AI辅助写作系统中的中国英语学习者语用失误特征研究

Characteristics of Pragmatic Failures among Chinese EFL Learners in AI-Assisted Writing Systems

摘要: 随着生成式AI技术在外语教育领域的深度应用，AI辅助写作已成为中国英语学习者提升写作能力的主流方式。然而，现有AI写作工具普遍存在“重语法轻语用”的缺陷，导致学习者频繁出现“语法正确但交际失效”的问题。本研究以120名非英语专业本科生为研究对象，构建了包含240篇平行文本的“AI辅助–传统手写”对比语料库，系统分析了中国学习者在Grammarly和批改网辅助写作中的语用失误特征。本研究发现：AI辅助写作中语用失误占总错误的42.7%，显著高于传统写作的28.3%；语用失误呈现三类典型特征：表层语用语言失误的模板化、深层社交语用失误的跨文化冲突性、算法诱导性失误的独特性；不同写作体裁与不同AI使用频率学习者的语用失误分布存在显著差异；AI系统的语用诊断效能严重不足，精确率仅为32.7%。本研究揭示了AI场景下语用失误的独有规律，为优化AI写作工具功能与英语写作教学提供了实证依据。

Abstract: With the widespread integration of generative AI in foreign language education, AI-assisted writing has become a primary method for Chinese EFL learners to enhance their writing skills. However, mainstream AI writing tools including Grammarly and Pigaiwang prioritize grammatical accuracy over pragmatic appropriateness, leading learners to produce grammatically correct but communicatively ineffective writing. This study established a 240-text paired corpus based on writing samples from 120 non-English major undergraduates, comparing pragmatic failures between AI-assisted and traditional handwritten writing. The results show that pragmatic failures occupy 42.7% of total errors in AI-assisted writing, significantly higher than the 28.3% in traditional writing. Three distinctive types of pragmatic failures are identified: templatized pragmatic-linguistic failures, cross-cultural sociopragmatic failures, and unique algorithm-induced failures. Their distribution differs notably across writing genres and AI usage frequencies. Additionally, current AI systems demonstrate poor pragmatic diagnosis performance, with an accuracy rate of only 32.7%. This study explores the unique patterns of pragmatic failures in AI-assisted writing. Its findings offer empirical support for optimizing intelligent writing tools and improving pragmatic-oriented English writing teaching.