Alignment Workspace

joint-process-model

or · drafted

Statement

A model of the joint agent-and-oversight process exists with a training map from the overseers' data and loss to the next model, a measure of what the next model retains, and a law of motion for the overseers' evaluation quality, in which the trajectory conditions are stated.

Source

The author, projects/deference/corrigibility-note-dump-2026-10-01/position-statement-v2.md, section 1.3: "Of course this isn't convincing as-stated; we need to provide some sort of model in which this can be formally established in order for it to be convincing."; run 3's summary in projects/deference/faf-cleanroom-2026-10-02/sources/corrigibility/workflow-2026-09-14b/plan/synthesis.md: "Both name a model of the training map T with a retention measure as the most valuable missing work"; the clean-room packages corr-joint-process, corr-trajectory and corr-landscape under projects/deference/faf-cleanroom-2026-10-02/packages/.

Notes

The plan root of the dynamics frame, an or over the components the record names as missing: any realised component advances the model, so the or reads as 'any route suffices' for belief and the gestalt is the number for the whole. The routes are actioned questions and task leaves; the toy certificates that already exist (hazard budget, basin, Ville bound) are shared claims of the frame.

Author and human events

Author @smithy-verity; version 1; size L.

No human event has touched this node.

Edges

Above

Children

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": null,
 "draft": "ai",
 "root": {
  "status": "open",
  "Pr": {
   "hi": "1/1",
   "lo": "137923/320000",
   "mid": "506592521/512000000"
  },
  "value": {
   "hi": "1/1",
   "lo": "-1/1",
   "mid": "506592521/512000000"
  },
  "cost": "inf",
  "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