Publish v0.3.0 research documentation

Revised methodology, codebook, ethics protocol and concise Spanish quick-start guide.

docs/codebook.md

@@ -1,132 +1,333 @@
 # Codebook
 
-## Core identifiers
-
-- `case_id`: stable incident identifier, lowercase ASCII with hyphens.
-- `case_family_id`: groups related incidents and election-wide records.
-- `claim_id`, `source_id`, `event_id`, `watch_id`: stable table-specific identifiers.
-- `election_id`: stable identifier for an administered contest or round.
-- `official_source_id`, `turnout_id`, `result_id`: stable identifiers in the official electoral-context layer.
-
-## Official electoral-context layer
+## Recommended reading order
 
-Official election data are descriptive context and outcomes. They never become manipulation claims merely because they are linked to a case.
-
-- `official_elections.election_status`: `completed`, `scheduled`, or `conditional`.
-- `official_elections.legal_result_status`: `annulled`, `certified`, `final`, or `not_yet_available`.
-- `official_elections.linked_case_ids`: pipe-delimited connection to case records; this is a relational link, not a causal assertion.
-- `official_turnout.data_status`: `final`, `pre_election`, `scheduled`, or `not_yet_available`.
-- `electorate_measure`: the authority's denominator or a transparent derivation of it.
-- `participation_measure`: the authority's ballots-cast or participation field definition.
-- `turnout_rate`: decimal fraction from `0` to `1`; blank when a defensible denominator is unavailable.
-- `official_results.valid_vote_share`: decimal fraction from `0` to `1`, calculated against the contest's valid-vote total.
-- `official_results.result_status`: `final`, `certified`, or `annulled`.
-- `official_results.coverage_status`: `complete` when all candidates/options for the scoped result are present; `partial` for an explicitly limited leading-candidate extract.
-- `official_data_sources.data_status`: status of the official data product at the research cutoff, not a claim about election integrity.
-
-Counts are preserved as published, including documented administrative differences that do not arithmetically reconcile. Notes define denominators and explain such discrepancies. A blank turnout rate is preferable to a ratio built from incompatible registration and ballot measures.
-
-## Case status
-
-- `retrospective`: election or incident has concluded.
-- `ongoing`: observed events exist and the election or adjudication is still active.
-- `prospective_monitoring`: monitoring file without an implied incident.
-
-## Event and watchlist status
-
-- `events.event_status`: `observed` means the dated event is sourced as having occurred; `contested` means the event occurred but legality, attribution, deception, or significance remains unresolved; `prospective` is a future date or deadline and never implies occurrence.
-- `watchlist.status=active`: the defensive indicator is operational or has already been activated by an observed signal and remains under review.
-- `watchlist.status=prospective_monitoring`: the rule or risk window is defined, but no activating signal is implied.
-- `watchlist.status=closed`: monitoring for that indicator has ended; later evidence may still revise the record.
-- `monitoring_window`: the inclusive period or named trigger for intensive assessment. It is not necessarily the activation date; an indicator can be `active` before the window starts when an earlier observed signal already activated it.
-
-## Inclusion and case-level assessment
-
-- `anchor_case`: `true` means the row is a direct member or subfile of one of the five user-defined anchor families; `false` marks linked evidence kept as a supporting case rather than a separate anchor family.
