Enhancing Learners' Deep Writing Performance in Generative AI-Assisted Environments through Self-Assessment Prompts

Abstract

As generative artificial intelligence (GenAI) rapidly permeates educational practice, its influence on students’ writing behaviors has drawn increasing scholarly attention. While GenAI can substantially reduce the cognitive burden of writing and enhance textual quality, concerns have emerged regarding its potential to induce cognitive offloading, thereby weakening learners’ higher-order thinking and reflective engagement. Addressing this issue requires pedagogical interventions capable of sustaining learners’ active cognitive participation during human–AI collaborative writing. To address this challenge, this study investigates whether introducing self-assessment prompts, as a form of metacognitive intervention, can effectively enhance learners' deep writing performance in GenAI-assisted environments. Employing a randomized pre-test-post-test control group design, the study assigned 62 university students to an experimental group (receiving self-assessment prompts) and a control group (no prompts). An analysis of data from the Deep Writing Process Scale using a mixed-design ANOVA revealed that: 1) a statistically significant interaction effect between time and group was found on the total scale score (p = .046), indicating that the self-assessment intervention significantly promoted the learners' overall deep writing process; 2) among the sub-dimensions, the intervention's positive impact was particularly significant on "Effective Communication" (p = .011), with a marginally significant positive trend for "Learning Perseverance" (p = .094); and 3) the effects on "Learning to Learn" and "Self-Directed Learning Tendency" were not significant. This research confirms that the enabling effects of technology are highly contingent upon effective pedagogical design. By functioning as a metacognitive scaffold, simple self-assessment prompts can effectively counteract cognitive inertia, ensuring learners maintain their core role as active, reflective agents in the human-AI collaborative writing process.

Keywords: Generative AI, EFL writing, Deep Writing, Self-Assessment

DOI: 10.7176/JEP/16-13-05

Publication date: December 30th 2025