Scholarlink

<p>While generative Artificial Intelligence (GAI) has shown considerable potential in improving writing quality, little is known about both how and to what extent students cognitively engage with GAI during writing tasks, for example, whether they critically evaluate or simply accept its suggestions. This mixed-methods study examined 41 undergraduate students’ interactions with ChatGPT-4o during academic writing revision using an adapted <i>Critical Interaction with GAI for Writing</i> framework. Thematic analysis of 102 prompts identified four engagement dimensions: <i>content revision, information seeking, writing presentation</i>, and <i>conversational engagement</i>, each coded for three levels of engagement (deep, moderate, or shallow). While most students engaged at shallow to moderate levels, a few students demonstrated deeper critical thinking engagement in the dialogic and evaluative use of GAI feedback. Additionally, results from linear mixed-effects models showed that content-focused engagement significantly facilitated the improvement in the idea presentation of academic writing. Findings from this study call for future pedagogical effort that promotes students’ critical thinking engagement when using GAI in academic writing.</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

Exploring Critical Interaction with GAI in Academic Writing: A Mixed-Methods Study

  • 期刊论文

  • 发表日期：

    2026年5月30日

查看全文
  • Qi Sun,
  • Shengkai Yin

摘要

While generative Artificial Intelligence (GAI) has shown considerable potential in improving writing quality, little is known about both how and to what extent students cognitively engage with GAI during writing tasks, for example, whether they critically evaluate or simply accept its suggestions. This mixed-methods study examined 41 undergraduate students’ interactions with ChatGPT-4o during academic writing revision using an adapted Critical Interaction with GAI for Writing framework. Thematic analysis of 102 prompts identified four engagement dimensions: content revision, information seeking, writing presentation, and conversational engagement, each coded for three levels of engagement (deep, moderate, or shallow). While most students engaged at shallow to moderate levels, a few students demonstrated deeper critical thinking engagement in the dialogic and evaluative use of GAI feedback. Additionally, results from linear mixed-effects models showed that content-focused engagement significantly facilitated the improvement in the idea presentation of academic writing. Findings from this study call for future pedagogical effort that promotes students’ critical thinking engagement when using GAI in academic writing.