crux-legitimacy-detection-attainable
cases · drafted
Statement
An agent's beliefs about which modifications are legitimate can be made accurate enough at high capability for the legitimacy-conditioned criterion to be the right target.
Source
The author's thesis in projects/deference/corrigibility-note-dump-2026-10-01/corrigibility-discussion-outline.md: it is crucial that such an agent have quite accurate beliefs about legitimacy, particularly at high capability; the author in projects/deference/corrigibility-note-dump-2026-10-01/position-statement.md, section 2.12: "because the absolute probability of legitimacy is in fact high ... the agent's conditional estimates of legitimacy can be high everywhere"; the negative result on forced trust between inductors (see no-forced-other-trust).
Notes
The crux the whole line turns on; the landed step step-legitimacy-conditioned-trust's accuracy clause is this claim. Pictures: learnable via retractions plus one unlearnable constraint; learnable because legitimacy is in fact high; attainable only with a human-side law of motion; not attainable because learning selects for undetectable manipulation; residual. The fourth row conditions on the no-forced-trust result.
Author and human events
Author @smithy-verity; version 1; size L.
No human event has touched this node.
Edges
Above
- given of no-forced-other-trust
- part of legitimacy-detection
- root (shared-claim) of legitimacy-detection
Children
- part: crux-legitimacy-detection-attainable.partition leaf/partition
- part: pic-detect-constraint-plus-retractions leaf/picture
- part: pic-detect-high-base-rate leaf/picture
- part: pic-detect-needs-oversight-law leaf/picture
- part: pic-detect-selects-undetectable leaf/picture
- part: pic-detect-residual leaf/picture
Ranking (the lens, not the structure)
{
"status": "open",
"diagnosis": null,
"flags": [
"AI number",
"given no-forced-other-trust",
"interval not a bound (negative evidential edge)",
"no belief",
"under attack"
],
"V": {
"V_prior": null,
"display": "2",
"headline": 2,
"ref": 2,
"valid": 2,
"weaker": "both"
},
"belief": null,
"draft": "ai",
"root": {
"status": "open",
"Pr": {
"hi": "1/2",
"lo": "1/2",
"mid": "1/2"
},
"value": null,
"cost": "273/1",
"margin": "no established leaves"
}
}Say something about it
Each verb opens a prefilled issue; a round files what lands as the ledger event it is.