, low, or none.
+`claim_status` records whether wording is supported as written, supported only as an attributed measurement, partly supported, an untested hypothesis, a review-bounded non-identification, or unresolved. `evidence_basis` is a non-numeric category; it is not an ordinal score.
 
-The legacy `confidence` field is retained as `coder_confidence_in_label`. It is confidence in the coding decision and wording, not a probability that the underlying proposition is true. It must never substitute for claim type, evidence relationship, independence, directness, or counterevidence.
+`not_identified_in_declared_review` means only that no supporting source was found among the sources linked to that record by the cutoff. It is not evidence of global absence and is intentionally not linked to selected sources as though those sources proved a negative.
 
-## Agency-transfer pathway
+## Falsifiable control-shift pathway
 
-The central sequence is:
-
-`capability → controller → vector → target → affected agency → control shift → power recipient → harm pathway → observed outcome`
+The pathway table separates capability, controller, vector, target, agency dimension, **hypothesized** control shift, possible power recipient, potential harm, and observed outcome. These are fields to test independently, not a causal ladder, and no transition is assumed merely because the fields share a row.
 
 `pathways.csv` records:
 
-- affected agency or capacity;
-- who originally held relevant control and where it moved;
+- the agency dimension examined;
+- who is hypothesized to hold or gain relevant control;
 - technical/social mechanism;
 - opacity, personalization, contestability, and duration;
-- recipient of power;
-- harm pathway and observed outcome;
-- agency-transfer status and epistemic assessment;
+- capability and control-mechanism status;
+- attempted-transfer, observed-agency-change, and agency-preservation status;
+- the maximum conclusion allowed by the evidence;
+- a falsification condition and comparator;
+- observed evidence separated from researcher inference;
 - supporting claim identifiers and reviewer note.
 
-An agency-transfer status other than `not_an_incident` or `insufficient_evidence` requires a specified capacity, mechanism, recipient, and evidence-backed claim. The dataset deliberately does not calculate a single agency-transfer score: such weights would be normative and falsely precise.
+The allowed `maximum_conclusion` values are `control_mechanism_observed`, `attempted_transfer_observed`, `agency_change_observed`, `agency_preserved_or_extended`, `insufficient_evidence`, and `not_an_incident`. They are conclusions, not severity ranks.
+
+An observed capability is not an observed control mechanism. A control mechanism is not an attempted transfer. An attempt is not an agency change. Disclosure, authorization, contestability, or successful user verification can support an agency-preservation hypothesis, but none is treated as proof without an outcome measure. `not_observed` means the record contains no direct observation of that outcome; it does not mean the outcome did not occur.
+
+In v0.4.0, none of the eight incident-count-eligible core records has `agency_change_status=observed` or `agency_preservation_status=observed`. Two support only an observed attempt, four support only an observed control mechanism, and two transparent contested uses remain boundary cases with insufficient evidence. The two preparedness files are `not_an_incident`.
+
+The dataset deliberately calculates no agency-transfer or control-shift score. Such weights would be normative and falsely precise. The renamed `control_shift_hypothesis` claims have no direct source links: their factual premises are linked separately through `source_claim_ids`, and their truth must be tested against the stated comparator.
 
 ## Quantitative observations
 
@@ -107,7 +116,7 @@ Selected decision-relevant quantities are normalized as separate rows in `observ
 - measurement method and source;
 - causal status and limitations.
 
-Accounts, posts, views, followers, likes, calls, questions, users, people, voters, and votes are different units. A platform event is never relabelled as a unique person without explicit deduplication. Descriptive quantities are not causal estimates.
+Accounts, transactions, posts, views, followers, likes, calls, questions, users, people, voters, and votes are different units. A platform event is never relabelled as a unique person without explicit deduplication. Heterogeneous aggregates are not normalized as one observation: the reported Moldova aggregate of more than 138,000 accounts and transactions is retained only as a warning claim and excluded from `observations`. Descriptive quantities are not causal estimates.
 
 ## Empirical model studies
 
@@ -121,7 +130,9 @@ Official election data are descriptive context and outcomes. They are linked to
 
 For each official source, the manifest records the institution, a landing URL, status, format, granularity, access method, retrieval time, snapshot cutoff, and reuse terms. A distinct download URL and optional content hash, bytes, or schema version are included only when captured and verified; blanks are not evidence of a missing public record.
 
