作者：刘清堂，蒋如意*，徐琦，郑欣欣，蒋新宇

出版刊物：Learning and Instruction

出版时间：2026年

内容摘要：

Background: Mathematics is increasingly important in the AI era, yet student engagement in mathematics classrooms remains low, constraining learning outcomes. Although teacher behavior plays a critical role in shaping engagement, prior research has largely examined isolated behavior types, offering limited insight into the differential and synergistic effects of teachers’ multimodal behaviors within classroom discourse, particularly in the Chinese cultural context. In addition, traditional approaches are often limited by subjectivity and small data scales.

Aims: This study proposes an ICAPD_Q framework for student engagement and examines the differential effects of primary mathematics teachers’ multimodal behaviors and their verbal–postural synergistic patterns on student engagement in China.

Sample: Data were drawn from 90 classroom video samples collected from primary mathematics classrooms in China.

Methods: Deep learning technologies were employed to extract large-scale process data on teachers’ multimodal behaviors and student engagement. A relational matrix network analysis was then conducted to systematically examine the differential impacts of teacher behaviors. Semi-structured teacher interviews were further used to complement and interpret the quantitative findings.

Results: While questioning verbal behaviors and adaptive postural behaviors occurred most frequently, directive verbal behaviors and proximal postural behaviors were most effective in enhancing engagement and promoting high-level engagement. Moreover, questioning-proximity and directive-proximity emerged as particularly effective dialogic discourse strategies, representing the most powerful verbal–postural synergistic patterns in the Chinese cultural context.

Conclusions: Different types of teacher verbal behaviors, postural behaviors, and their synergistic patterns exerted significantly different effects on student engagement within Chinese mathematics classrooms.

背景：数学在人工智能时代日益重要，然而数学课堂中的学生投入度仍然偏低，制约了学习成效。尽管教师行为在塑造学生投入度方面起着关键作用，但以往研究大多考察孤立的行为类型，对课堂话语中教师多模态行为的差异化效应与协同效应，尤其是在中国文化情境下的探讨，所提供的见解有限。此外，传统方法常受主观性和数据规模较小的限制。

目的：本研究提出一个面向学生投入度的ICAPD_Q框架，并考察中国小学数学教师多模态行为及其言语—体态协同模式对学生投入度的差异化影响。

样本：数据取自中国小学数学课堂收集的90个课堂视频样本。

方法：采用深度学习技术提取教师多模态行为与学生投入度的大规模过程数据。随后进行关系矩阵网络分析，以系统考察教师行为的差异化影响。并进一步采用半结构化教师访谈来补充和阐释定量研究发现。

结果：提问类言语行为和适应性体态行为出现频率最高，而指令类言语行为和近距体态行为在提升投入度和促进高水平投入方面最为有效。此外，提问—近距和指令—近距成为特别有效的对话式话语策略，代表了中国文化情境下最有力的言语—体态协同模式。

结论：不同类型的教师言语行为、体态行为及其协同模式对中国数学课堂中的学生投入度产生了显著不同的影响。