Translation of Arabic Expressions of Impossibility by AI and Student-Translators: A Comparative Study

Expressions of impossibility (EIs) refer to events that never or rarely happen, things that are impossible to find, tasks that are difficult or impossible to perform, or things that are of no use. They are a kind of proverbs with a metaphorical meaning that speakers of a language use in daily interaction and communication. This study aims to compare the difficulties that Artificial Intelligence (AI) and student-translators have in translating EIs, the kinds of translation errors that both make, which direction is easier English-Arabic or Arabic-English, their translation strategies and the causes of errors for both. A sample of English and Arabic EIs was collected and translated by Microsoft Copilot (MC) and undergraduate students majoring in translation at the College of Language Silences, King Saud University. Data analysis showed that Arabic-English translation was easier for MC than English-Arabic translation. MC mostly gave literal word-for-word translation (once in a blue moon مرة واحدة في القمر الأزرق instead of مرة في العمر/مرة في السنة), which sometimes sounded meaningless and culturally awkward. The students translated fewer than 35% of the EIs correctly, compared to by 52% correct translations by MC. They left many blank. Expressions similar in both languages were easy to translate, whereas opaque expressions were more difficult (near the knuckle, ghost of a chance, dance on a land mine). Both MC and students gave more correct Arabic-English than English-Arabic translations. The most common translation strategy used by both was word-for-word translation. Paraphrase/explanation, partial, and extraneous translation were the most frequently used strategies by students. MC did not leave any expressions blank. Translation errors by students were due to lack of mastery of English, limited exposure to English idioms and proverbs, unfamiliar words, lack of background knowledge, cultural gaps, and inadequate translation competence. Although MC can explain the underlying meaning on an EI, it cannot make conceptual alignment because MC translation models prioritize direct linguistic accuracy, i.e., word-for-word translation, over natural, culturally adapted phrasing. Examples of correct and faulty translations by AI and students, translation strategies, sources of errors by both and recommendations for improvement are given.