-- `inclusion_tier`: `A` means the full claim-level corpus criteria are met: atomic claims, explicit provenance, a dated event record, counterevidence or uncertainty coding, and review. Comparative-candidate rows are not tiered.
-- `overall_confidence`: a reviewer synthesis of confidence in the case description and occurrence as worded. It is not an arithmetic average and does not express confidence in attribution, behavioural effect, or electoral effect; those conclusions remain governed by the atomic claims.
-
-## AI role
-
-- `core_generation`: AI generated or cloned the deceptive or persuasive content.
-- `material_amplification`: algorithmic ranking, bots, or automation materially distributed content.
-- `supporting_tool`: AI reduced cost or assisted production, targeting, translation, or analysis.
-- `detection_response`: AI is used mainly by defenders or institutions.
-- `none_established`: AI is alleged or discussed but not established in the incident.
-- `mixed`: multiple roles apply.
-
-## Generative-AI status
-
-- `confirmed`: supported by provenance, admission, forensic finding, or authoritative record.
-- `probable`: multiple indicators support generation, but provenance is incomplete.
-- `alleged`: a source alleges synthetic generation without sufficient independent verification.
-- `not_established`: no adequate evidence that the relevant content was generative AI.
-- `not_applicable`: the case concerns other algorithmic or coordinated mechanisms.
-
-## Occurrence and effect
-
-- `occurrence_status`: `confirmed`, `partly_confirmed`, `alleged`, or `not_observed`.
-- `reach_status`: `measured`, `platform_reported`, `estimated`, `qualitative`, or `unknown`.
-- `behavioural_effect_status`: `measured`, `indicated`, `not_detected`, or `unknown`.
-- `electoral_effect_status`: `measured`, `institutionally_asserted`, `not_detected`, or `unknown`.
-
-`institutionally_asserted` means an institution relied on or stated an effect in a formal process. It does not convert that assertion into a causal estimate.
-
-## Attribution status
-
-- `admitted`: the actor acknowledged the conduct.
-- `adjudicated`: a competent body reached a final finding relevant to attribution.
-- `officially_attributed`: a public authority attributed the conduct, without implying final adjudication.
-- `platform_attributed`: a platform or threat-intelligence report made the attribution.
-- `credibly_reported`: multiple high-quality independent sources identify the actor.
-- `alleged`: an interested or single source makes the attribution.
-- `unknown`: insufficient evidence.
-- `mixed`: different components have different attribution strength.
-
-## Mechanism classes
-
-Multiple values are separated by `|`:
-
-- `voice_clone`
-- `synthetic_video`
-- `synthetic_image`
-- `synthetic_text`
-- `recommender_amplification`
-- `coordinated_inauthentic_behaviour`
-- `bot_amplification`
-- `paid_influence`
-- `microtargeting`
-- `telecom_robocall`
-- `spoofed_media`
-- `illicit_finance`
-- `voter_suppression`
-- `institutional_response`
-- `liars_dividend`
-
-## Evidence labels
-
-See the dataset card. `confidence` must be a decimal from `0` to `1`. A claim marked `established_evidence` normally requires `confidence >= 0.8`; exceptions require a note.
+| Table | Unit | Use |
+|---|---|---|
+| `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 |
+| `observations` | one quantity | Reach, production, engagement, response, or context |
+| `claims` | atomic assertion | Evidence status and wording |
+| `claim_evidence` | claim–source link | Normalized support/counterevidence |
+| `model_evaluations` | empirical study | Capability or experimental evidence |
+| `official_election_metrics` | one official metric | Legally and semantically explicit electoral context |
 
