Beyond the interface: Supporting beginner-level CEFR mediation and learner autonomy with ChatGPT

Abstract

This study explores how ChatGPT, used as a dialogic mediator, supports CEFR-based mediation and learner autonomy among multilingual university students at A1 to A2 proficiency levels in Spanish. A key finding is the contrast between learners’ observable autonomy during AI-mediated tasks and the deeper strategic awareness revealed in post-task reflections. Grounded in sociocultural theory and the CEFR (Common European Framework of Reference for Languages) Companion Volume (2020), the study examines how learners used ChatGPT to interpret texts, clarify vocabulary, and engage in structured argumentation. All tasks were conducted in Spanish, which was neither the students’ native language nor the primary language in which ChatGPT was trained. Twenty-four students completed structured Spanish-language tasks with ChatGPT and submitted written reflections in their native languages. Qualitative content analysis, guided by CEFR mediation descriptors, revealed frequent use of strategies related to Mediating Texts and Facilitating Communication. During the tasks, students often relied on ChatGPT for linguistic support. However, their reflections demonstrated higher levels of metacognitive autonomy, including goal-setting, critical evaluation of AI feedback, and comparison of digital tools. These findings highlight the importance of combining AI-supported interaction with structured reflection to promote communicative competence and digital agency. The study also points to the need for adapting CEFR mediation frameworks to better capture the dynamics of AI-mediated language learning.