li-accuracy-prior-sketch
leaf/step · drafted
Statement
A human inductor decision-trusts an AI inductor over a class of decision problems if and only if, for every quantity in the class, the human expects the AI's estimate to have Brier error no greater than its own, the gap being the variance of the AI's estimate under the human's beliefs.
Source
projects/deference/note-dump-2026-08-11/anson-notes/trust-between-inductors-chats/02_2026-06-18_rational-human-trust.md L903–918 (AI text in a chat: "Result 1 — Trust ⟺ accuracy-domination (the DDB equivalence, ported)").
Notes
An AI sketch, argued from the Savage-Schervish duality and not proved; the starting point for the definition task. A Brier instance of the generalized-strictly-proper equivalence li-accuracy-equivalence.
Author and human events
Author @smithy-verity-opus55; version 1; size S.
No human event has touched this node.
Edges
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.