-## Source roles
-
-- `primary_record`
-- `corroboration`
-- `context`
+## Identifiers and normalized joins
+
+- `case_id`: stable claim-coded record identifier.
+- `case_family_id`: related records sharing an electoral/institutional context.
+- `pathway_id`, `observation_id`, `claim_id`, `source_id`: stable table identifiers.
+- `election_id`: administered contest, round, or explicitly labelled election bundle.
+- `official_source_id`, `metric_id`, `result_id`: official-context identifiers.
+- `catalog_id`: identifier from a core case, screening lead, or empirical study.
+
+Use these normalized tables instead of parsing legacy pipes when possible:
+
+- `claim_evidence`
+- `analytic_record_claims`
+- `case_elections`
+- `election_sources`
+- `case_sources`
+
+## Case catalog
+
+### `record_layer`
+
+- `claim_coded_core`: complete claim/source/pathway record.
+- `screening_register`: lead or comparison awaiting promotion, disaggregation, or exclusion.
+- `empirical_model_study`: audit or experiment; not an incident.
+
+### `record_type`
+
+- `election_wide_case`
+- `observed_network`
+- `observed_incident`
+- `observed_campaign_set`
+- `preparedness_file`
+- `candidate_lead`
+- `model_evaluation_or_experiment`
+
+`record_type` defines the unit. Do not pool different types without an explicit design.
+
+### `incident_count_eligible`
+
+Boolean. It is always `false` for preparedness files, screening leads, and empirical model studies. `true` does not imply that all eligible records are exchangeable or equally harmful.
+
+### `manipulation_assessment`
+
+- `manipulation_confirmed`
+- `manipulation_probable`
+- `mixed_documented_manipulation`
+- `transparent_contested_use`
+- `manipulation_not_established`
+- `no_manipulation_observed`
+- `preparedness_not_incident`
+- `not_an_incident`
+- `not_yet_claim_coded`
+
+AI generation by itself is not a manipulation finding.
+
+## Cases
+
+### `case_status`
+
+- `retrospective`
+- `ongoing`
+- `prospective_monitoring`
+
+### AI role
+
+- `core_generation`: AI creates or clones focal content.
+- `material_amplification`: ranking or automation materially distributes content.
+- `supporting_tool`: AI lowers cost or assists production/translation/targeting.
+- `detection_response`: AI is mainly defensive.
+- `none_established`: AI is alleged or discussed but not established.
+- `mixed`: multiple roles; consult `technology_uses` and pathways.
+
+### `generative_ai_status`
+
+- `confirmed`
+- `probable`
+- `alleged`
+- `not_established`
+- `not_applicable`
+
+The screening register uses more granular provisional codes; normalize before promotion.
+
+### Occurrence and effects
+
+- `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`
+
+`institutionally_asserted` records institutional reliance or wording. It is not a causal estimate.
+
+Legacy `reach_estimate` and `reach_unit` remain for compatibility. Use `observations` for analysis.
+
+## Claims
+
+### `claim_type`
+
+- `occurrence`
+- `mechanism`
+- `ai_role`
+- `reach`
+- `attribution`
+- `intent`
+- `behavioural_effect`
+- `electoral_effect`
+- `institutional_outcome`
+- `response`
+- `legal_status`
+- `counterevidence`
+- `agency_transfer`
+
+### `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.
+- `open_question`: unresolved and not filled by assumption.
+
+### `claim_status`
+
+- `supported`
+- `supported_as_attributed_measurement`
+- `partly_supported`
+- `unresolved`
+
+### `evidence_strength`
+
+- `high`
+- `moderate`
+- `low`
+- `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.
+
+## Claim evidence
+
+### `relation`
+
+- `supports`
 - `counterevidence`
-- `methodology`
-- `ongoing_monitoring`
+- `qualifies` where retained in compatible tooling
+
+### `evidence_scope`
+
+- `underlying_fact`
+- `assertion_was_made`
+- `institutional_action`
+- `measurement`
+- `researcher_inference`
+
+### `directness`
+
+- `firsthand_artifact_or_measurement`
+- `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.
+
+## Sources
+
+### `source_quality`
+
+- `A`: authoritative primary record or peer-reviewed study directly relevant to the claim.
+- `B`: strong independent reporting, civil-society investigation, or transparent technical report.
+- `C`: interested-party/platform self-report or materially incomplete source.
