Evidrai Methodology Transparency Statement

1. Purpose

This statement explains the principles behind Evidrai's credibility and verification methodology.

Evidrai is designed to help users understand information more clearly by examining available evidence, source quality, context, uncertainty, and contradiction.

2. Claim-First Assessment

Evidrai assesses the specific claim being reviewed. A source may be reputable in general but weak for a particular claim if it does not address the relevant facts directly.

The platform aims to distinguish between evidence, context, commentary, allegation, and repetition.

3. Source Classification

Evidrai may classify sources into broad categories to establish initial source-quality priors before assessing claim fit. These categories are used as starting points, not final judgments.

• Scientific sources: research, medical, standards, academic, peer-reviewed, or preprint sources where relevant
• Government sources: official public agencies, statistics, public services, and intergovernmental sources
• Legal sources: statutes, courts, legislation, parliamentary material, and official legal records
• Primary sources: direct original evidence such as filings, datasets, transcripts, recordings, source documents, or official records
• News sources: reporting outlets and wire services, treated cautiously where coverage may share source chains or editorial framing
• Secondary sources: reputable analysis, expert synthesis, or institutional summaries that are not original evidence
• Contextual sources: background material that may explain a topic but does not by itself verify or falsify the claim

4. Source Quality Factors

Individual source scores may consider multiple factors. These factors are combined to help rank and explain sources, but they do not alone determine the final verdict.

• Authority: whether the source is likely to be authoritative for this type of claim
• Relevance: whether the source addresses the specific claim being checked
• Directness: how close the source is to original evidence rather than interpretation or commentary
• Independence: whether the source adds a distinct evidence chain rather than repeating the same article, briefing, or wire story
• Recency: whether the source is temporally appropriate for the claim
• Bias risk: whether the source has incentives, framing, or conflicts that may require additional corroboration

5. Corroboration, Contradiction, and Independence

Evidrai does not treat volume as proof. Multiple sources repeating the same original claim, press release, wire report, or narrative cluster may count as amplification rather than independent confirmation.

Contradictory evidence is considered directly. A credible contradiction can materially reduce confidence in a claim even when several other sources appear supportive.

6. Confidence Indicators

Evidrai may present confidence levels to indicate how strongly available evidence supports an assessment. Confidence is based on the quality, directness, independence, consistency, and completeness of the evidence set.

Confidence should not be interpreted as certainty. It reflects the available evidence at the time of review.

• Availability of primary or direct evidence
• Quality and relevance of reviewed sources
• Agreement or conflict between evidence chains
• Recency or temporal fit
• Ambiguity, missing context, or uncertainty
• Whether evidence appears independently corroborated or merely repeated

7. Source Transparency

Where possible, Evidrai provides supporting sources, source scores, evidence summaries, and explanations so users can review the basis of an assessment.

Public-facing explanations may simplify internal scoring details, but the underlying methodology is designed to be auditable and versioned.

8. Limits of Methodology

Evidrai may not always have access to all relevant information. Assessments may be incomplete where evidence is missing, conflicting, private, newly emerging, paywalled, jurisdiction-specific, or difficult to verify.

Source categories and scoring factors are decision-support signals, not guarantees of truth. Human judgment remains important, especially for high-stakes decisions.

9. Updates and Governance

Assessments may change as new evidence becomes available.

Evidrai is committed to continuous improvement of its source classification, scoring policy, model behaviour, auditability, and transparency practices. Changes to scoring policy are managed as methodology changes and should be reviewable over time.

10. Contact

Questions about methodology may be directed to trust@evidrai.com.