Speaker
Description
Title
When students lead feedback: Ethical AI-supported peer assessment in academic writing
Abstract
Peer assessment is widely recognized as an assessment-for-learning practice in second language writing, yet its implementation in higher education often remains inconsistent, raising concerns about feedback quality and fairness in technology-mediated environments. Although AI-supported learning is increasingly adopted, limited research has examined how AI can support peer assessment while preserving learner agency and ethical assessment practices.
This study investigated an ethical AI-supported peer assessment framework implemented in a seven-week (105-hour) preparatory English course for English-medium instruction (EMI). Participants were 96 first-year non-English major students (CEFR B2–C1) who engaged in stable peer groups of three to four in iterative cycles of writing, peer feedback, AI-supported scaffolding, and revision. A convergent parallel mixed-methods design was adopted, combining paired-sample t-tests, Pearson correlation, and thematic analysis.
The findings showed significant improvements in feedback specificity, criterion alignment, and constructiveness, with large effect sizes (d = .91–1.09, p < .001). Assessment literacy also improved significantly (M = 3.26 to 3.98, d = .96, p < .001). AI-supported engagement was moderately associated with feedback quality (r = .57, 95% CI [.41, .69], p < .001, n = 96). Qualitative findings indicated that students used AI as a scaffold while retaining responsibility for evaluative decisions. Overall, the study provides evidence that ethically designed AI-supported peer assessment can strengthen feedback practices, assessment literacy, and learner agency in academic writing for EMI readiness.
Keywords: artificial intelligence, assessment literacy, learner agency, ethical assessment, EMI writing