+- `D`: lead only.
+
+### Independence and transparency
+
+- `source_independence`: `independent_or_official | independent_secondary | interested_or_mixed | unknown`
+- `method_transparency`: `sufficiently_described_or_reproducible | not_applicable_official_record | limited_to_published_account | opaque`
+- `access_status`: status of the URL or archive at the cutoff.
+
+Quality, independence, directness, and scope are separate concepts.
+
+## 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.
+
+### Attribution status
+
+- `admitted`
+- `adjudicated`
+- `officially_attributed`
+- `platform_attributed`
+- `credibly_reported`
+- `alleged`
+- `unknown`
+- `mixed`
+
+Provider identification does not imply that the provider designed, approved, funded, or operated a campaign deployment.
+
+## Technology uses
+
+Key fields:
+
+- `provider`
+- `product_or_model`
+- `version`
+- `access_type`
+- `technology_role`
+- `use_status`
+- `provenance_status`
+- `evidence_label`
+
+An unknown version must be explicit in `notes`. Do not infer model provider, open/closed status, API route, prompt, retrieval, or safeguard configuration from output appearance.
+
+## Content items
+
+`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.
+
+## Pathways
+
+### `agency_transfer_status`
+
+- `full_chain_observed`
+- `partial_chain_observed`
+- `mechanism_observed_effect_unknown`
+- `mechanism_inferred`
+- `plausible_hypothesis`
+- `no_transfer_observed`
+- `insufficient_evidence`
+- `not_an_incident`
+
+### `assessment_label`
+
+- `established_evidence`
+- `strong_inference`
+- `plausible_hypothesis`
+- `speculative_scenario`
+- `open_question`
+- `not_applicable`
+
+The pathway fields describe a proposition about control. `observed_outcome` must state what is actually observed and preserve unknown behavioral/electoral effects.
+
+## Observations
+
+### Values
+
+- `value_low`, `value_best`, `value_high`: numeric interval where supported.
+- `value_operator`: exact, approximately, at least, more than, campaign-reported, or sample-specific operator.
+- `unit`: the measured object, not a generic “reach.”
+- `denominator`: population, events, sample, platform counter, or other base.
+- `unique_entity_status`: whether entities are deduplicated, unknown, or not applicable.
+
+### `causal_status`
+
+- `descriptive_only`
+- `association_only`
+- `experimental_causal_estimate`
+- `quasi_experimental_causal_estimate`
+- `causal_estimate`
+- `not_applicable`
+
+A causal estimate requires a suitable outcome claim, comparison/design, uncertainty, and limitations. Official totals alone are never a causal design.
+
+## Model evaluations
+
+Important fields are systems and versions, interface, language, sample unit and size, design, metrics, headline results, causal scope, data availability, and source.
+
+`sample_size` must be interpreted with `sample_unit`. For example, 3 experiments is not 3 participants; 810 party–question pairs is not 810 voters.
+
+## Sampling frame
+
+- `search_status`: `complete | systematic | substantial | partial | ongoing | not_started`
+- `negative_search_recorded`: boolean.
+- `known_gap`: explicit limitation preventing stronger coverage claims.
+
+Asia and Africa aggregate named subregions for coverage counts. A false negative-search flag triggers a warning and blocks prevalence inference.
+
+## Official electoral context
+
+### Election status
+
+- `completed`
+- `scheduled`
+- `conditional`
+
+`contest_type=election_bundle` indicates several separately administered contests sharing a date. It is not one national result.
+
+### Legal/result status
+
+Legal status may be a chronology such as `validated_then_void` or an explicit court-confirmation string. Tally, validation, and process status should not be collapsed.
+
+### Turnout wide table
+
+- `ballots_cast`: retained compatibility field. For Moldova it stores protocol `D` (participants), as originally imported.
+- `turnout_numerator`: numerator used for the displayed/legal rate.
+- `turnout_numerator_measure`: semantic definition and protocol field.
+- `authority_reported_turnout_rate`: rounded authority value.
+- `turnout_rate`: computed rate with full precision.
+
+Prefer `official_election_metrics` for cross-jurisdiction work.
 
