Impact of artificial intelligence on the development of academic and intercultural communicative competence in students of non-language majors during foreign language learning at university

Abstract

The text analyzes how the systematic integration of generative artificial intelligence tools into foreign language teaching at university influences the formation of academic and intercultural communicative competence in students of engineering and economic majors, for whom language serves as a tool for professional and scientific communication. The theoretical argumentation connects the spread of machine translation and automated text generation with the need to rethink reproductive methods and enhance cognitive engagement of learners, including the development of critical thinking and information verification skills, as well as the recognition of possible cultural biases and stereotypes transmitted by digital agents. The effect of AI implementation is evaluated by comparing control and experimental student groups (period 2021-2024, total 482 persons) using standardized linguodidactic measures oriented towards CEFR descriptors, diagnostics of intercultural sensitivity based on situational cases, and expert assessment of written and oral products; statistical verification of differences and correlations is performed using the t-test, Mann-Whitney U test, correlation analysis, and χ². The presented indicators demonstrate a significant increase in lexical density and quality of academic writing under AI support (Type/Token 0.61 vs. 0.48; 2.83 errors per 100 words vs. 5.12; academicity index 28.94% vs. 18.32; text coherence 8.47 vs. 6.89), as well as pronounced shifts in intercultural components, particularly in politeness strategies (high level 49,38% vs. 19.47) and sociocultural awareness (40,59% vs. 14.16). Additionally, an increase in learning autonomy and engagement is recorded alongside a reduction in result dispersion, which is interpreted as an effect of personalization and immediate feedback. Pedagogically significant limitations are discussed: the risk of «illusion of competence», technological dependence, and the need for methodological control, under which AI functions as a catalyst for speech-thinking processes rather than a replacement for independent production.