GitHub Copilot in Computing Education: A Field Study Dataset

Description
This dataset was generated from a field study conducted to evaluate the integration of the AI pair programmer, GitHub Copilot, into project-based computer science and software engineering curriculums. The primary aim was to assess how undergraduate students perceive and utilize AI-assisted coding tools during a realistic development task, and to measure the tool's impact on their productivity, learning, and final project quality.
Methodology
Participants: 114 senior undergraduate students from Software Engineering and Computer Science programs in Three Universities in Kurdistan Region of Iraq.
Task: Participants were tasked with building a static personal portfolio website using HTML, CSS, and JavaScript.
Intervention: GitHub Copilot was used as the primary AI-assisted coding tool within the Visual Studio Code IDE.
Design: The study followed a structured, project-based learning protocol. Students participated in a guided competition to complete the task, adhering to a predefined set of functional and non-functional requirements.
Data Collection: Three streams of data were collected:
Perceptual Survey: A post-task survey using a 5-point Likert scale measured five constructs adapted from the Technology Acceptance Model (TAM).
Objective Performance Metrics: Actual project completion time (in hours) and a final project score assessed via a detailed rubric.
Qualitative Feedback: An open-ended question prompted students to describe a critical engagement with an incorrect or suboptimal AI suggestion.

Dataset Description

Column Name; Description; Data Type Allowed Values / Range
No: Unique participant identifier Integer 1-114
Gender: Participant gender String (Nominal) m (Male), f (Female)
Perceived Ease of Use: TAM construct: Ease of using GitHub Copilot Ordinal (Integer) 1 (Strongly Disagree) - 5 (Strongly Agree)
Perceived Code Quality: TAM construct: Belief that Copilot improved output code quality Ordinal (Integer) 1-5
Perceived Time Efficiency: TAM construct: Belief that Copilot saved time Ordinal (Integer) 1-5
Support in Learning Construct: Belief that Copilot aided in learning new concepts Ordinal (Integer) 1-5
Willingness to Recommend: TAM construct: Intention to adopt/recommend the tool Ordinal (Integer) 1-5
Actual Completion time Total project duration in hours: Continuous (Float) 8-15 hours
Final Project Score-Functionality: Rubric sub-score for functional requirements. Continuous (Float) 0-10
Final Project Score-Readability: Rubric sub-score for code structure & readability. Continuous (Float) 0-10
Final Project Score-UI: Rubric sub-score for design & user interface. Continuous (Float) 0-10
Final Project Score-Responsiveness: Rubric sub-score for responsive design. Continuous (Float) 0-10
Final Project Score-Total Overall project score (average of sub-scores): Continuous (Float) 0-10
Critical Engagement: Qualitative response describing a critical interaction with Copilot String (Text) Free-text response

数据集说明
本数据集源自一项实地研究，旨在评估AI结对编程工具GitHub Copilot在基于项目的计算机科学与软件工程课程中的集成应用情况。本研究的核心目标为探究本科生在真实开发任务中如何感知并使用AI辅助编码工具，并量化该工具对学生编程效率、学习效果以及最终项目质量的影响。

研究方法
参与者：来自伊拉克库尔德斯坦地区三所高校软件工程与计算机科学专业的114名大四本科生。
任务：要求参与者使用HTML、CSS及JavaScript构建静态个人作品集网站。
干预手段：在Visual Studio Code集成开发环境中，将GitHub Copilot作为核心AI辅助编码工具进行使用。
研究设计：本研究遵循标准化的基于项目学习范式，学生需遵循预先设定的功能性与非功能性需求，参与带有指导的竞赛式任务以完成项目。
数据采集：共采集三类数据：
1. 感知调查问卷：采用5点李克特量表开展任务后调研，对改编自技术接受模型（Technology Acceptance Model，TAM）的5个构念进行测量。
2. 客观性能指标：包括实际项目完成时长（单位：小时）以及通过详细评分细则评定的最终项目得分。
3. 质性反馈：设置开放式问题，要求学生描述其与AI生成的错误或次优代码建议产生关键交互的场景。

数据集详情
列名; 字段说明; 数据类型; 允许值/取值范围
编号; 唯一参与者标识符; 整数型; 1-114
性别; 参与者性别; 字符串（标称型）; m（男）、f（女）
感知易用性; 技术接受模型构念：对GitHub Copilot的使用便捷性的认知; 有序整数型; 1（非常不同意）-5（非常同意）
感知代码质量; 技术接受模型构念：认为Copilot提升了输出代码质量的认知; 有序整数型; 1-5
感知时间效率; 技术接受模型构念：认为Copilot节省了时间的认知; 有序整数型; 1-5
学习支持度; 构念：认为Copilot有助于学习新概念的认知; 有序整数型; 1-5
使用意愿; 技术接受模型构念：采纳/推荐该工具的意向; 有序整数型; 1-5
实际完成时长; 项目总耗时（单位：小时）; 连续浮点型; 8-15小时
最终项目得分-功能完整性; 功能性需求的评分细分子得分; 连续浮点型; 0-10
最终项目得分-代码可读性; 代码结构与可读性的评分细分子得分; 连续浮点型; 0-10
最终项目得分-用户界面; 设计与用户界面的评分细分子得分; 连续浮点型; 0-10
最终项目得分-响应式设计; 响应式设计的评分细分子得分; 连续浮点型; 0-10
最终项目总得分; 项目总得分（各子得分的平均值）; 连续浮点型; 0-10
关键交互经历; 描述与Copilot产生关键交互的质性反馈文本; 字符串（文本型）; 自由文本回复