AIDetection: A Generative AI Detection Tool for Educators Using Syntactic Matching of Common ASCII Characters As Potential 'AI Traces' Within Users' Internet Browser
Project Overview
The document explores the role of generative AI in education, highlighting the development of AIDetection, a JavaScript-based tool designed to help educators identify AI-generated content in student essays. This tool uses a heuristic approach to uncover syntactic traces indicative of AI writing, aiming to bolster academic integrity and support adherence to policies on AI usage among students. AIDetection offers advantages such as the ability to process documents in bulk and ensure compliance with GDPR regulations, but it also encounters challenges, including the risk of false positives and difficulties in analyzing non-English texts. Overall, the document emphasizes the importance of tools like AIDetection in fostering a responsible educational environment as generative AI continues to influence academic practices.
Key Applications
AIDetection.info
Context: Educational context targeting educators and students in writing courses, particularly those using AI tools.
Implementation: The tool was implemented as a web application that educators can use to scan multiple documents for AI-generated content.
Outcomes: Educators can effectively monitor compliance with AI use policies, streamline grading processes, and receive downloadable reports summarizing AI detection results.
Challenges: The tool does not provide definitive proof of AI-generated content and may yield false positives, especially in essays discussing AI topics.
Implementation Barriers
Technological
The tool's performance depends on local computing resources, which may limit its effectiveness for large-scale document processing.
Proposed Solutions: Encouraging users to process documents in smaller batches to avoid performance issues.
Accuracy
The tool may generate false positives, incorrectly flagging non-AI content as AI-generated.
Proposed Solutions: Educators should use the tool as a heuristic aid rather than definitive proof, ensuring contextual understanding.
Project Team
Andy Buschmann
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Andy Buschmann
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