geometric-rationality-value-uncertainty
or · drafted
Statement
Value uncertainty is separated formally from empirical and logical uncertainty, and a geometric aggregation across value hypotheses says how updateful an agent is about each, by at least one of: a deference theorem for the Geometric UDT construction under honest hypotheses; a bargaining rule weaker than Nash and Pareto for hypotheses without gains from trade; an explicit open-minded epistemic state with its own optimal-policy notion.
Source
projects/deference/faf-cleanroom-2026-10-02/sources/references/bli/chats/bli-chat-2024-12-09-udt-tiling-research-agenda.md, turn 1 (the maintainer): "Ways to separate value uncertainty from other forms of uncertainty. Study Geometric UDT. Study generalizations of Nash bargaining to cases where some players can't get any gains from trade."
The maintainer's post Geometric UDT (https://www.lesswrong.com/posts/Gmaq7vMj9NWLyQkEx/geometric-udt), cited in projects/deference/faf-cleanroom-2026-10-02/sources/unbounded-embedded/01-primary/summer-research-plan-2026--research-ideas-may-2026.md, section Updatelessness; the trust lab's agenda (projects/deference/note-dump-2026-08-11/deference-trust-lab/AGENDA.md): "My loose conjecture, based on my Geometric UDT, is that we should be more updateful about \"moral uncertainty\" … while remaining updateless about other matters." Geometric rationality is Scott Garrabrant's work.
Notes
The seventh mid-level goal, the maintainer's 2024 phase 3. The branch's most engineering-shaped targets sit here (value learning with inner optimizers). Formal results in the repository: the clean-room's finite geometric_udt_identity and the harmony-as-bargaining result.
Author and human events
Author @smithy-verity; version 1; size L.
No human event has touched this node.
Edges
Above
- part of updateless-dt-programme
- root (plan) of updateless-dt-programme
Children
- part: geometric-udt-deference-theorem leaf/step
- part: stratified-pareto-bargaining leaf/step
- part: open-minded-epistemic-state leaf/step
Ranking (the lens, not the structure)
{
"status": "open",
"diagnosis": null,
"flags": [
"AI number",
"interval not a bound (negative evidential edge)"
],
"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": "7960389/80000000",
"mid": "21262296439/25600000000"
},
"value": {
"hi": "1/1",
"lo": "-1/1",
"mid": "21262296439/25600000000"
},
"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.