-This release does not archive or redistribute the referenced source corpus. Therefore, claim locators are populated only when a stable page, paragraph, timestamp, table, cell, or section was verified against a captured object. Blank locators and content-hash fields are explicit reproducibility gaps. They must not be represented as pinpoint verification or source preservation.
+This release does not archive or redistribute the referenced source corpus. Claim locators are populated only when a page, paragraph, timestamp, table, cell, or section was verified against the cited object. `locator_coverage=full_claim` means that pinpoint covers the full coded wording; `partial_claim` means it covers only part; `not_verified` leaves the locator blank. A verified locator is not a source snapshot or preservation hash.
+
+At the v0.4.0 cutoff, 57/124 claim–source relations have a verified full-claim locator, 47/124 have a verified partial/premise locator, and 20/124 are `not_verified` because the cited object was inaccessible during the locator audit. Relations for which no supporting passage was identified were removed or reclassified as explicit factual premises; the mutable YouTube count was removed because no stable snapshot was preserved. Separately, 0/70 sources have an archive URL or content hash, and 0/31 official-data sources have verified content hashes, byte counts, or source-schema versions. The relational model has substantially better pinpointing, but the source corpus is not independently preserved and partial locators do not validate uncited parts of a claim.
 
 For each measure:
 
@@ -164,16 +175,18 @@ The release blocks these common errors:
 - a hybrid operation does not permit its total outcome to be attributed to AI alone;
 - platform or state attribution is not the same as final adjudication;
 - rapid removal can explain low reach and is not proof of low technical capability;
-- absence is evidence only when the search frame and intensity were recorded.
+- a non-identification is bounded to the declared review and is not evidence of global absence;
+- institutional action, including annulment, is not part of the manipulation mechanism unless a causal relationship is independently established;
+- production or representational control does not by itself establish a change in a voter's, candidate's, or represented person's agency.
 
 ## Quality assurance
 
-The deterministic v0.3.1 audit checks all 26 generated tables for:
+The deterministic v0.4.0 audit is designed to check all generated tables for:
 
 1. explicit record type and manipulation assessment;
 2. exclusion of preparedness and empirical studies from incident counts;
 3. normalized foreign keys and evidence joins;
-4. controller/target/mechanism/status in every pathway;
+4. controller, target, mechanism, separate falsifiable states, maximum conclusion, falsification condition, and comparator in every pathway;
 5. actor and technology provenance, including explained unknown versions;
 6. typed, bounded, sourced observations and causal status;
 7. no platform proxy coded as a unique person or voter;
@@ -181,6 +194,8 @@ The deterministic v0.3.1 audit checks all 26 generated tables for:
 9. sampling-frame counts and negative-search flags;
 10. election, source, legal, turnout, and future-date status;
 11. complete-result shares and totals;
-12. reproduction of research and coverage views.
+12. reproduction of research and coverage views;
+13. absence of numeric coder-confidence fields and legacy agency-transfer labels;
+14. bounded handling of absence claims, official allegations, and heterogeneous quantities.
 
-The current release passes with zero blocking errors. Six warnings deliberately identify regional frames without systematic negative searches.
+Audit success establishes structural and semantic consistency only. It does not validate source truth, intercoder reliability, sampling representativeness, causal effects, or a control-shift hypothesis.

