Scholarlink

<p>Generative artificial intelligence (GenAI) tools such as ChatGPT have rapidly spread across higher education, offering new possibilities for feedback, problem-solving support, and collaborative idea generation. Yet empirical evidence from real-world STEM classrooms where GenAI use is permitted—particularly in mathematics—remains limited. This study investigated relationships between students’ engagement in course-embedded safeguards (CLA and RLJ) and academic achievement in a first-year university calculus course in South Korea (<i>N</i> = 230; 16-week semester) where GenAI use was permitted throughout. The course integrated Collaborative Learning Activities (CLA) and Reflective Learning Journals (RLJ), with learning outcomes assessed via exams and team projects. RLJ participation showed a strong association with academic achievement (<i>r</i> = .705, <i>p</i> &lt; .001), with achievement increasing across participation levels before approaching a plateau at the highest levels. Students also scored higher on exam items aligned with CLA than on non-aligned items, although the effect was small (paired-samples <i>t</i> test, <i>n</i> = 211, <i>t</i>(210) = 2.53, <i>p</i> = .012, <i>dz</i> = 0.17). In exploratory comparisons, performance did not differ significantly between an AI-solvable and an AI-unsolvable exam item, suggesting that achievement in a GenAI-permitted context may depend less on direct AI assistance and more on students’ reflective verification and peer dialogue. Overall, the findings underscore the value of designing AI-permitted environments that cultivate reflective checking and collaborative reasoning, alongside guidance for responsible use.</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

Integrating Generative AI in University Mathematics: reflective and collaborative learning as pedagogical safeguards

  • 期刊论文

  • 发表日期：

    2026年3月19日

查看全文
  • Won-Young Choi

摘要

Generative artificial intelligence (GenAI) tools such as ChatGPT have rapidly spread across higher education, offering new possibilities for feedback, problem-solving support, and collaborative idea generation. Yet empirical evidence from real-world STEM classrooms where GenAI use is permitted—particularly in mathematics—remains limited. This study investigated relationships between students’ engagement in course-embedded safeguards (CLA and RLJ) and academic achievement in a first-year university calculus course in South Korea (N = 230; 16-week semester) where GenAI use was permitted throughout. The course integrated Collaborative Learning Activities (CLA) and Reflective Learning Journals (RLJ), with learning outcomes assessed via exams and team projects. RLJ participation showed a strong association with academic achievement (r = .705, p < .001), with achievement increasing across participation levels before approaching a plateau at the highest levels. Students also scored higher on exam items aligned with CLA than on non-aligned items, although the effect was small (paired-samples t test, n = 211, t(210) = 2.53, p = .012, dz = 0.17). In exploratory comparisons, performance did not differ significantly between an AI-solvable and an AI-unsolvable exam item, suggesting that achievement in a GenAI-permitted context may depend less on direct AI assistance and more on students’ reflective verification and peer dialogue. Overall, the findings underscore the value of designing AI-permitted environments that cultivate reflective checking and collaborative reasoning, alongside guidance for responsible use.