当代外语研究

当代外语研究 ›› 2025, Vol. 25 ›› Issue (5): 156-169.doi: 10.3969/j.issn.1674-8921.2025.05.016

• 思源学脉 • 上一篇    下一篇

人机协同反馈对学习者投入与修订质量的影响——基于文学翻译译后编辑的实证研究

吕倩兮1(), 姜兆坤1,2()   

  1. 1.上海交通大学,上海,200240
    2.新加坡国立大学,新加坡,138601
  • 出版日期:2025-10-28 发布日期:2025-11-07
  • 作者简介:吕倩兮,博士,上海交通大学助理教授、硕士生导师。主要研究方向为翻译与口译研究、心理语言学、计算语言学。电子邮箱:lvqianxi@sjtu.edu.cn;
    姜兆坤,新加坡国立大学博士研究生。主要研究方向为口笔译评估、语言测试、计算语言学。电子邮箱:jiangzhaokun@u.nus.edu。
  • 基金资助:
    *教育部人文社会科学研究青年基金项目“基于机器学习的对外新闻英译情感分析及其传播效果研究”(编号GKZY1400022);上海外国语大学语言科学与多语智能应用重点实验室年度开放课题“ChatGPT、神经机器翻译与人工翻译的多维对比及人机合作策略研究”(编号KLSMAI-2023-OP-0002)

The Impact of GenAI-Instructor Collaborative Feedback on Learner Engagement and Revision Quality: An Empirical Study Based on Literary Translation Post-Editing

LV Qianxi1(), JIANG Zhaokun1,2()   

  • Online:2025-10-28 Published:2025-11-07

RichHTML

1

PDF (PC)

4

摘要：

在生成式人工智能(GenAI)快速融入翻译能力培养的背景下,如何优化反馈机制以促进学习投入,已成为亟须解决的问题。本研究采用混合研究方法,系统考察反馈模式(GenAI单一反馈 vs. GenAI-教师协同反馈)与反馈复杂度如何影响学习者在机器翻译译后编辑(MTPE)学习中的情感、认知、行为投入及再修订质量。24名高年级本科生完成包含典型修辞手法的文学MTPE任务,数据来自其修订结果、问卷测量与半结构化访谈。结果显示:(1)协同反馈显著提升三维投入与修订质量;(2)学习者偏好简洁且情境契合的反馈,反馈复杂度需依任务与个体差异动态调整;(3)情感与认知投入的协同作用有助于推动有效修订。研究证实协同反馈在可信度、可理解性与可操作性上优于单一反馈,更能激发学习投入。这一结果为优化人工智能时代的MTPE教学反馈机制提供了实证支持,并为人机协同的智能化教育干预提供了参考。

关键词: 协同反馈, 投入, 生成式AI, 反馈复杂度, 机器翻译译后编辑

Abstract:

The integration of GenAI into translation training is calling for the optimization of feedback mechanisms to enhance learner engagement. This study investigates the influence of feedback modality (GenAI-only feedback vs. GenAI-instructor collaborative feedback) and feedback complexity on learners’ emotional, cognitive, and behavioral engagement, as well as on the quality of subsequent revisions. Twenty-four senior undergraduate students completed a literary MTPE task incorporating typical rhetorical devices. Data were obtained from their revised outputs, questionnaire responses, and semi-structured interviews. Our findings reveal that: (1) collaborative feedback significantly enhanced three dimensions of engagement and revision quality; (2) learners favored concise and contextually relevant feedback, indicating that feedback complexity should be dynamically tailored to task characteristics and individual learner profiles; and (3) the interaction between emotional and cognitive engagement facilitated more effective revisions. Overall, collaborative feedback demonstrated superior credibility, comprehensibility, and actionability compared to GenAI-only feedback, thereby fostering learner engagement more effectively. This research offers empirical evidence for refining feedback mechanisms in MTPE instruction in the era of AI and provides practical insights for the design of intelligent, human—machine collaborative educational strategies.

Key words: collaborative feedback, engagement, generative AI, feedback complexity, MTPE

中图分类号: