Alignment Workspace

pic-correction-informative

leaf/picture · drafted

Statement

The operator's correction carries information about what the agent values, so on the algebra that contains the correction the agent's utility depends on it and the correction changes the agent's information rather than its objective.

Source

The author's thesis of 2026-09-13 as republished in projects/deference/corrigibility-note-dump-2026-10-01/corrigibility-discussion-outline.md (any change of utility function can be represented as a change in belief in an expanded sigma-algebra); projects/deference/corrigibility-note-dump-2026-10-01/value-change-as-epistemic-update.md, section 4.1: for a Gandhi agent the correction carries no such information (corrections are noise and resisting them is right), for a corrigible agent in the author's sense it does (the public copy's wording).

Notes

The single place where the corrigibility line attacks the landed risk story; its weight is the branch's main evidential link between the thesis and the anti-naturality root.

Author and human events

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

No human event has touched this node.

Edges

Above

Ranking (the lens, not the structure)
{
 "status": "open",
 "diagnosis": null,
 "flags": [
  "AI number"
 ],
 "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