Evidence and uncertainty

The initial dataset is editorial seed knowledge. There are no claimed expert reviews, verified experts, or scientific sources in this release. Definitions and example associations are proposed for inspection, not presented as established research conclusions.

Status vocabulary

• unreviewed: submitted with no completed assessment.
• community_proposed: a community proposal awaiting review.
• editorial_seed: initial editorial material without established supporting evidence.
• expert_reviewed: an explicit claim has a recorded expert assessment; this does not mean unanimous agreement or research support.
• research_supported: supporting primary research is attached to the particular claim.
• systematic_review_supported: the claim is supported by an identified systematic review within its scope.
• disputed: significant objections or contradictory evidence are retained.
• deprecated: preserved for historical identity, discouraged for current use.

These labels are not an ordinal truth scale. An emotion's record status does not validate every relationship, dimension, or contextual statement.

Claims and sources

Create an evidence record with subject, predicate, optional object, a narrow claim, context, population, and supporting/contradictory source IDs. A field-level claim omits object; a relationship claim names an existing target. Validation checks source references and requires support for research-supported labels. Human review still has to assess whether the source actually supports the assertion and whether the study design and population transfer.

Do not infer a numerical probability from weak/moderate/strong. These qualitative strength labels are allowed but seed associations remain unknown. Numerical strength uses the same provenance and scale requirements as other dimensions.

Expert review and disagreement

Reviews target evidence records. Outcomes are agree, qualified agreement, disagree, insufficient evidence, needs revision, or outside expertise. Review comments, expertise domain, date, version, and conflicts are mandatory. Qualifications are retained alongside the original claim. A reviewer is linked to public contributor metadata; credential status and scientific validity are different concepts.

Scientific editors should seek independent review, check credentials with consent, and preserve disputes in Git history and active records. No reviewer counts, credential badges, or votes should be converted into a scientific truth score.

Reading the graph

Projection edges include claim IDs and claim statuses separately. The structural edge defaults to editorial/unreviewed rather than inheriting an entity's research status. Follow supporting and contradictory sources through the Evidence nodes to see which claim they address.