Evidence Pilot
Evidence-led research workspace

Evidence work should stay traceable.

Evidence Pilot helps researchers, writers, analysts, and teams, including journalists, investigators, policy writers, reviewers, and institutions, ingest source libraries, dedupe references, extract claims, build evidence matrices, manage redaction boundaries, and keep every research decision tied to the material it came from.

Open Research Workspace Review Accuracy Standards Contact Support
Research Dossier

From library import to evidence matrix.

Evidence Pilot turns source intake into a local research dossier with source-ingestion receipts, DOI/RIS/BibTeX/Zotero-style import QA, provider-bundle receipts for Crossref/OpenAlex/PubMed-style metadata, Zotero RIS export, duplicate detection, extracted claims, persisted screening queue decisions, evidence-matrix rows, confidence bands, redaction/privacy boundaries, reviewer permissions, and audit receipts.

  • Source ingestion and dedupe Normalize DOI-ready, RIS, BibTeX, Crossref-style, OpenAlex-style, PubMed-style, and Zotero-style records while keeping provider verification separate from local metadata QA.
  • Import health meter Preview recognized DOI, RIS, BibTeX, Zotero-style records, duplicates, missing title/author/date rows, and provider-observation conflicts before treating imports as evidence. Conflict rows require human review; provider_verification_claim: false.
  • Claim extraction and evidence matrix Pull reviewable claims from drafts or supplied claim lists, then connect each claim to source matches, source confidence, and publication blockers.
  • Protocol reproducibility readiness Protocol readiness attaches local protocol templates, reproducibility checks, claim-risk flags, subscription-tier readiness, and reviewer checkpoints to every field-specific dossier. Professional advice, independent reproduction, and live provider verification remain separate review steps.
  • Redaction and privacy boundaries Flag confidential, anonymous-source, contact-detail, minor, medical, sealed, or private-person material before export.
  • Institutional packet receipts Package reference-export QA, screening queue inclusion/exclusion reasons, reviewer roles, signoff boundaries, packet-export QA, and local institutional audit receipts for human review.
Workflow Profiles
Researcher literature reviewReference import QA, dedupe receipts, evidence matrices, citation export receipts, and source-confidence summaries.
Journalist investigationSource provenance, claim review queues, redaction/privacy boundaries, and publication blockers.
Policy analysisEvidence strength, study or record context, unsupported claims, memo-ready sources, and institutional receipts.
Field-Specific Evidence Depth

Templates for specialist review paths.

Evidence Pilot now carries local field-profile receipts for specialist customer paths. Each field link opens the same research workspace with a field profile target. These are local workflow profiles and review-packet defaults that guide source templates, claim checks, reviewer roles, export receipts, and safety boundaries without claiming live provider verification, institutional SSO, IRB approval, professional advice, or customer-visible publication.

Clinical specialists Neurologist, Cardiologist, and Oncologist paths cover guidelines, trials, endpoints, adverse events, redaction, and clinician-review boundaries.
Physical sciences Physicist, Astronomer, Climate Scientist, and Chemist paths cover preprints, experiments, observations, models, uncertainty, protocols, and safety context.
Methods and systems Epidemiologist, Computer Scientist, Psychologist, Biomedical Engineer, Social Scientist, and PhD Student paths cover study design, benchmarks, constructs, devices, populations, and dissertation literature-review workflows.
Protocol readiness Field-specific defaults now connect protocol inputs, reproducibility checks, claim-risk flags, citation/source custody, privacy/redaction boundaries, reviewer packet status, export readiness, and tier availability.
Review scope The packet is local review scaffolding for preparing a dossier. Evidence Pilot does not claim professional advice, independent reproduction, live external database verification, provider verification, or publication approval — these remain separate review steps.
NeurologistPubMed-style clinical literature, trial evidence, guideline trail, and care-boundary notes.Open profile
CardiologistGuideline updates, outcome trials, risk-factor evidence, and local citation QA.Open profile
PhysicistPreprint records, method assumptions, instrument notes, and source provenance.Open profile
AstronomerObservation logs, catalog references, telescope or mission data, and data-release notes.Open profile
EpidemiologistCohort definitions, incidence/prevalence sources, confounder notes, and population boundaries.Open profile
OncologistTrial phase, endpoints, treatment guideline trails, therapy evidence, and clinical boundaries.Open profile
Climate scientistDataset version, model scenario, observation record, uncertainty note, and scenario boundary.Open profile
Computer scientistBenchmark provenance, repository/version records, reproducibility checklist, and artifact links.Open profile
PsychologistSample detail, measure validity, replication context, ethics/privacy note, and effect-size trace.Open profile
ChemistReaction or assay conditions, compound identity, measurement source, method custody, and safety boundary.Open profile
Biomedical engineerDevice specifications, validation datasets, standards references, and design-control trace.Open profile
Social scientistSurvey/interview methods, coding frame, population context, IRB/ethics notes, and data-use boundaries.Open profile
PhD studentLiterature review queue, advisor-ready notes, citation export, chapter evidence map, and draft claim trace.Open profile
Research Brief

Start with the decision, then inspect the source.

Capture a clear question, the decision it informs, and the first source context before opening a larger dossier. This local guide helps prepare a reviewer conversation; it does not rate truth, verify a provider record, or clear a claim for publication.

1. Intake guidance

Record before using the source

2. Source decision support

Decision lens waiting for intake.

Choose a source type and record its locator, context, and limitations to see the local review prompts.

Claim Custody

From source import to claim-level risk audit.

Evidence Pilot extends the source dossier with a full claim custody chain. Every extractable claim in the manuscript carries a risk level, source audit trail, and contradiction signal before any source is marked ready for publication. This is local editorial triage — it does not determine truth or replace human review.

  • Claim-risk flags Each extracted claim receives a risk level — high (no source, contradicted), medium (partial sourcing, time-sensitive), or low (supported) — with specific reasons and chapter location. High-risk and medium-risk claims are flagged for human review before the manuscript advances.
  • Source audit trail Every source in the dossier carries an audit trail entry: title, type, year, DOI or URL, access date, verified-by field, and integrity flags (no locator, no year, retraction signals, withdrawal notices). Archive sources are tracked separately.
  • Contradiction checks The contradiction radar compares claims across notes, source excerpts, manuscript passages, and transcript segments. It flags passages where number disagreements, year conflicts, negation mismatches, or counterargument patterns suggest that the same subject gets contradictory treatment in different sections.
  • Reproducibility packet A protocol readiness check for evidence-based writing: methods documentation, search terms, source versions, access dates, retraction scan, and redaction boundaries. Shows which checks passed automatically and which require manual reviewer confirmation.
Claim-risk API POST /api/v1/bc/workbench/claim-audit-chain — returns claim count, high/medium/low risk breakdown, source audit trail, and contradiction signals. Evidence custody packet, not external verification.
Reproducibility API POST /api/v1/bc/workbench/reproducibility-packet — returns a scored checklist of six automated checks plus two manual-review items. Local review scaffolding only; does not claim IRB approval or publication review.
Scope of claim custody All claim custody features are deterministic, local, and owned by BellerCreatives. They do not call external databases, fabricate citations, or claim that a reviewed dossier constitutes professional advice, institutional approval, or publication clearance.
Trust model: Evidence Pilot is the BellerCreatives suite for evidence-led research. BellerCreatives remains the home for accounts, billing, privacy, policy, and support.