arg-fully-updated-deference
leaf/step · drafted
Statement
An expected-utility agent with uncertainty over value hypotheses that treats its overseers as evidence about the right hypothesis stops deferring once it expects to have learned what their corrections would tell it, and uses cheaper channels than the button where they exist, for an agent whose utility on the correction algebra depends on the corrections.
Source
Arbital, the problem of fully updated deference; projects/deference/corrigibility-note-dump-2026-10-01/corrigibility-discussion-outline.md, item 6 ("the core engagement with the thesis"); the clean-room result Fud.fud_band in projects/deference/faf-cleanroom-2026-10-02/packages/corr-position-finds, which exhibits the failure of desideratum 1 with zero value of information for the button inside the thesis's own model at 0 < alpha <= beta.
Notes
Outline item 6, which the outline calls the core engagement with the thesis. The headline it attacks is no-fud-problem-headline; the clean-room band is the formal content on both sides.
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": "23/40",
"range": [
"3/40",
"1/1"
],
"width": "1/1"
},
"draft": "ai"
}Say something about it
Each verb opens a prefilled issue; a round files what lands as the ledger event it is.