Abstract:

The proposed research investigates the use of generative AI to personalize second language (L2) writing instruction through a corpus-driven approach. It examines the effect of integrating generative AI and corpus analysis in tailoring L2 reflective writing instruction to address the unique linguistic profiles of English for Academic Purposes (EAP) learners at a public university in Oman, using the DIEP (Description, Interpretation, Evaluation, Planning) Model. The study utilizes a quasi-experimental design featuring two existing sections/classes of the same Preparatory Studies Center course as intervention and comparison groups. Initially, learners from both the intervention and comparison groups are asked to complete a short in-class pretest that is in-line with the curriculum and based on a reflective writing prompt in order to establish baseline proficiency in English essay writing. During the first 6 weeks of writing instruction following the pretest, students in both groups continue with regular instruction based on the curriculum and prescribed materials. In the proceeding 6 weeks, the comparison group continues with regular instruction while the teacher of the intervention group uses a trained GenAI chatbot to customize instruction and student tasks/activities according to each learner’s specific needs. To achieve this, samples of the intervention group’s normal class-based writing work are compiled by the research team in a corpus specifically designed for each learner. GPT-4o is then employed to generate a linguistic profile of each learner’s unique writing needs. Subsequently, these profiles are analyzed with GPT-4o to develop personalized exercises and activities designed to improve student writing which the intervention group teacher applies in class. After the instructional intervention phase is complete, both groups take an in-class posttest, with results compared with the pretest to determine if statistically significant improvements in writing have occurred due to the integration of generative AI information in writing instruction. Study results show the potential to enhance writing instruction across English-medium programs and courses in Oman, and to pave the way for larger-scale research into the application of generative AI to enhance general teaching efficiency at the university and beyond.

Keywords:

GenAI, data-driven learning, Corpus-based Language Pedagogy (CBLP), reflective writing, L2 writing instruction.