Scholarlink

<p>Analyzing the teaching and learning environment (TLE) through student feedback is essential for identifying curricular gaps and improving teaching practices. However, traditional feedback analysis methods, particularly for qualitative data, are often time-consuming and prone to human bias. Large Language Models (LLMs) offer a promising solution by facilitating multimodal data analysis. This study presents a novel framework that utilizes multi-agent LLMs to analyze feedback from 7,160 medical residents, integrating both quantitative and qualitative data to generate comprehensive reports. Quantitative analysis was conducted using SPSS for descriptive statistics, trend analysis, and non-parametric tests, while LLMs handled sentiment analysis and topic identification for qualitative feedback, with results visualized through word clouds. The findings from both qualitative and quantitative analyses were integrated as multimodal data, comprising text and images. The multi-agent system, comprising report generation and modification agents, analyzed multimodal data to produce and refine reports. This iterative process significantly improved report quality in terms of balance, clarity, semantics, readability, and coherence. The findings suggest that multi-agent LLMs improve the efficiency and depth of feedback analysis while providing a scalable solution for generating customized and comprehensive reports. This approach has far-reaching implications for educational data analysis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Analyzing the teaching and learning environments through student feedback at scale: a multi-agent LLMs framework

  • 期刊论文

  • 发表日期：

    2025年6月4日

查看全文
  • Chang Cai,
  • Shengxin Hong,
  • Min Ma,
  • Haiyue Feng,
  • Sixuan Du,
  • Minyang Chow,
  • Winnie Li-Lian Teo,
  • Siyuan Liu,
  • Xiuyi Fan

摘要

Analyzing the teaching and learning environment (TLE) through student feedback is essential for identifying curricular gaps and improving teaching practices. However, traditional feedback analysis methods, particularly for qualitative data, are often time-consuming and prone to human bias. Large Language Models (LLMs) offer a promising solution by facilitating multimodal data analysis. This study presents a novel framework that utilizes multi-agent LLMs to analyze feedback from 7,160 medical residents, integrating both quantitative and qualitative data to generate comprehensive reports. Quantitative analysis was conducted using SPSS for descriptive statistics, trend analysis, and non-parametric tests, while LLMs handled sentiment analysis and topic identification for qualitative feedback, with results visualized through word clouds. The findings from both qualitative and quantitative analyses were integrated as multimodal data, comprising text and images. The multi-agent system, comprising report generation and modification agents, analyzed multimodal data to produce and refine reports. This iterative process significantly improved report quality in terms of balance, clarity, semantics, readability, and coherence. The findings suggest that multi-agent LLMs improve the efficiency and depth of feedback analysis while providing a scalable solution for generating customized and comprehensive reports. This approach has far-reaching implications for educational data analysis.