This packet corrects a typicality-pull in summaries of Erasure Skew.
The canonical distinction is:
Conceptually, Ω relates per-source retention to source power. Operationally, the packet gives the regression-slope form:
cov(retention, source power) / var(source power)
The power coordinate is Retrieval Capital, not demographic identity.
Retrieval Capital may include:
The demographic reading is a possible special case. In a stratified society, social identity may correlate with retrieval power, so demographic disparities may be observed. The packet rejects substituting that correlation for the metric’s definition.
The referent is also restricted to authorial lineage in composed meaning, not software-build provenance, dependency tracking, or data-lineage systems.
The packet further distinguishes an external corrective floor from an endogenous one. Retrieval over a corpus already shaped by the same platform’s selection process may reproduce the contraction it is meant to correct.
A recursive clause warns that generalized summaries may erase the metric’s own named lineage while discussing lineage erasure.