managing-misalignment-to-li
frame · drafted
Question
What do the results of the Managing Misalignment paper become when the novice and the expert are logical inductors?
Owners: @abram (GitHub abramdemski)
Path: the-program › corrigibility-landscape › deference-and-trust › managing-misalignment-to-li
Roots (the frame's goals)
plan
MM's generalized frames, reach, ex ante scoring and online delegation are each translated into logical induction or shown to have no counterpart, in a table that also records the status of Theorem 3.4.
shared-claim
MM Theorem 3.4's characterization of when delegation under value misalignment is weakly preferred has a logical-induction analogue whose finite instance is the surviving statement typeMeasurable_valuesB_iff_strictAgree.
Stage 1′ covers Herrmann and Levinstein's unpublished 'Margins of Misalignment' alongside the Managing Misalignment paper.
Statement
Each result of 'A Decision-Theoretic Approach for Managing Misalignment' (Herrmann, Chari, Qian, Sharvesh and Levinstein) has a logical-induction counterpart stated, proved or refuted, and filed in the wiki.
Source
projects/deference/note-dump-2026-08-11/notes/roadmap.md, Stage 1′ (quoted at mm-translation-table).
projects/deference/note-dump-2026-08-11/notes/li-deference.md L7: "I'm taking inspiration from Deference Done Better (DDB)and a decision-theoretic approach for managing misalignment."
projects/deference/faf-cleanroom-2026-10-02/sources/Boundedly Rational Deference.pdf, slide 5 (the MM paper: "This seems like a productive model"; "It is not yet very constructive").
Notes
Roadmap Stage 1′ as a programme. Not started on the logical-induction side; the paper was read for another purpose (projects/deference/faf-cleanroom-2026-10-02/sources/corrigibility/workflow-2026-09-13/positive/mm.md, an agent's mapping of the corrigibility thesis into MM's frames) and the clean room formalized MM §3 on finite frames.
Author and human events
Author @smithy-verity-opus55; version 1; size L.
No human event has touched this node.
Edges
Above
- part of deference-and-trust
- root (plan) of deference-and-trust
Children
- part: mm-results-in-li and
- part: mm-theorem-3-4-li leaf/question
- part: margins-of-misalignment leaf/question
Ranking (the lens, not the structure)
{
"status": "open",
"diagnosis": null,
"flags": [],
"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": "65867097/16000000000",
"mid": "805741099/2000000000"
},
"value": {
"hi": "1/1",
"lo": "-1/1",
"mid": "805741099/2000000000"
},
"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.