paperbot · PL 论文追踪

全部论文 待读 收藏

P2 Does ChatGPT Help With Introductory Programming?An Experiment of Students Using ChatGPT in CS1

ChatGPT 有助于入门编程吗?学生在 CS1 中使用 ChatGPT 的实验
ICSE Proceedings of the 46th International Conference on Software Engineering: Software Engineering Education and Training2024匹配度 1/10 · 论文研究的是 CS1 教学中 ChatGPT 对学生学习效果和资源使用行为的影响,属于计算机教育/HCI 方向,不涉及静态分析、程序分析、形式化验证、浮点分析或贝叶斯程序分析,也与 Agentic 程序分析无关。引用 118 inbox
Yuankai Xue, Hanlin Chen, Gina R. Bai, Robert Tairas, Yu Huang
crossref: 2026-09-12dblp: 2026-09-12

AI 速览

一句话:通过课堂对照实验发现,CS1 中使用 ChatGPT 对学习成绩无显著提升,却降低学生探索其他资源的意愿。

本文以课堂对照实验研究 ChatGPT 对 CS1 学生学习成效与编程行为的影响。56 名学生参与;实验组可在完成 UML 图与编程任务时使用 ChatGPT 3.5 及其他在线资源,对照组仅禁用 ChatGPT,并全程屏幕录制。分析结合 UML、编程与后测分数,以及任务用时、资源交互和调查反馈。结果显示:使用 ChatGPT 对任务式学习表现无显著影响;学生一旦使用 ChatGPT,会更少探索传统教育资源并倾向依赖它,而这种依赖未带来更好成绩;多数学生持中立态度,但担忧伦理问题与跨任务表现不稳定。

问题
在 CS1 入门编程课堂中,ChatGPT 是否以及如何影响学生的学习成效、编程行为和资源使用?
方法
开展受控课堂实验:56 名参与者被分为实验组和对照组,实验组可使用 ChatGPT 3.5 及其他在线资源,对照组仅禁用 ChatGPT;全程屏幕录制并收集任务成果、后测和问卷数据。
评估
以 UML 图、编程实现和后测分数衡量学习成效,结合任务用时、与 ChatGPT 和其他资源的交互记录及学生感知调查,进行定量与定性分析。
原文摘要(Abstract)

Generative AI, notably ChatGPT, has garnered attention in computer science education. This paper presents a controlled experiment that explores ChatGPT's role in CS1 in a classroom setting. Specifically, we aim to investigate the impact of ChatGPT on student learning outcomes and their behaviors when working on programming assignments. Participants were tasked with creating a UML diagram and subsequently implementing its design through programming, followed by a closed-book post-evaluation and a post-survey. All the participants were required to screen-record the whole process. In total, 56 participants were recruited, with 48 successful screen recordings. Participants in the Experimental Group can access ChatGPT 3.5 and other online resources, such as Google and Stack Overflow when creating the UML diagram and programming; however, participants in the Control Group can access all online resources except for ChatGPT (i.e., the only design variable is the access to ChatGPT). Finally, we measured and analyzed participants' learning outcomes through their UML diagram, programming, and post-evaluation scores. We also analyzed the time participants took to complete the tasks and their interactions with ChatGPT and other resources from the screen recordings. After finishing the tasks, student participants also provided their perceptions of using ChatGPT in CS1 through a post-survey. With rigorous quantitative and qualitative analysis, we found that (1) using ChatGPT does not present a significant impact on students' learning performance in the CS1 assignment-style tasks; (2) once using ChatGPT, students' tendency to explore other traditional educational resources is largely reduced (though available) and they tend to rely solely on ChatGPT, and this reliance on ChatGPT did not guarantee enhanced learning performance; (3) the majority of students hold neutral views on ChatGPT's role in CS1 programming but most of them raised concerns about its potential ethical issues and inconsistent performance across different tasks. We hope this study can help educators and students better understand the impact of ChatGPT in CS1 and inspire future work to provide proper guidelines for using ChatGPT in introductory programming classes.

链接与引用

DOI 原文 · PDF(开放获取) · DBLP

BibTeX
@inproceedings{XueCBTH24,
  title = {Does ChatGPT Help With Introductory Programming?An Experiment of Students Using ChatGPT in CS1},
  author = {Yuankai Xue and Hanlin Chen and Gina R. Bai and Robert Tairas and Yu Huang},
  booktitle = {ICSE},
  year = {2024},
  doi = {10.1145/3639474.3640076}
}