• Dedakhanov Abdumalik Mutalliyevich

  Author

The rapid development of Large Language Models (LLMs) like the conversational systems based on GPT has transformed from a research curiosity to a tool that can now be found in the daily lives of English language teachers and learners, raising pressing questions about how it can be best used in teaching and learning. This paper reviews the potential for the use of the LLM capabilities (such as automated feedback generation, error detection, and conversational practice) to transform English teaching workflows, and presents results from a classroom-based evaluation of an LLM-supported teaching support system. The design was a mixed-method quasi-experimental design, with 18 English teachers using LLM-generated feedback and lesson support materials for 10 weeks, and 18 English teachers who had been using a set of lesson-planning and feedback methods for a period of time continuing to do so without the use of LLM support materials. The data were collected using pre- and post-intervention teacher surveys, learner writing samples, classroom observation (notes), and semi-structured interviews, and then analyzed using descriptive statistics and thematic coding. Results reveal that educators employing the language model-assisted system stated that preparation time declined and provided students with individual feedback more frequently, while students in their classes demonstrated a more considerable improvement in grammar and revision than the control group. The suggested model incorporates the content-generation level, the level of grammar and bias supervision, the teacher-administered system, and the classroom-based feedback cycle, putting teachers' decision-making in the center of language model utilization instead of treating the system as an independent teacher. The investigation raised certain questions regarding the high dependence on the system, reliability of the output, and changes in the professional self-conception of educators acting as the evaluators of the work done. To sum up, it can be said that large language models can enhance efficiency and personalization in English teaching if used in the framework of teacher supervision and control, while it is essential to avoid direct application of the language models and ensure ongoing learning of teachers and continuous ability to monitor the models' output.

• PDF