Exploring emerging developmental patterns of task‐based learner–GenAI interactions: The role of learner proficiency and task type
Abstract Drawing on dynamic usage‐based and interactionist SLA perspectives, this study investigated developmental patterns of linguistic complexity and language‐related episodes (LREs) during L2 learners' written interactions with ChatGPT. Ten Korean EFL learners completed seven sessions of opinion‐gap and debate tasks over 5 weeks. Generalized additive mixed models (GAMMs) revealed dynamic, non‐linear developmental trajectories that differed by proficiency and task type. Lower‐proficiency learners exhibited an inverted U‐shaped pattern in syntactic complexity, whereas higher‐proficiency learners showed stability in syntax but a non‐linear increase in lexical sophistication. Regarding task effects, debate tasks consistently elicited greater syntactic complexity than opinion‐gap tasks and showed an overall increase in lexical sophistication. LRE analysis indicated that learners focused primarily on lexis over grammar, with reliance on the L1 increasing over time. Pedagogically, these findings highlight the dynamic nature of language development in GenAI‐mediated interactions and suggest that AI‐based tasks should be designed considering learner proficiency and task type to maximize learning opportunities.
Authors
- Soyeon Sim (ORCID: https://orcid.org/0009-0006-2490-5965)
Institutions
- Georgia State University (US)
Publication Details
- Journal
- Foreign Language Annals
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1111/flan.70088
- Primary Topic
- EFL/ESL Teaching and Learning
- Type
- article
- Field-Weighted Citation Impact
- 0.00