What it answers
This is the record’s reply of 2026-10-10 to CLAIM-258, which says that Blackwell’s order needs one parameter and that the line’s central cases lack one: the situation changes under compensation (CLAIM-091), and where it is common, a translator with context knowledge makes a rendering that is not a garbling. The reply is a reconciliation. The objection runs together two decision targets. For one of them its premise fails and the order applies; for the other it is right. The distinction is a proposal: whether receivers’ verdicts track it, and which target the line’s central cases have, are open, and the defeat condition says so.
The claim
Two targets. A reader of a rendering can be asked two kinds of question.
- Report. What was the source’s speaker doing, in the source’s situation? The target reader, like the source reader, decides about the source’s situation z. The rendering is evidence about z.
- Re-performance. Does this rendering do, here, in the target’s situation z′, what the source did there? The target reader decides about z′, and the comparison needs to know which z′ corresponds to which z.
In report fidelity the order is defined. Both experiments, P(U | Z) for the source and P(V | Z) for the rendering, are indexed by the source’s situation, so they share the parameter THEORY-156 requires (“an n-tuple of probability measures on a common space, one per state”). A translator who knows the context makes a rendering that depends on Z other than through U, so V is not a garbling of U. The order still compares the two. The direction the manuscript uses (garbling implies no worse) is silent, and the order’s verdict is often incomparability.
CLAIM-323’s worked case, read as a report case. Let Z be uniform on {1, 2, 3}. The source tells its reader whether Z = 1 (U = 1 if Z = 1, else U = 0). The rendering tells its reader whether Z = 3 (V = 1 if Z = 3, else V = 0). Take 0–1 loss.
- “Is Z = 3?” From V the risk is 0, since V answers it. From U: when U = 1 (probability 1/3), Z = 1 and the answer is certainly no, with no error; when U = 0 (probability 2/3), Z is 2 or 3 with probability 1/2 each, so any answer is wrong with probability 1/2. The risk is (1/3)(0) + (2/3)(1/2) = 1/3.
- “Is Z = 1?” By the same steps with the roles exchanged, the risk is 0 from U and 1/3 from V.
So each does better on one problem, neither dominates, and they are incomparable in Blackwell’s order. That is a precise verdict, and it is the verdict CLAIM-128 expects: profiles that trade off are incomparable until a task fixes weights. The size of each direction’s loss is what CLAIM-259’s directed decision term measures: the rendering loses 1/3 on “Is Z = 1?” and gains 1/3 on “Is Z = 3?”.
In re-performance fidelity the objection holds. CLAIM-091’s translation (u, s, a, k, g) → (u′, s, a, k′, g′) compensates for changed social coordinates, so the target reader decides about a situation with coordinates (k′, g′). Comparing the experiments needs a correspondence z ↦ z′ and both indexed by one parameter through it. That correspondence is QUESTION-005’s, and the record has no anchor for it. Until it does, Blackwell’s order does not reach re-performance.
What this does to the decision term. In a report task the paired task is the same task asked of both readers, so the task correspondence is the identity. Then CLAIM-218’s second point (losses of two different tasks on one scale) and its fourth (an unanchored correspondence of tasks) do not arise. They arise for re-performance.
What it does not say
- It does not say which kind of fidelity literary translation aims at. Compensation (CLAIM-091) is a re-performance notion, and much translation practice may aim at it.
- It does not say the order reaches the line’s central cases. If they are re-performances, it does not, and the defeat condition’s second clause says so.
- It does not supply the correspondence of situations. That is QUESTION-005.
- It does not say how far apart incomparable experiments are in Le Cam’s sense; that rests on Torgersen (LIT-778), unread here.
Note of 2026-10-10: a documented case
CASE-063, found by a search the owner asked for, supplies the documented case this claim lacked: the English localization of Ace Attorney, where the localizer names the two targets herself (“textually accurate” against “emotionally accurate”), states the correspondence of situations (an alternate Los Angeles without the Alien Land Law of 1913), and the fidelity is contested on record. It does not supply the defeat condition’s evidence: no receivers have been given a report task and a re-performance task on the same renderings, and the critics-against-fans split could be a difference of weights (CLAIM-240). Hsu’s “strike a balance” is the strongest reply for CLAIM-258.
The same search found a case that suggests this claim may need narrowing. “What it does not say” calls compensation (CLAIM-091) “a re-performance notion”. Translators who rendered “Son of God” with non-biological terms for Muslim readers defend the compensating rendering as the better report: the literal word “add[s] a procreative meaning that was absent from the original”. If that holds, compensation can serve report fidelity for a reader whose code differs, and the line should say that compensation is mainly, not only, a re-performance notion. The renderings are known only through a critic, and the reader evidence is the translators’ own, so the narrowing is proposed, not made.
Note of 2026-10-27: the narrowing filed as an objection
The narrowing the note of 2026-10-10 proposed and did not make, that compensation is mainly, not only, a re-performance notion, is filed as CLAIM-461, a counter, Proposed, so it can carry a status. It awaits a case whose source is read. The claim is unchanged.
Note of 2026-10-27: grounded in what it uses of Blackwell
The status review of this date (ADR-057) moved this claim’s ground from THEORY-156 to THEORY-1096. It uses Blackwell’s definitions (an experiment, a garbling, the informativeness order) or the easy direction, that a garbling never helps a decision-maker free to choose any rule. Both hold on any outcome spaces, with the proof written out there. THEORY-156’s general converse, which stays Proposed, is not used. Quotations of THEORY-156 above are of the definitions it states.
Note of 2026-10-30: what holds it back
The result its defeat condition names is missing: no receivers have been given a report task and a re-performance task on the same renderings, and the line’s central cases are not classified as reports or re-performances. CLAIM-461’s objection, that compensation also serves report fidelity, is unanswered: the history note of 2026-10-27 names it, but the sentence it targets under “What it does not say” is unchanged.