-## Delimiters and missingness
+### Official long-form metrics
 
-- Multi-value fields use `|`.
-- Missing values are empty strings, not the literal strings `null`, `NA`, or `unknown` unless `unknown` is a defined category.
-- Human-readable summaries use full sentences; categorical fields use codebook values.
-- Percentages and confidence values are stored as decimal fractions, so `0.525` means 52.5%.
+- `metric_code` and `source_field_code`: normalized and original codes.
+- `value`: numeric value.
+- `numerator_metric`, `denominator_metric`: components of a rate.
+- `authority_reported_rate`, `computed_rate`: published rounding and reproducible calculation.
+- `legal_role`: protocol component, legal numerator/denominator, validity rate, participation rate, or context.
 
-## Comparative-candidate workflow
+### Official results
 
-`candidates` rows are not equivalent to included `cases` rows. `research_status` records the next research action, such as promotion after claim coding, boundary comparison, reserve, or a named second-source requirement. Candidate status vocabulary is intentionally workflow-oriented and may be more granular than the core case enums.
+- `valid_vote_share`: decimal fraction from 0 to 1.
+- `coverage_status`: `complete_national_contest | complete | partial`.
+- `ballot_designating_entity`: entity formally designating a candidate, kept separate from party membership or endorsement.
+- `seats`: official allocated mandates when applicable.
+- `outcome=excluded_ballots_invalidated`: registered competitor with no valid votes after exclusion; not ranked as an ordinary zero-vote competitor.
 
-Candidate `occurrence`, `attribution`, `reach`, `generative_ai`, and related categorical fields use provisional research codes and are not interchangeable with the core enums. Normalize every field before promotion to the included corpus.
+## Missingness and delimiters
 
-`source_urls` contains pipe-delimited exact links. A candidate normally requires at least two sources; a one-source record is permitted only when its `research_status` and `main_caveat` explicitly identify the missing corroboration and keep it in reserve.
+- Missing values are blank.
+- The literal `unknown` is used only when it is a controlled category.
+- Dates use ISO 8601 (`YYYY-MM-DD`).
+- Percentages and shares are decimal fractions; `0.525` means 52.5%.
+- Legacy multi-value fields use `|`; normalized joins are preferred.

docs/ethics-and-release.md

@@ -12,6 +12,12 @@ Public officials, candidates, organisations, and adjudicated or officially named
 
 Publishing monitoring indicators during an election can itself create risk. Indicators are therefore defensive and aggregate: disclosure compliance, official complaints, verified synthetic-media incidents, coordinated-network takedowns, election-authority notices, and changes in platform response. The dataset excludes evasion tactics and targeting vulnerabilities.
 
+Preparedness files are explicitly excluded from incident counts. Model-capability tests and randomized experiments remain in a separate empirical-study layer so that laboratory evidence cannot be misrepresented as deployed manipulation.
+
+## 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.
+
 ## 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.

docs/guia-rapida-es.md

@@ -0,0 +1,71 @@
+# Guía rápida para investigación
+
+## Qué es y qué no es
+
+Esta versión es un índice de evidencia trazable sobre IA, elecciones y transferencia de agencia. Permite comparar mecanismos, actores, sistemas, objetivos, métricas y límites causales.
+
+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.
+
+## Las cinco tablas que conviene abrir primero
+
+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.
+
+En Excel, estas capas aparecen primero y el detalle queda después.
+
+## Cómo leer una fila correctamente
+
+Antes de usar una cifra o conclusión, pregunte:
+
+- ¿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?
+
+## Regla central de inferencia
+
+No transforme esta secuencia:
+
+`contenido observado → alcance de plataforma → alarma institucional → resultado electoral`
+
+en esta afirmación:
+
+`la IA cambió votos`.
+
+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.
+
+## Estudios de modelos
+
+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.
+
+## Datos electorales oficiales
+
+Use `official_election_metrics.csv` para comparaciones semánticas. La tabla ancha `official_turnout.csv` se conserva por compatibilidad.
+
+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.
+
+Los resultados oficiales describen votos, escaños y estatus legal. No identifican exposición ni causalidad.
+
+## Qué puede afirmar con seguridad
+
+Puede describir:
+
+- 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.
+
+No debe afirmar prevalencia global, impacto agregado ni votos cambiados sin otra muestra y un diseño específico.
+
+## Estado de calidad
+
+La auditoría v0.3 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.

docs/methodology.md

@@ -2,113 +2,177 @@
 
 ## Research question
 
-When election-related influence systems move citizens from persuasion toward agency transfer, which actor gains control, through which technical and institutional pathway, and what evidence exists of democratic harm?
+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?
 
