Human-AI collaborative assessment in central Asian EFL classrooms: An action research study

This paper presents the results of a longitudinal action research study in Central Asia on human-AI collaborative assessment of English language writing among 16–18-year-olds. It aimed to enhance teacher productivity and provide timely, effective student feedback. Five instructors guided learners in generating handwritten essays, delegating the assessment to Microsoft 365 Copilot. Rubrics were assembled using the International English Language Testing System (IELTS) writing guidelines and the host educational entity's assessment policy. An AI prompt was developed, tested, and refined; teachers' experiences were documented via observations and open-ended questions. A key strength of this study is the introduction of a “Dictation Bridge,” a human-in-the-loop solution addressing the handwritten-to-digital gap in non-Western contexts. The findings indicate that successful integration is a complex, iterative process requiring collaboration, vision, and robust support. Crucially, paired t-tests confirmed that teacher-mediated AI feedback resulted in highly significant improvements in students' IELTS band scores (p < .001). The study demonstrates that AI can significantly reduce grading time from days to minutes, enabling educators to focus on higher-value pedagogical activities, such as building student self-efficacy and deeply exploring feedback. This outcome fosters a “cyborg workflow” that enhances the educator's role rather than replacing it. Notably, the study found that unmediated AI feedback may put high-achieving students at risk of stylistic flattening. While the findings are from a single study site, the paper details the methods, team dynamics, and challenges involved, providing a replicable human–AI-mediated workflow for handwritten EFL contexts and redefining the modern educator's role.