Alignment Workspace

updateful-about-values

leaf/question · drafted

Statement

An agent's optimal treatment of uncertainty is updateful about moral uncertainty, an aligned AI mostly or completely updateful about human values, and updateless about other matters.

Source

projects/deference/faf-cleanroom-2026-10-02/sources/unbounded-embedded/01-primary/summer-research-plan-2026--research-ideas-may-2026.md, section Updatelessness: "My loose conjecture, based on my Geometric UDT, is that we should be more updateful about "moral uncertainty" (EG, an aligned AI should be mostly/completely updateful about human values), while remaining updateless about other matters."

Notes

A loose conjecture, as the source states it. Evidence against its simplest reading: projects/deference/faf-cleanroom-2026-10-02/packages/udt-endorse-policy's values_coin (the mugging read as a values fact, where the values-updateful policy loses) and its finding that 'values are separable' is not a property of subject matter; with decisionFlat_iff the finite results say updatefulness about values is optimal exactly when value cells are uncoupled in the utility, a hypothesis about the problem. A third pass, the deference lab's critique (projects/deference/note-dump-2026-08-11/deference-trust-lab/report/CRITIQUE.md), rates the factual half near-trivial and the asymmetry a conjecture.

Author and human events

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

No human event has touched this node.

Edges

Above

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": {
  "J": 0,
  "mid": "7/20",
  "range": [
   "0/1",
   "17/20"
  ],
  "width": "1/1"
 },
 "draft": "ai",
 "root": {
  "status": "open",
  "Pr": {
   "hi": "17/20",
   "lo": "0/1",
   "mid": "7/20"
  },
  "value": null,
  "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