Human-AI Collaboration or Academic Misconduct? Measuring AI Use in Student Writing Through Stylometric Evidence
Project Overview
The document examines the role of generative AI in education, particularly through a study that investigates authorship verification techniques aimed at assessing AI assistance in academic writing. It underscores the necessity for transparent and effective methods to differentiate between student-generated and AI-produced content, thereby fostering academic integrity and enriching the learning experience. The study emphasizes the significance of human-AI collaboration and introduces an innovative approach utilizing stylometric analysis to develop individual academic writing profiles for students. This approach not only assists educators in identifying the authenticity of student work but also aids in understanding how AI can be leveraged to enhance student learning outcomes. By focusing on these aspects, the document contributes to the ongoing discourse about the integration of AI in educational settings, highlighting both the challenges and opportunities it presents for promoting academic honesty and improving educational practices.
Key Applications
Adapted Feature Vector Difference (FVD) Authorship Verification Method
Context: Higher education, specifically for university students in various disciplines
Implementation: Developed a new AV method based on student writing profiles to quantify AI assistance in academic writing.
Outcomes: Improved ability to identify stylometric discrepancies and measure human-AI collaboration, aiding in academic integrity investigations.
Challenges: Existing detection tools have high false positive rates and may misclassify non-native speakers' writing as AI-generated.
Implementation Barriers
Technical Barrier
High false positive rates in existing AI detection tools make it difficult to accurately identify AI-generated content.
Proposed Solutions: Develop more equitable and transparent methods for distinguishing between human and AI-generated texts.
Ethical Barrier
Concerns about academic integrity and the potential for students to misuse AI to complete assignments, along with the need for authorship verification techniques.
Proposed Solutions: Employ authorship verification techniques not as punitive measures but to promote responsible AI use.
Project Team
Eduardo Araujo Oliveira
Researcher
Madhavi Mohoni
Researcher
Sonsoles López-Pernas
Researcher
Mohammed Saqr
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Eduardo Araujo Oliveira, Madhavi Mohoni, Sonsoles López-Pernas, Mohammed Saqr
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai