Dataset from a Quasi-Experimental Study on Enhancing Academic Integrity and Reducing Plagiarism through a Generative AI-Based Paraphrasing Tool in Multilingual Higher Education

This dataset is derived from a quasi-experimental study titled “Enhancing Academic Integrity and Reducing Plagiarism through a Generative AI-Based Paraphrasing Tool: A Quasi-Experimental Study in Multilingual Writing Education.” The study investigated the impact of Languafrasa, a generative AI-based paraphrasing tool specifically designed to support ethical academic writing, on the performance, ethical awareness, and perception of undergraduate students engaged in multilingual education.

The study involved 300 undergraduate participants divided equally into experimental and control groups. Over a six-week period, the experimental group utilized the AI tool to complete weekly paraphrasing tasks integrated into a learning management system, while the control group performed the same tasks without access to AI. The dataset includes demographic profiles (age, gender, and field of study), pretest and posttest similarity scores (including gain scores), rubric-based evaluations of paraphrasing quality across five dimensions—semantic fidelity, syntactic transformation, lexical diversity, citation ethics, and clarity—along with measures of engagement and ethical awareness, and student perception survey responses in both Likert and open-ended formats.

All statistical procedures were conducted using JASP, including descriptive statistics, assumption testing, paired and independent samples t-tests, one-way ANOVA, MANOVA, and Wilcoxon signed-rank tests. Results are presented in six summary PDF files corresponding to the study’s three research questions. Each file contains structured outputs, effect size calculations, and narrative interpretations.

The dataset offers a comprehensive, transparent, and replicable resource for researchers, educators, and instructional designers interested in academic integrity, AI-assisted writing instruction, and multilingual learning environments in higher education.

这个数据集源自一项题为《通过生成式AI（Generative AI）辅助改写工具提升学术诚信与减少剽窃行为：多语言写作教育中的准实验研究》的准实验研究。该研究旨在探究Languafrasa——一款专为支持伦理学术写作设计的生成式AI辅助改写工具——对参与多语言教育的本科生在学习表现、伦理意识及认知方面的影响。

研究招募了300名本科生参与者，平均分为实验组与对照组。在为期六周的研究中，实验组使用该AI工具完成集成于学习管理系统的每周改写任务，而对照组则在无AI辅助的情况下完成相同任务。数据集包含以下内容：参与者人口统计学特征（年龄、性别及研究领域）、前测与后测相似度得分（含增益分数）、基于评分量表的改写质量评估（涵盖语义保真度、句法转换、词汇多样性、引用伦理及清晰度五个维度）、参与度与伦理意识的测量指标，以及学生感知调查的结果（含李克特量表与开放式问答两种形式）。

所有统计分析均通过JASP软件完成，包括描述性统计、假设检验、配对样本t检验与独立样本t检验、单因素方差分析（ANOVA）、多变量方差分析（MANOVA）及威尔科克森符号秩检验。研究结果以六份摘要PDF文件呈现，对应研究的三个核心问题。每份文件均包含结构化输出、效应量计算及叙述性解读。

该数据集为关注学术诚信、AI辅助写作教学及高等教育多语言学习环境的研究者、教育工作者及教学设计人员提供了一个全面、透明且可复现的资源库。