生成语法
计算机科学
学术写作
纠正性反馈
第二语言写作
数学教育
心理学
生成模型
人工智能
语言学
语言习得
同行反馈
混合学习
开发(拓扑)
学业成绩
高等教育
培训转移
自然语言处理
书面语
第二语言
人机交互
控制(管理)
作者
Hanjing Yu,Marjolijn H. Verspoor,Hongying Peng
标识
DOI:10.1080/17501229.2026.2618187
摘要
The present study explores how generative GenAI feedback and hybrid feedback affect different dimensions of text quality, as measured by Functional Adequacy (FA), in three versions of an academic writing task. 73 Chinese EFL students produced (1) an initial draft written without GenAI assistance, (2) a revision based on GenAI feedback, and (3) a further revision incorporating both human feedback and prompt training. The study first compares students’ FA scores across the three versions with traditional statistics. Then we use unconditional individual growth curve models (IGCM) that allow for variation among learners to trace changes in the four FA subdimensions: content, task requirements, comprehensibility, and coherence and cohesion. Results show significant improvements in overall FA scores and across all four subdimensions, with rapid initial gains followed by relatively slower but sustained progress. GenAI feedback effectively facilitated FA development in all dimensions, extending beyond surface-level error correction. Hybrid feedback provided added value for continued enhancement in content, task requirements, and coherence/cohesion. These findings offer both theoretical and practical insights for L2 writing research and pedagogy.
科研通智能强力驱动
Strongly Powered by AbleSci AI