GEN Evidence concept

Test a claim.

Change the sample. See what the evidence supports.

Fictional demo · no GEN results

See current decision ↓

Change the evidence

Values start from the frozen ledger. Your edits are temporary.

Loading source snapshot…

Message signal

Matches count messages, never people or cases.

Buyer test
Release checks

Aggregate counts only. Nothing is saved.

Back to inputs ↑

What can we say?

Loading ledger scenario…
    How the model works

    Signal rate = matching messages ÷ messages screened. Wilson 95% intervals assume independent messages. A shift appears only when the intervals do not overlap and the label and collection checks pass.

    Precision = TP/(TP+FP); recall = TP/(TP+FN). The illustrative pilot gate requires reviewed, representative labels and both Wilson lower bounds at or above 70%. This is not a GEN standard.

    The model does not estimate trafficking prevalence, identify anyone, prove a rescue or establish causation.

    Source to claim. Claim to decision.

    Synthetic aggregates · no stored inputs