-## Unit of analysis
+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 core unit is an incident or tightly bounded campaign. Related incidents can share a `case_family_id`. A whole election is used as the unit only when the institutional record itself concerns election-wide conditions, as in the annulment and repeat of Romania's presidential election.
+## Research layers and units
+
+The release has three non-interchangeable layers.
+
+1. **Claim-coded core.** A bounded record with atomic claims, sources, actors, technology, pathway assessment, and case–election links. `record_type` identifies whether the unit is an election-wide case, observed network, observed incident, campaign set, or preparedness file.
+2. **Screening register.** Comparative leads, boundary cases, null samples, reserves, and records needing disaggregation or corroboration. These are not included incidents.
+3. **Empirical model studies.** Model audits and experiments. They measure capability or experimental outcomes under a specified design, not real-world incident prevalence.
+
+Preparedness files and empirical studies have `incident_count_eligible=false`. The remaining core records still require stratification by `record_type`; they are not automatically exchangeable observations.
 
 ## Inclusion criteria
 
-A case is included when all of the following apply:
+A core record must:
+
+1. be materially connected to an election, referendum, voter participation, candidate choice, or electoral trust;
+2. involve a documented AI, synthetic-media, recommender, automated-distribution, coordinated-network, or AI-enabled response mechanism;
+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.
+
+Transparent or authorised AI may be retained as a comparison when it tests the boundary between representation, persuasion, manipulation, and agency preservation.
 
-1. It is materially connected to an election, referendum, voter participation, candidate choice, or confidence in electoral administration.
-2. It involves at least one mechanism in scope: synthetic media, generative AI, recommender or ranking systems, automated or coordinated amplification, bots, AI-assisted targeting, AI-enabled impersonation, or AI-supported detection and response.
-3. It has at least two credible independent sources, or one authoritative primary record plus sufficient corroboration.
-4. Its factual occurrence can be separated from interpretations of intent, attribution, reach, and effect.
+## Screening and coverage
 
-Prospective files may be included before an election only when they contain official rules, documented preparations, already observed incidents, and observable monitoring indicators. Scenarios alone are stored only in `watchlist` and labelled `speculative_scenario`.
+Selection is purposive. Anchor cases were chosen for legal, institutional, technical, or comparative value. Candidate leads were then added to broaden geography, mechanism, authorization status, model access, channel, and positive/null comparison.
 
-The `candidates` config is a research queue, not part of the included case corpus. It preserves comparative leads, boundary cases, reserves, exclusions, source links, and the principal reason a record has not yet been promoted into full claim-level coding.
+`sampling_frame.csv` records:
 
-## Exclusion criteria
+- region and selection basis;
+- search status and languages;
+- source channels;
+- included and candidate counts;
+- whether a negative/null search was recorded;
+- known gaps and last search date.
 
-- Generic political misinformation with no material automation, AI, algorithmic-distribution, or synthetic-media component.
-- Demonstrations or hypothetical content never deployed in an election context.
-- Claims supported only by anonymous reposts or unverifiable screenshots.
-- A political use of AI that is transparent, authorised, and lacks a plausible manipulation mechanism, unless retained as a labelled comparison or boundary case.
-- Unsupported claims that exposure, impressions, or virality changed votes.
+The release may be described as comprehensive within its declared evidence files. It must not be described as globally representative. A regional prevalence statement requires a defined universe, systematic search, recorded intensity, positive and negative decisions, and comparable denominators.
 
 ## Evidence workflow
 
-1. Build a source inventory, prioritising courts, election authorities, regulators, law enforcement, international election-observation missions, platform transparency reports, and peer-reviewed work.
-2. Atomise source statements into claims.
-3. Code each claim by type and epistemic status.
-4. Add independent corroboration and counterevidence.
-5. Separate the content creator, sponsor, distributor, platform, attributed state actor, and responding institution.
-6. Record the maximum conclusion warranted by the evidence, not the most dramatic interpretation.
-7. Re-verify links and ongoing-case status before each release.
+1. Inventory official, legal, regulatory, platform, technical, peer-reviewed, civil-society, and independent reporting sources.
+2. Atomize source statements into claims.
+3. Distinguish the fact itself from the fact that an actor or institution made an assertion.
+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.
 
