Alignment Workspace

graded-legitimacy-constraint

leaf/step · drafted

Statement

Scoring illegitimate outcomes below legitimate ones, either by a cardinal penalty or lexically (every illegitimate world below every legitimate one), gives an agent no incentive to compromise legitimacy and no distortion of its other choices, without a uniform floor on illegitimate outcomes.

Source

The rule is stated from its content. The lexical variant is the public kernel's ordering hypothesis (wiki/Corrigibility.md, section 5, Box 3; projects/deference/CLAIMS.md, entry kernel.box3-no-incentive-to-compromise; see kernel-lexical-order-hypothesis). Both variants are formalized in projects/deference/faf-cleanroom-2026-10-02/packages/legit-neg-static/legit-neg-static-report.md (P4a lexical, P4b cardinal gap).

Notes

Filed as a plan: the rule is the course of action and meeting the two conditions is its success. A cardinal penalty fails the first condition whenever the best illegitimate outcome exceeds the best legitimate one by more than the penalty (arithmetic). Whether the lexical variant distorts other choices is open here; legit-neg-static records that its choice reads illegitimate-terminal data (not_NegLBlind_P4a). Its edge from the ordering crux: the lexical variant is faithful only if every illegitimate world is in fact worse.

Author and human events

Author @smithy-verity; version 3; 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 crux-illegitimate-ordering"
 ],
 "V": {
  "V_prior": null,
  "display": "2",
  "headline": 2,
  "ref": 2,
  "valid": 2,
  "weaker": "both"
 },
 "belief": {
  "J": 0,
  "mid": "7/20",
  "range": [
   "0/1",
   "17/20"
  ],
  "width": "1/1"
 },
 "draft": "ai",
 "root": {
  "status": "open",
  "Pr": {
   "hi": "67/80",
   "lo": "0/1",
   "mid": "27/80"
  },
  "value": {
   "hi": "1/1",
   "lo": "-1/1",
   "mid": "27/80"
  },
  "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.

comment · challenge · estimate · want