1
Inspect the method and brief
Open the instance and confirm the question, scope, method versions, profile criteria and selection count.Review the proposed brief through its configured decision point. Check exclusions and the handling of insufficient evidence or ties.
2
Inspect research and identities
Review the selected source evidence and extraction results. Resolve proposed merges and disputed associations through the authorised curation route.Open representative relationships. Distinguish reported facts, method-based interpretations and unresolved alternatives.
3
Check the graph computation
Inspect the frozen graph selection, item versions, code version, algorithm and parameters.Compare the returned measures and shortlist with the configured rule. Record disconnected or omitted data. Do not treat a centrality measure as a direct measure of personal influence.
4
Review the proposed profiles
Check each profile against cited sources and the declared framework. Inspect contrary evidence, missing information and uncertain identity.The standard scenario selects five candidates by default; the configured count can differ. Psychometric interpretations require evidence and must remain distinct from measured attributes.
5
Review and publish the assessment
Inspect the assembled report, supporting assets and method references. Follow the configured human review points and request changes where needed.Confirm publication of the exact approved version. Agent agreement cannot replace the required human decision.