-## Official electoral-data protocol
+`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 status
+
+`evidence_label` preserves the reasoning form:
+
+- `established_evidence`
+- `established_as_campaign_report`
+- `strong_inference`
+- `open_question`
 
-The official electoral-context layer is acquired directly from the institution that administers, totalizes, certifies, adjudicates, or officially schedules the relevant contest. It is kept separate from the influence-claims layer and linked only through stable election and case identifiers.
+`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.
 
-For every retained official measure:
+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.
 
-1. Store the administering authority, exact URL, format, granularity, data status, retrieval date, and reuse note.
-2. Prefer signed or explicitly final protocols over live dashboards and preliminary snapshots.
-3. Preserve the authority's raw denominator definition. Do not create turnout rates when registration and ballot-eligibility measures are incompatible.
-4. Distinguish complete result groups from partial leading-candidate extracts.
-5. Preserve legal status independently from observed counts; an annulled contest can retain a final count before annulment.
-6. Do not create result rows for future elections. Pre-election electorate and candidate files retain `pre_election` status.
-7. Exclude personal identifiers, addresses, contact details, and exact polling-place coordinates from the public release.
+## Agency-transfer pathway
 
-Institution-specific reconciliation rules are documented in row notes. Romania's final national protocols include postal votes omitted from precinct CSV exports. Moldova workbooks mix aggregation levels and therefore require explicit national-summary or station-row selection. New Hampshire party-registration and primary-ballot measures are not forced into an invalid turnout denominator. Brazil's national and state partitions are mutually exclusive ingestion strategies and must never be concatenated.
+The central sequence is:
 
-Official results establish the outcome reported by the authority. They do not measure exposure to manipulation or identify a causal effect. Cross-election change is not a natural experiment, and precinct or district correlations cannot support person-level inference.
+`capability → controller → vector → target → affected agency → control shift → power recipient → harm pathway → observed outcome`
 
-## Source quality
+`pathways.csv` records:
 
-`source_quality` is an ordinal assessment:
+- affected agency or capacity;
+- who originally held relevant control and where it moved;
+- technical/social mechanism;
+- opacity, personalization, contestability, and duration;
+- recipient of power;
+- harm pathway and observed outcome;
+- agency-transfer status and epistemic assessment;
+- supporting claim identifiers and reviewer note.
 
-- `A`: authoritative primary record or peer-reviewed study with directly relevant evidence.
-- `B`: high-quality independent reporting, established civil-society investigation, or transparent technical report.
-- `C`: interested-party or platform self-report, credible but incomplete secondary reporting, or source with material uncertainty.
-- `D`: lead only; not sufficient to support a published claim without corroboration.
+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.
 
-Source quality is not the same as truth. A regulator's allegation is an authoritative record that the allegation was made, but not necessarily proof that the alleged conduct occurred.
+## Quantitative observations
+
+Every quantity is a separate row in `observations.csv`. Required elements are:
+
+- metric family and name;
+- low, best, high, and operator where available;
+- unit and denominator;
+- unique-entity status;
+- platform, geography, and time window;
+- 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.
+
+## Empirical model studies
+
+`model_evaluations.csv` keeps model tests outside the incident corpus. A row states systems/versions, interface, languages, sample unit and size, design, metrics, result summary, data availability, and causal scope.
+
+Randomized experiments can support causal claims only for the assigned interaction and measured outcome in the experimental population. They do not establish covert deployment, durable effects, actual ballots, or aggregate election outcomes. Point-in-time model audits cannot be generalized to later versions, all prompts, consumer interfaces, or all user populations.
+
+## Official electoral-data protocol
 
-## Claim model
+Official election data are descriptive context and outcomes. They are linked to cases but never treated as evidence that influence changed votes.
 
-Claims are typed as `occurrence`, `mechanism`, `ai_role`, `reach`, `attribution`, `intent`, `behavioural_effect`, `electoral_effect`, `institutional_outcome`, `response`, `legal_status`, `counterevidence`, or `agency_transfer`.
+For each source, the manifest records the institution, landing and download URLs, status, format, granularity, access method, retrieval time, snapshot cutoff, reuse terms, and optional content hash/bytes/schema version.
 