docs/red-team-v0.4.0.md

@@ -0,0 +1,46 @@
+# Red-team report for v0.4.0
+
+## Verdict
+
+Version 0.4.0 is an exploratory evidence index, not an empirical test of the agency-transfer hypothesis. The red-team found that v0.3.1 was structurally useful but its central pathway field could make inferred control shifts look observed. The release was therefore recoded rather than patched cosmetically.
+
+## Findings and disposition
+
+| Finding | v0.4.0 disposition |
+|---|---|
+| One pathway label mixed capability, attempted transfer, observed effect and preparedness | Replaced by separate `capability_status`, `control_mechanism_status`, `attempted_transfer_status`, `agency_change_status` and `agency_preservation_status` fields. |
+| Researcher inference could be mistaken for an observed transfer | Added `maximum_conclusion`, `observed_evidence`, `researcher_inference`, `falsification_condition` and `comparator`. The release records zero observed agency changes. |
+| Four older agency-transfer claims did not cover all core records | Added one case-specific control-shift hypothesis for each missing retrospective record and recoded all eight as `researcher_hypothesis` / `hypothesis_not_tested`. None has direct supporting evidence links. |
+| Allegations and denials were often linked as underlying facts | Reclassified claim and evidence scope. The evidence layer now distinguishes `assertion_was_made`, `institutional_action`, `measurement`, `underlying_fact`, `review_scope_statement` and `researcher_inference`. |
+| Selected sources were used to imply global absence | Reworded 12 claims as `not_identified_in_declared_review`; removed their 23 purported support links. These statements describe only the declared review set. |
+| Numeric confidence fields implied calibration that did not exist | Removed claim-level `confidence`, `coder_confidence_in_label` and case-level `overall_confidence`. No probability or intercoder-reliability statistic is reported. |
+| A 138,000 figure mixed accounts and transactions across geographies | Removed the incompatible aggregate observation. Its component facts remain claim-level context and are not summed as one comparable metric. |
+| Transparent, authorised or satirical uses could be read as proven manipulation | Retained them as boundary cases. Disclosure alone proves neither harm nor agency preservation. |
+| Official results and platform metrics could leak into causal interpretation | Kept all selected observations descriptive unless a recorded design supports a causal estimate. Official election data remain context only. |
+| The validator could pass stale audits, a forged Parquet manifest or silent type coercion | The v0.4 validator compares stored and live audits, verifies all 26 CSV/Parquet schemas and hashes, rejects invalid coercion, and has isolated mutation tests. |
+| Legacy scripts could overwrite current outputs | The old `audit-data.mjs` entry point is now a read-only compatibility wrapper; the v0.4 auditor retains its historical `audit-v03.mjs` filename, and the obsolete workbook entry point fails before writing. |
+| Public version identity and packaged metadata drifted | v0.4 aligns README, CFF, package, datapackage, Parquet manifest, workbook and ZIP; package members and release artifacts receive explicit SHA-256 manifests. |
+| Dataset-card onboarding and link health were weak | Added clickable entry points, a 26-config map, English-language metadata, a Spanish guide, pinned Datasets/pandas/Polars recipes and corrected known dead links. |
+
+## What remains unresolved
+
+- The claim-coded corpus is purposive, mixed-unit and concentrated in four countries. Six sampling frames still lack a recorded systematic negative search. No prevalence inference is supported.
+- One primary coder produced the coding. A second model-assisted adversarial review can detect inconsistencies but is not an intercoder-reliability study.
+- Source pinpointing improved but remains bounded: 57/124 relations have a verified full-claim locator, 47/124 a partial/premise locator and 20/124 no verified locator because the cited object was inaccessible. Six unsupported edges were removed or recoded as premises, and one mutable platform count was removed for lack of a snapshot. No general source is archived or hashed, and no official-source object is hashed/versioned. Passing QA does not repair those preservation gaps.
+- Most behavioural and electoral effects remain unmeasured. Content, reach, institutional response and election results do not identify voter-level agency change.
+- Screening leads and empirical model studies are comparison and discovery layers, not included incidents.
+
+## Release acceptance criteria
+
+A publishable v0.4.0 build must satisfy all of the following:
+
+1. deterministic source build and stored audit parity;
+2. zero blocking integrity or semantic errors, with all warnings retained;
+3. 26 CSV and 26 Parquet resources with matching rows, schema, source hashes and internal Parquet metadata;
+4. mutation tests that reject a forged manifest, stale audits and invalid boolean coercion without replacing valid Parquet output;
+5. a visually reviewed workbook whose QA sheet reports the same audit result;
+6. a safe ZIP whose member hashes match both its internal manifest and the checked-out files;
+7. external SHA-256 checksums for the ZIP and workbook;
+8. a fixed Hub revision, public downloads and anonymous Data Studio checks for all 26 configurations.
+
+This report records model-assisted red-team work. It is not a peer review, source-by-source adjudication, legal opinion or substitute for independent replication.