Methods · Language & communication · Short guide
G8
Translation critique
Compare your translation against alternatives and defend choices: training judgment about meaning, not only word swaps.
When to use it
- You are translating a short passage and need judgement, not a fluent takeover.
- You want to defend tone, information, and ambiguity choices line by line.
- You catch yourself picking the AI’s “best” English because it sounds nicer.
What to ask AI
Type something like: “Passage: [text]. My translation: [yours]. Give two alternative renderings without declaring a winner. Ask me to choose per sentence and justify. Probe where I smoothed away ambiguity or changed register.”
Always translate yourself first. Alternatives come second.
Keep the hard part on your side: you choose and defend; the AI does not crown a winner.
How it helps
Translation is a precision gym for bilingual control.
The learning is not finding prettier words. It is seeing what was gained or lost: tone, information, ambiguity.
AI can offer alternatives quickly. If you accept the fluent option without defending choices, you practised taste, not judgement.
How to run it
- Translate a short passage yourself first.
- Request two or three alternatives without a declared winner, or write a second draft yourself.
- Diff on meaning, tone, and register, not “which sounds nicer.”
- Defend or revise contested lines one by one.
- Back-translate a sensitive line as a check.
Example exchange
Here's a line: « Elle l'a quitté sans un mot. » I translated it as “She left him without a word.” Can you give two alternatives without crowning a winner? I want to choose and say why.
Alt A: “She walked out on him without saying anything.” Alt B: “She left him, and she didn't say a word.” Which would you keep for a literary short story, and what does yours preserve that A risks losing?
I keep mine. A adds “walked out,” which judges the leaving. I want the bare fact and the silence.
Fair. Does “without a word” undersell any emotional heat the narrator implied? Offer one revised clause only if you think there's a gap. Don't let me rewrite the whole line.
Copyable prompt
Passage: [ ].
My translation: [ ].
Give two alternative renderings without declaring a winner. Ask me to choose
per sentence and justify. Probe where I smoothed away ambiguity or changed
register.
The Tell
Here is how you know this method has flipped: you accept the model’s “best” translation with no defended choices.
Fluency is not the same as fidelity to tone and ambiguity.
If you cannot say what you kept and what you refused, the AI translated; you watched.
Principle evidence
Strength of the underlying learning idea, not a claim about AI products.
The underlying learning idea is rated moderate. Critiquing and defending translation choices builds metalinguistic awareness and justified decision-making, standard in translator education. The strength is in the defence step, not in producing a single polished target sentence.
AI delivery evidence
Whether an AI tutor delivers this method well is a separate question.
AI “better” translations and critiques are contested. Models favour fluent, unambiguous English and may erase productive ambiguity. Your first attempt and your justifications are the method; treat AI alternatives as options, never as a verdict.