<![CDATA[Using Generative AI in English language writing classes has sparked major concerns about academic honesty, since students often rely on it to write their papers for them. This quasi-experimental study compares the learning outcomes of 50 eleventh-grade students using ethically-regulated DeepSeek AI scaffolding (n=26) against those engaged in conventional peer collaboration (n=24). Guided by a Prompt Guidebook, the AI group utilized DeepSeek solely for brainstorming, vocabulary, and grammar, while the control group participated in Padlet-based discussions and peer review. Pre-test and post-test writing scores were analyzed using Wilcoxon and Mann-Whitney U tests. Results reveal that both instructional approaches yielded significant within-group improvements (p=0.000). Critically, between-group comparison showed no statistically significant difference in writing achievement (Mann-Whitney U = 295.500, p=0.748). This parity challenges the narrative of AI superiority, demonstrating that ethically-guided AI use produces outcomes equivalent to well-structured human collaboration. These findings constitute a pedagogical breakthrough: they affirm the resilience of teacher-led pedagogy, prove that AI can support learning without replacing cognitive effort, and underscore the imperative of ethical frameworks to preserve students' intellectual agency in AI-enhanced classrooms.]]>
Copyrights © 2026
