Source Reconciliation Under Asymmetric Narrative Propagation

We analyse the asymmetric propagation of headline-level narratives across primary, secondary, and tertiary sources. Reconciliation is treated as the construction of a consistency graph over claims, with edge weights determined by publisher reliability, timestamp proximity, and structural independence of derivation.

Reconciled posterior μ = softmax(W · c) where c is the vector of claim indicators and W encodes pairwise source weights. Independence is approximated through citation-graph analysis.

Tertiary aggregation collapses independent signal into apparent consensus. Without explicit citation graph analysis, source counts are misleading; one claim cited five times by derivative outlets carries the evidential weight of one source.

Source confidence is reported separately from signal confidence on every memo. A high source count without independence is treated as low-evidence.

Citation-graph reconstruction is imperfect; we under-count rather than over-count independence.