PERSONALIZATION WITH AI IN FOREIGN-LANGUAGE TEACHING: POSSIBILITIES AND LIMITS

Abstract

This article examines how artificial intelligence (AI) can support personalized learning in foreign-language teaching and translation studies, and where sensible boundaries of its use should be drawn. The empirical base comprises a survey of 100 participants (70 students and 30 instructors) and a comparative check of Google Translate, DeepL, and ChatGPT on educational texts. We assessed semantic adequacy, contextual accuracy, and stylistic appropriateness. The survey used five-point scales with two open-ended prompts; after a short pilot, rating criteria were aligned to improve inter-rater agreement.

Findings suggest that AI helps tailor learning paths to students’ levels, speeds up work with vocabulary and corpora, handles routine operations (draft translation, terminology lookup, first-pass post-editing), and frees classroom time for higher-order tasks. At the same time, we observed clear limits. Some learners develop a habit of skimming rather than reading critically, as ready-made prompts nudge them toward shallow solutions. Machine translation occasionally drops cultural references, domain terminology, or implied meaning; stylistic drift and “confident errors” also occur. Dependence on tools tends to be stronger among novices and declines as proficiency grows. AI has become a mainstream study aid that speeds routine tasks and supports personalization; at the same time, sizeable cohorts perceive cognitive costs. Automated MT is serviceable in technical/scientific genres but does not replace human translation for culturally dense texts; pedagogically, this points to the need for critical verification and systematic post-editing of AI outputs.