生成式AI辅助下高中生英语写作"人机协同"反馈模式的行动研究

摘要

高中英语写作反馈长期面临班级容量大、反馈周期长、难以兼顾深度与个性化等困境。生成式人工智能（GAI）的兴起为破解这一难题提供了技术可能，但现有研究多聚焦"教师—AI"二元混合，鲜少将同伴反馈纳入同一框架系统探讨三者的协同机制。本研究采用行动研究范式，通过三轮"计划—行动—观察—反思"循环，在46名高二学生中构建并优化了"教师反馈—AI反馈—同伴反馈"三位一体的协同反馈模式。研究运用混合方法收集数据，发现：（1）协同模式显著提升了学生写作质量，尤其在语言准确性与篇章结构方面效果突出；（2）三种反馈形成功能互补：AI侧重语言形式即时纠错，同伴促进内容逻辑与读者意识，教师发挥高阶诊断与情感支持作用；（3）运行中存在"AI依赖""面子问题""教师负担"等张力，可通过批判性使用培训、话语支架与聚焦式精批等策略有效调适。研究提出了"反馈分工梯度模型"，为生成式AI时代的英语写作教学反馈体系优化提供了实践参考。

Abstract

Feedback provision in senior high school English writing instruction has long been plagued by challenges including large class sizes, protracted feedback cycles, and the difficulty of simultaneously ensuring depth and personalization. While the advent of Generative Artificial Intelligence (GAI) presents a technological avenue for ameliorating these issues, extant research has predominantly adopted a binary "teacher-AI" comparative framework, with scant attention paid to integrating peer feedback into a unified system to examine the tripartite collaborative mechanism. Drawing upon an action research paradigm, this study constructed and refined a triadic collaborative feedback model—encompassing teacher feedback, AI feedback, and peer feedback—through three iterative cycles of planning, acting, observing, and reflecting, involving 46 Grade-11 students. Utilizing a mixed-methods design, the findings revealed that: (1) the collaborative model significantly enhanced students' writing quality , with particularly marked improvements in linguistic accuracy and textual coherence; (2) the three feedback modalities established a functionally complementary division of labor, wherein AI feedback targeted immediate linguistic error correction, peer feedback fostered content logic and reader awareness, and teacher feedback provided high-order diagnostic insights and affective support; and (3) the model's implementation was characterized by operational tensions, including AI over-reliance, face-saving dilemmas in peer review, and unsustainable teacher workload, which were effectively mitigated through critical-use training, structured discourse scaffolds, and focused intensive marking. The study culminates in the proposal of a "Graduated Feedback Division Model," contributing practical insights for the optimization of feedback ecosystems in English writing pedagogy within the generative AI landscape.

参考文献