Scholarlink

<p>Artificial intelligence (AI) integration into education—especially in programming courses—has attracted a lot of interest. This paper investigates how ChatGPT, the generative language model of OpenAI, improves students’s computational thinking ability and programming competency in first programming classes. We compared results between ChatGPT-assisted and control groups using a pretest-posttest control group design including 160 undergraduates registered in a “Problem Solving using Python”. With effect sizes ranging from <i>p</i> &lt; 0.001, η² <b>=</b> 0.31 to 0.43, results shown statistically significant improvements (<i>p</i> &lt; 0.001) across all computational thinking dimensions, especially in algorithmic thinking and problem-solving. Comparing the experimental group—PEP-8 compliance—against conventional tools, the group showed 22% better code quality. Although ChatGPT produced tailored explanations and debugging help, we found a 15% error rate in its outputs. The paper addresses ethical issues of artificial intelligence integration in education and suggests a Guided Use Framework to reduce over-reliance risks. These results provide useful information for teachers using artificial intelligence tools while keeping academic rigidity.</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

Generative AI in Programming Education: Evaluating ChatGPT’s Effect on Computational Thinking

  • 期刊论文

  • 发表日期：

    2025年6月10日

查看全文
  • M. Sivasakthi,
  • A. Meenakshi

摘要

Artificial intelligence (AI) integration into education—especially in programming courses—has attracted a lot of interest. This paper investigates how ChatGPT, the generative language model of OpenAI, improves students’s computational thinking ability and programming competency in first programming classes. We compared results between ChatGPT-assisted and control groups using a pretest-posttest control group design including 160 undergraduates registered in a “Problem Solving using Python”. With effect sizes ranging from p < 0.001, η² = 0.31 to 0.43, results shown statistically significant improvements (p < 0.001) across all computational thinking dimensions, especially in algorithmic thinking and problem-solving. Comparing the experimental group—PEP-8 compliance—against conventional tools, the group showed 22% better code quality. Although ChatGPT produced tailored explanations and debugging help, we found a 15% error rate in its outputs. The paper addresses ethical issues of artificial intelligence integration in education and suggests a Guided Use Framework to reduce over-reliance risks. These results provide useful information for teachers using artificial intelligence tools while keeping academic rigidity.