Skip to main content Skip to navigation

ChatGPT or academic scientist? Distinguishing authorship with over 99% accuracy using off-the-shelf machine learning tools

Project Overview

The document explores the transformative role of generative AI, particularly ChatGPT, in education, highlighting its significant impact on academic writing. It underscores the importance of distinguishing between AI-generated and human-written content in higher education settings, where integrity and originality are crucial. To address this challenge, the authors have developed a robust model employing machine learning techniques that accurately classifies text authorship, achieving an impressive accuracy rate exceeding 99%. This innovative approach not only serves as a tool for detecting AI involvement in academic writing but also offers broader implications for enhancing educational practices and improving plagiarism detection methods. By integrating such technologies, educational institutions can better navigate the complexities introduced by generative AI, ensuring that the integrity of academic standards is maintained while also embracing the potential benefits of AI in educational contexts.

Key Applications

Model for discriminating ChatGPT-generated text from human academic writing

Context: Higher education, targeting academic scientists and educators

Implementation: Developed a supervised classification model using linguistic features to differentiate between human and AI text

Outcomes: Achieved over 99% accuracy in classifying authorship; reduced misclassification rates significantly compared to existing methods

Challenges: Adaptation of the model to various academic writing styles; ensuring the model remains effective across different disciplines

Implementation Barriers

Technical barrier

Challenges in adapting AI detection models to the unique styles of various academic disciplines

Proposed Solutions: Developing tailored models that account for specific linguistic features of different fields; expanding the dataset to include diverse writing styles

Educational barrier

Concerns about students using AI to complete assignments and the implications for academic integrity

Proposed Solutions: Implementing educational frameworks that guide appropriate AI use; enhancing student understanding of authorship and originality

Project Team

Heather Desaire

Researcher

Aleesa E. Chua

Researcher

Madeline Isom

Researcher

Romana Jarosova

Researcher

David Hua

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Heather Desaire, Aleesa E. Chua, Madeline Isom, Romana Jarosova, David Hua

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

Let us know you agree to cookies