The objection
At U62 the owner asked “how would you suggest we update or modify the theses?” The assistant’s answer (A240) recommended for CLAIM-228: “Retain as a methodological proposition, not the definition of all fidelity”, because “Its procedure-level observational score is well motivated, but single-token evaluations can still have meaning under other metrics”. Its second refinement:
Second, distinguish a procedure-level transport claim from an item-level fidelity judgment. A family of stochastic kernels fitted across source items can be tested for generalization. But it does not follow that an individual translated utterance has no fidelity properties. It means only that the proposed procedure-level distributional score is not identified by one isolated pair.
and its reason: “That prevents a repair to one metric from turning into an unnecessarily broad metaphysical assertion.”
Against the title as written, it is right. Version 1’s title ended: “for a single rendering of a single item it is not defined”. The subject of that clause is observational fidelity, not the measure, so read alone it denies that a single rendering has the property at all.
The body says less, three times:
- “So it is not defined as a measure.”
- “A single rendering such as Henley’s Villon (CASE-010) is judged as a sample of the procedure that produced it, or by another measure stated as such”.
- “It does not say a single translation cannot be judged, only that this measure does not judge it.”
The objection it answers says the same: CLAIM-230’s remedy is “For single renderings, a different measure, stated as such.” So the derivation shows only what A240 says it shows: one source–target pair does not identify the procedure-level score, because for one pair a kernel can make the per-item term zero (CLAIM-207).
Granted because the point is checked by reading the entry against its own title. The repair is CLAIM-228’s version 2, which restates the title and leaves the derivation, the defeat condition and the body unchanged.
What it does not say
- It does not say the procedure-level score is wrong, or that the constant-kernel argument fails for one pair. Both stand.
- It does not say any per-item measure exists or is any good. It says the title should not deny one.
- It does not take up A240’s wider phrase, “not the definition of all fidelity”. The claim was always scoped to observational fidelity as D_obs measures it; the overreach was the title’s “not defined”.
Note of 2026-10-28: moved to ARG-017
This objection was about what CLAIM-228’s derivation shows: that one source–target pair does not identify the procedure-level score, and not that a single rendering lacks fidelity. Under ADR-061 that is the argument ARG-017, from CLAIM-207, CLAIM-011 and CLAIM-381 to CLAIM-228. It is Active for the repaired title, and its body records that the inference to version 1’s title failed, as this claim said.