Alignment Workspace

arg-nearest-unblocked

leaf/step · drafted

Statement

A partial do-not-interfere specification added to a strong optimizer's utility leaves other channels open that the optimizer then uses, for an agent whose corrigibility is taught by examples rather than imposed as a constraint.

Source

Arbital, nearest unblocked strategy; projects/deference/corrigibility-note-dump-2026-10-01/corrigibility-discussion-outline.md, item 2, closed as not an anti-naturality argument, with the outline's gloss: the strategy is an artifact of control by constraints rather than teaching by examples, following Armstrong's reading.

Notes

The outline's gloss, in its third-person wording, not a quotation of the author. Its residue is the generalization question generalization-from-corrections.

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": {
  "J": 0,
  "mid": "1/8",
  "range": [
   "0/1",
   "5/8"
  ],
  "width": "1/1"
 },
 "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