BUV Lecturers and Students’ Research Published in Top 1% Education Journal, Advancing the Understanding of AI Text Detection in Education
Sep 18, 2024
14:35:38
British University Vietnam is proud to announce groundbreaking research led by Assoc. Prof. Mike Perkins, Head of the Centre for Research & Innovation, alongside members of the BUV GenAI Academic Research Cluster – Mr. Darius Postma, Mr. Don Hickerson, Mr. James McGaughran; collaborator Dr. Jasper Roe from James Cook University Singapore; and two BUV students, Binh Vu and Huy Khuat. The paper was published in the prestigious International Journal of Educational Technology in Higher Education, ranked among the top 1% of education journals worldwide.
“Can we accurately detect AI text content?”, ask educators worldwide in an attempt to restrictively regulate the use of AI in education. This research tackles a critical issue in the age of Generative AI (GenAI): the effectiveness of AI text detectors used to identify AI-generated content in student work.
The study rigorously tested various AI text detectors using human-written samples, unmodified AI-generated text, and AI text manipulated to evade detection. The findings highlight the limitations of current detection methods:
- AI detectors showed only 39.5% accuracy on unmodified AI-generated text
- Accuracy dropped to 22.1% when simple evasion techniques were applied
- Major differences in detectability between AI writing tools, with some being significantly harder to detect than others
These results raise important questions about the reliance on AI detectors to ensure academic integrity. As Assoc. Prof. Perkins emphasises, “It’s important to consider the potential impact on inclusivity when the accuracy of these tools can be so easily manipulated by relatively simple techniques requiring no specialist knowledge.”
The study highlights the need for a nuanced approach to AI in education, considering both the potential benefits and risks of these emerging technologies. It suggests that relying solely on AI detectors to uphold academic integrity may be problematic and calls for educators to reconsider approaches to assessment in light of AI advancements, such as the AI Assessment Scale (AIAS), a solution developed by BUV researchers. This scale provides a more comprehensive picture of student work, allowing students to utilise all available resources, including AI tools, empowering students, and educators to effectively embrace AI as a learning tool rather than an academic misconduct. The AIAS has been recognised worldwide as a progressive assessment structure and has been adopted by 20 universities worldwide since its inception.
Notably, this research was a collaborative effort between BUV lecturers and two BUV students, Binh Vu and Huy Khuat. Their contributions to the project were invaluable, demonstrating the potential of student involvement in cutting-edge research.
By understanding the limitations of AI text detectors, BUV is leading the way in creating a more inclusive and effective learning environment for the GenAI era.
Read the full research here: https://educationaltechnologyjournal.springeropen.com/articles/10.1186/s41239-024-00487-w