Select any node to see what it is, where it came from, and what consumes it.
Nothing in this graph is decoration - every node is a document, an extracted fact with an anchor, a resolved entity, an engine detector, or a finding that examine() actually produced.
There is no risk score here, and no ranking of claimants. The only number attached to a person is corroboration depth - how many independent documents assert the same thing - which is explainable to the person it describes.
Findings are reconciliations, not decisions. Every one carries recommendation: "review" and the innocent explanations that would also fit.