FROM DICTIONARY TO LLMs: RETHINKING LANGUAGE LEARNING, ASSESSMENT, AND TEACHER ROLES IN THE ERA OF GENERATIVE AI

Abstract

Assessment in language learning has become increasingly complex in the era of generative artificial intelligence, particularly when learners’ integrity and ethical awareness are not yet sufficiently developed. This study aims to map the shifting landscape of language learning amid the rapid emergence of generative AI tools such as ChatGPT and Gemini. Employing a literature review methodology, this study analyzed 20 relevant peer-reviewed articles published between 2023 and 2025 that examine language learning practices in the context of generative AI. The findings reveal an evolutionary shift in language learning practices, moving from passive reference tools such as printed and digital dictionaries toward interactive, adaptive, and dialogic systems. The expanded access to linguistic input, real-time feedback, and personalized practice is concentrated on practical and procedural aspects of language learning, including writing support, grammar correction, vocabulary practice, translation, and automated feedback rather than deeper understanding of structures, meaning-making processes, and the construction of knowledge in long-term memory. This imbalance indicates that current AI-supported practices tend to enhance efficiency and surface-level performance rather than durable language competence. The findings suggest an urgent need to redesign language assessment practices that align with emerging cultures of AI-assisted autonomous learning while maintaining academic integrity and meaningful learning.