Integrating large language models into EFL writing instruction: effects on performance, self-regulated learning strategies, and motivation

This study aimed to investigate the efficacy of utilizing large language models (LLMs) to enhance self-regulated learning (SRL) strategy instruction in English as a Foreign Language (EFL) writing. An LLM-supported Cognitive Academic Language Learning Model (CALLA-LLM) was developed and examined for its potential to improve elementary students' EFL writing performance, SRL strategy use, and writing motivation. In a randomized controlled trial, 65 elementary school students were divided into an experimental group receiving CALLA-LLM instruction and a control group receiving traditional CALLA instruction. Both groups learned SRL strategies over 5 weeks, with data collected pre-intervention, post-intervention, and at a one-month follow-up. Results showed that the CALLA-LLM group made significant improvements in writing performance, SRL strategy use, and writing motivation, maintained most of the gains at follow-up, and significantly outperformed the control group. Findings provide empirical evidence for the efficacy of the CALLA-LLM model in enhancing EFL writing strategy instruction, lending support for integrating AI technologies such as LLMs into English language teaching. Moreover, the study underscores the importance of the "Humans in the Loop" approach, which emphasizes the essential role of human educators in AI-assisted language instruction.