It Takes Two to Tango: A Longitudinal Mixed-Methods Investigation of Human-AI Co-Adaptation Across Iterative Dialogues

Abstract

Given AI’s capacity to emulate human-like cognition, its integration into educational contexts, particularly in second language (L2) learning, has drawn increasing scholarly attention (Han, 2024; Zhang, 2023). Studies have shown that AI tools can support the development of lexical diversity, grammatical accuracy, and coherence in L2 writing (e.g., Chen, 2025; Escalante et al., 2023). There is also growing interest in how these tools influence learner autonomy, engagement, and motivation (Blake, 2007; Wei, 2023). However, despite this growing body of literature, little remains known about how learners and AI systems mutually influence one another during interaction.