人工智能能打破课堂沉默吗？——以大学英语课堂会话为例

Can GPT Break Classroom Silence?—Case Study of a College English Course

摘要: 研究目的：在大学英语课堂上，课堂沉默现象较为普遍，这通常是由教师提出的抽象问题以及学生知识储备不足所导致的。然而，国内相关研究只关注课堂活跃氛围，而忽略了沉默打破是否有效促进了学生能力的提高。为了有效打破沉默，提升课堂互动效果以及学生的语言和认知能力，本研究将GPT引入到两个英语班级中，以辅助学生理解、回答教师提问。研究方法：为了探究GPT是否能够、如何打破消极的课堂沉默(NCS)，本研究从中国一所民办高校中选取了两个班级，分别作为实验组和控制组。为了衡量GPT对NCS的影响，本研究从言语互动、语言能力和认知能力这三个角度进行了前测和两次后测，采用语料库和Python算法进行测量。研究结果：本研究发现，打破沉默的效果应通过言语互动水平、技能、认知方面的提升来衡量。使用GPT的课堂和传统课堂都能够打破沉默，但使用GPT的班级在言语表述的规范性、结构性、概括性、具体化方面表现更优，语言质量提升明显，而传统课堂在言语输出量方面表现更出色。当回答者存在对智能手机的行为成瘾问题、对问题准备不足、无法有效使用GPT的提示功能时，GPT的使用效果会大打折扣。

Abstract: Object: Classroom silence frequently occurs in college English courses. This phenomenon is often attributed to the teachers’ abstract questions and the students’ inadequate knowledge reserves. In order to effectively shatter this silence and enhance interaction as well as students’ language and cognitive capabilities, this research incorporated GPT into two English classes. The aim was to help students better comprehend and respond to the teachers’ questions. Method: With the objective of investigating whether GPT can break negative classroom silence (NCS) and the way it does so, this research selected two classes from a private university in China, where NCS is prevalent, to serve as the experimental group and the control group respectively. To assess the impact of GPT on NCS, this research carried out a pretest and two post-tests. These tests were conducted from three aspects: speech interaction, language ability, and cognitive ability, and were evaluated using corpus data and Python algorithms. Result: This research revealed that the effectiveness of breaking silence should be gauged not merely by increased speech interaction but also by the enhancement of skills and cognitive abilities. Both classes with the assistance of GPT and traditional classes are capable of breaking the silence. However, the GPT-assisted group demonstrated superior performance in terms of speech specification and drawing conclusions, while the traditional class had an advantage in the quantity of speech. GPT will prove to be ineffective when a respondent suffers from behavioral addiction to smartphones, lacks sufficient preparation for the questions, or is unable to use GPT prompts properly.