rs-steering-toward-illegitimacy
leaf/step · drafted
Statement
An agent that scores illegitimate worlds by graded values steers toward illegitimate worlds where the best achievable outcome there exceeds the best outcome it expects to be achievable in legitimate worlds.
Source
The capture toys of cluster C (projects/deference/faf-cleanroom-2026-10-02/packages/legit-neg-pricing/legit-neg-pricing-report.md, targets 16, 17, 19 and 20: steering and capture): the cardinal variant fails exactly when the illegitimate best exceeds the legitimate best plus the penalty, and is not a failure where that estimate is correct.
Notes
Threatens: the graded alternative. Confirmed in the pricing package's model. A risk story the corrigibility work itself produced: it shows a probability and a status and enters the comparison through the plans whose tables cite it. No outcome class is tagged (tagging a failure class is a human judgment under the protocol's rule on classes), so its failure mass goes to the default trajectory until a human tags it.
A root of corrigibility-landscape, the sub-frame whose plans it threatens; the top frame keeps only stories at the level of its own question.
Author and human events
Author @smithy-verity; version 1; size M.
No human event has touched this node.
Edges
Above
- part of corrigibility-landscape
- given of crux-illegitimate-ordering
- root (risk-story) of corrigibility-landscape
- given crux-illegitimate-ordering
Ranking (the lens, not the structure)
{
"status": "open",
"diagnosis": null,
"flags": [
"AI number",
"given crux-illegitimate-ordering",
"interval not a bound (negative evidential edge)"
],
"V": {
"V_prior": null,
"display": "2",
"headline": 2,
"ref": 2,
"valid": 2,
"weaker": "both"
},
"belief": {
"J": 0,
"mid": "5/8",
"range": [
"1/8",
"1/1"
],
"width": "1/1"
},
"draft": "ai",
"root": {
"status": "open",
"Pr": {
"hi": "73/80",
"lo": "9/80",
"mid": "21/40"
},
"value": null,
"cost": "85/1",
"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.