Repository path: wiki/arms/ARM-realia-channel.md · rendered 2026-09-09

ARM-realia-channel — what a reader is actually placing

Deliverable advanced (T3, wiki/tracks.md §Deliverables): the evidence base under cultural-mediation — the sense whose entire content is how culture-bound items are handled — and framework/v0.2 §7, which as of S156 makes claims about located and unlocated English with no measurement of what a reader's sense of location is made of. cultural-mediation's current evidence is RS-20260727e-sense-boundary, which measured only whether raters sort decisions into the right box. Nothing has ever measured what those decisions do to a reader.

The question, in one sentence

A translated page is located twice over — by the world it describes and by the English it is written in — and this arm asks whether a reader can tell those two channels apart, because the translator can move one of them and not the other.

Why it exists

RS-20260811-floor (S156) asked three judges which national variety of English a passage was written in, forbade them in terms to answer from subject matter, and got back Akaki Akakievitch, roubles, sheepskin, Czar's, 'Stute Fish. Token concordance with the orthographic detector was 0.283 and the gate G1b failed. That run treated the result as a defect of its instrument and recorded it in §6 as a limit.

It is not only a defect. It is a finding about reading that nobody has measured, and it lands on the oldest decision in translation: whether to keep the source-culture word. If a reader's judgment of whose English this is is partly made of what the book is about, then a translator who keeps the roubles is also, silently, buying a stronger impression that the prose itself is nationally placed — a cost cultural-mediation does not know about. If the two channels are separable, the translator is free, and framework/v0.2 §7 can say so.

Steps and budget — 2 sessions

Outcome — the arm closes resolved at 2 of 2

cultural-mediation now carries a measured statement about what a realia decision does, which is exactly what the completion criterion asked for, and it is not the statement the arm expected. The sense's entry and framework/v0.2 §7.6 both now say:

There is no road out of a culture-bound item that keeps the world and leaves the prose unplaced. Transferring the source word places the world (24 of 24 on the world question) and pushes the prose away from the target's own nation (0.150 against the location-free arm). Domesticating pulls both toward it (24 of 24 against transfer on the prose question, 8 of 8 passages, sign P = 0.0039; and only 1 of 17 on the world question, i.e. the story reads less foreign too). Neutralising is the only road that leaves the prose unplaced, and it buys that by deleting the world.

What the arm learned about its own step 1 is worth as much as the result. Step 1 conflated transfer with domestication inside one KEEP arm and included proper names in its spans; both are why its P1 could only return a null. The successor separated the arms, excluded names, and tested the asymmetry directly — the last of those only because the pre-run critic noticed that the claim's central comparison was never being made.

And step 1's null is not overturned. Absolute variety verdicts did not move at n = 7 passages; side-by-side the difference is perfectly detectable. Both are true and the sense entry says so.

Completion criterion

cultural-mediation's entry in wiki/goodness-senses.md carries a measured statement about what realia handling does to a reader's placement of a translation — or a written null saying the two channels did not separate under this design, and why. Either closes the arm.

Log

• 2026-08-11 (S157) — constituted. T3 was named by check_balance.py at 5 sessions since its last principal unit and was arm-free.
• 2026-08-11 (S162) — step 2 run and frozen; the arm closes resolved at 2 of 2. Deliverable reached: cultural-mediation carries a measured statement of consequence and framework/v0.2 §7 has its paragraph. The session's translation limb, T-jutrenje-R06-v1, is the project's twentieth source language and its first South Slavic one, and its frozen three-way site table is the experiment's three arms. 2 of 2 used, consecutive: 1.
• 2026-08-11 (S157) — step 1 run and frozen. 1 of 2 used, consecutive: 1. What step 2 owes is now narrower than when the arm was written: the sense entry and the framework paragraph, plus one decision — whether the halfpennies finding (the leak runs through the translator's domestications, not through the retained foreign matter) is worth a design of its own or a sentence in cultural-mediation.