arg-newer-formalisms-checkpoint
leaf/question · drafted
Statement
The corrigibility notion that survives the walk matches at least one of shutdown instructability (Carey and Everitt 2023), the corrigibility transformation (Hudson 2025), CAST (Harms) and the broad-safety notion of deployed systems.
Source
projects/deference/corrigibility-note-dump-2026-10-01/corrigibility-discussion-outline.md, item 9; projects/deference/corrigibility-note-dump-2026-10-01/corrigibility-reading-list-v1.md, the two organising facts (at least five non-equivalent notions circulate; the shutdown button is the standard toy case rather than the whole problem); clean-room results bearing on it: Carey and Everitt's Theorem 10 vacuous and Theorem 14's converse false under the paper's fixed utility domain (projects/deference/faf-cleanroom-2026-10-02/packages/corr-scim-cid); Hudson's Theorem 3.1 corrigibility half definitional under one reading and the transformation inert on a physical channel (projects/deference/faf-cleanroom-2026-10-02/packages/lit-mdp-corrigible, findings F-1 and F-17).
Notes
Outline item 9. Its edge from five-notions-of-corrigibility: with five non-equivalent notions in circulation, a match to at least one is more likely and means less.
Author and human events
Author @smithy-verity; version 1; size M.
No human event has touched this node.
Edges
Above
Ranking (the lens, not the structure)
{
"status": "open",
"diagnosis": null,
"flags": [
"AI number",
"given five-notions-of-corrigibility"
],
"V": {
"V_prior": null,
"display": "2",
"headline": 2,
"ref": 2,
"valid": 2,
"weaker": "both"
},
"belief": {
"J": 0,
"mid": "1/2",
"range": [
"0/1",
"1/1"
],
"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.