Alignment Workspace

correction-changes-objective

cases · drafted

Statement

An operator's correction of the agent changes the agent's objective rather than its information.

Source

Reconstructed from Soares et al. 2015, section 1 and the shutdown model of section 2, where a correction replaces the utility function U by U*; the paper does not state it in this form.

Notes

The assumption of the anti-naturality root, now a table: the author's thesis lands here as the picture on which the correction is informative, and the i-flat reading on which it is noise is the picture under which the Soares et al. sentence holds. The Statement is the landed one, unchanged; the table was filed by the content stage of 2026-10-05 from projects/deference/corrigibility-note-dump-2026-10-01/value-change-as-epistemic-update.md, section 4.1, read by the filing session as naming this assumption as the hidden one. The landed range on the leaf is marked stale by the protocol's rule for a leaf that became a table.

Author and human events

Author @smithy-verity; version 1; size M.

No human event has touched this node.

Edges

Above

Children

Ranking (the lens, not the structure)
{
 "status": "open",
 "diagnosis": null,
 "flags": [
  "AI number",
  "attribution unvetted",
  "no belief",
  "stale: leaf became a table",
  "under attack"
 ],
 "V": {
  "V_prior": null,
  "display": "2",
  "headline": 2,
  "ref": 2,
  "valid": 2,
  "weaker": "both"
 },
 "belief": null,
 "draft": "ai"
}

Say something about it

Each verb opens a prefilled issue; a round files what lands as the ledger event it is.

comment · challenge · estimate · want