-The wording controls the confidence score. For example:
+For each measure:
 
-- “The FCC issued a $6 million forfeiture order” can be established by the order.
-- “The robocall changed votes” remains unknown without behavioural evidence.
-- “Moldovan authorities alleged that a foreign-funded network paid voters” is established as an official allegation; the underlying conduct and attribution require their own claims.
+1. retain the original field code and label when available;
+2. separate numerator, denominator, authority-reported rate, and computed rate;
+3. distinguish legal role from descriptive role;
+4. preserve legal validation or annulment chronology separately from the tally;
+5. never create future result rows;
+6. never force a turnout rate from incompatible denominator concepts;
+7. reconcile complete result groups to the official valid-vote total.
 
-## Agency-transfer model
+### Moldova denominator rule
 
-Each case is assessed along this path:
+Moldovan protocols distinguish:
 
-`capability → controlling actor → influence vector → targeted population → affected state or capacity → agency transferred → power concentrated → democratic harm → mitigation`
+- `A`: electors on basic lists;
+- `B`: electors on supplementary lists;
+- `C`: electors who received ballots;
+- `D`: electors who participated;
+- `E`: `C − D`;
+- `F`: invalid ballots;
+- `H`: valid votes;
+- `I/J`: additional ballot-accounting fields where available.
 
-The affected state or capacity may be belief, preference, attention, trust, participation, autonomous deliberation, agenda-setting, or institutional discretion.
+For the 2024 presidential rounds and referendum, the Constitutional Court used `C/(A+B)` for the legal validity threshold. The prior wide table used `D/(A+B)`, which is a descriptive participation rate but not the legal rate. Both are now retained. For the 2025 parliamentary election, the authority-reported rate uses `D/(A+B)`.
 
-## Ongoing cases
+### Result completeness
 
-Brazil and the United States in 2026 are versioned monitoring files. Dated events use:
+Romania 2024 first round, Romania 2025 first round, Moldova 2024 first round, Moldova 2024 second round, the Moldovan referendum, and Moldova 2025 parliament are complete at national contest level. New Hampshire remains `partial`; its source remains authoritative for the full field.
 
-- `observed`: an event has occurred and is sourced.
-- `contested`: the event occurred but legality, attribution, deception, or significance is unresolved.
-- `prospective`: a future election, deadline, or meeting is recorded without implying occurrence.
+## Causal safeguards
 
-Watchlist indicators use a separate status vocabulary: `active` when an indicator is operational or activated by an observed signal, `prospective_monitoring` when only a rule/window exists, and `closed` after monitoring ends. `monitoring_window` is the intensive assessment interval or trigger and need not equal the initial activation date.
+The release blocks these common errors:
 
-Every prospective case file includes an `as_of_date`; event rows use `event_date`, and watchlist rows use `last_checked` plus `monitoring_window`. No forecast is silently converted into a fact.
+- institutional alarm or annulment does not prove votes changed;
+- a close result is not evidence of influence;
+- views, engagement, or followers do not prove exposure or persuasion;
+- cross-election change is not a counterfactual;
+- aggregate geography does not identify individual behavior;
+- 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.
 
 ## Quality assurance
 
-- Referential integrity across case, claim, source, event, and watchlist identifiers.
-- Every direct claim-to-source use has a matching case-source link whose declared support types include that claim type.
-- No claim without at least one source or an explicit `open_question` status.
-- No electoral-effect claim inferred solely from reach.
-- No foreign attribution based solely on political statements.
-- Platform self-reports labelled as such.
-- Dates stored as ISO 8601.
-- URLs preserved as plain provenance fields; copyrighted source text is not reproduced.
-- Unique official election, turnout, result, and source identifiers with foreign-key checks.
-- Future elections cannot contain final turnout or result rows.
-- Complete national result groups must sum to the official valid-vote total and their valid-vote shares must sum to one within tolerance.
-- Official source status, result coverage, legal status, and turnout denominator definitions are explicit.
+The deterministic v0.3 audit checks:
+
+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;
+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;
+8. model-study sample/design/scope;
+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.
+
+The current release passes with zero blocking errors. Six warnings deliberately identify regional frames without systematic negative searches.