LARGE LANGUAGE MODELS IN FOREIGN LANGUAGE LEARNING AND ITS PEDAGOGICAL AND ETHICAL IMPLICATIONS

Abstract

This study investigates the use of large language models (LLMs) by university students in foreign language learning, analysing usage patterns, perceived benefits and limitations, cognitive effects, and academic integrity awareness. A survey of 398 students at the State University of Trade and Economics revealed that 84% actively use LLMs, primarily for information retrieval, translation verification, and text editing rather than conversational practice. While students report reduced anxiety and improved access to learning support, concerning trends emerge: over half experience at least occasional emotional dependence on AI, and significant proportions report declining memorisation abilities and reduced speaking confidence. Students prioritise efficiency-oriented benefits over language-specific affordances, a pattern partly attributable to the challenging learning conditions caused by Russia’s ongoing war of aggression against Ukraine. Notably, the overwhelming majority of students believe that LLM use should be regulated rather than forbidden, reflecting a broad recognition that these tools require structured institutional frameworks to be used responsibly and effectively. The findings underscore the need for clear ethical guidelines, responsible integration strategies, and assessment approaches that ensure LLMs enhance rather than replace the development of genuine language competencies.