From Data to Discourse: Leveraging Learning Analytics to Personalize English Language Instruction in Higher Education

Abstract

     The amount of student data generated by higher education English classes is growing, but this data is often only used to watch screens instead of being turned into feedback, participation support, and teaching conversation. From student traces to discourse-sensitive educational action, this mixed-methods quasi-experimental intervention study looked at how learning analytics can personalize English language training. 180 college English as a foreign language (EFL) students were split evenly between a standard-instruction control group and a personalized group based on analytics. Multi-source factors included LMS involvement, task completion, quiz performance, writing development, rewriting behavior, feedback views, feedback uptake, forum activity, speaking participation, and English performance measures before and after the study. It was found that there are five theoretically separate groups of learners: busy but low-accuracy learners, low-engagement/high-potential learners, feedback-resistant learners, and quiet consistent learners. Higher post-test scores, stronger standardized learning gain, more feedback uptake, better rewrite quality, and more noticeable involvement among initially quiet or low-engagement learners were all linked to analytics-informed personalization compared to standard teaching. Most of the links between analytics and English learning gain were best when teachers turned signs into focused feedback, rewriting supports, and learner-facing conversation. Additionally, evidence from conversations showed that analytics changed the way teachers and students talked to each other from general correction to more targeted review questions, comments based on strategies, and support that was tailored to each student's preferences. In higher education, the study adds a data-to-discussion approach for ethical, teacher-mediated, analytics-informed English language learning.