Universities issue own rules to curb AI-assisted theses

Amid reports that students are increasingly using artificial intelligence tools for academic tasks, and following recent Chinese government measures to restrict the use of generative AI in scientific research, a growing number of Chinese universities have released their own updated measures to curb AI-assisted academic writing.

University World News has identified at least five universities that have issued their first notices regarding AI-generated content (AIGC) in graduation theses over the past month. They include the Beijing University of Technology, Southeast University, Tianjin University of Science and Technology, Fuzhou University and Hubei University.

Dedicated detection systems

Under the new guidelines, the universities will start testing the proportion of AI generated content using dedicated detection systems. These include systems such as VPCS and Gocheck, which are widely used by universities in the country. In its promotional literature, Gocheck claims a detection “accuracy rate” of above 99%.

Students found to have used more AI generated content than is deemed acceptable — the proportion is not always specified in university documents — will be warned and asked to “self-correct”.

In a 28 April notification, Fuzhou University in China’s southern Fujian Province told students it will test both the draft and final versions of their graduation papers using an external platform. The university said the new rules aimed to strengthen academic integrity, foster a “good academic ecosystem”, and standardise the scientific and reasonable use of AI technology.

Tianjin University of Science and Technology set out more specific rules, stating that once the proportion of AI-generated text in the undergraduate graduation thesis is found to exceed 40%, it will issue a warning to the student. “After receiving the warning, the student must make serious revisions before the university conducts another inspection through the relevant AIGC detection channels,” it said.

Similarly, Hubei University’s undergraduate school will also identify whether ‘AI ghostwriting’ is used in theses and issue a corresponding report for reference. If the paper is found to be at “high risk” of being AI-ghostwritten, students will be asked to make changes “under professional guidance”, it said.

New law includes ‘ghostwriting’

Some university circulars have quoted from China’s newly-enacted Academic Degrees Law, which states that the malpractice of ghostwriting, plagiarism, and counterfeiting may result in a degree being revoked. China adopted the new law in March, to take effect on 1 January 2025. The bill is the culmination of efforts since 2018 to formalise longstanding degree regulations.

Last year, the draft version of the legislation set out conditions under which degrees may not be conferred or can be revoked. In this draft, “AI-assisted ghostwriting” was included for the first time, attracting widespread attention.

Although references to AI were ultimately omitted from the bill passed by the National People’s Congress in March, academic misconduct such as ghostwriting and plagiarism will now face penalties, including revocation, under the new law, leaving it to universities to identify the ghostwriting.

June marks China’s annual graduation season. Final-year students are expected to finalise their graduation theses by April before the oral examination known as thesis defence. This year, a record 11.7 million graduates are set to enter the job market across China, according to the government’s latest work report.

Scale of AI use

Studies have found that Chinese university students increasingly turn to AI tools for help with academic tasks. In March, a research project from Shanghai-based East China Normal University revealed that around 40% of university students in Shanghai had used ChatGPT.
Project lead Yao Rong, associate professor at the Institute of Higher Education in the Faculty of Education at East China Normal University, said students find it a “novel and trendsetting” experience to use tools like ChatGPT.

Although ChatGPT is not officially available to mainland Chinese users, students gain access by purchasing foreign phones numbers and using VPNs (virtual private networks) which hide their IP address to bypass the Chinese ‘Firewall’.

Despite its popularity, educators are now grappling with the risks and limitations of generative AI in the academic setting.

“One major challenge is the accuracy of AI-generated information … a lot of [it] is false, or fabricated out of thin air,” said Hu Dan, assistant professor at Beijing Foreign Studies University.

“It’s dangerous when the databases that AI platforms are built on are of questionable quality. Experienced teaching staff may be able to distinguish between right and wrong, but without an in-depth understanding of the subject matter, it would test the student’s ability to make the right judgement,” she told University World News.

AI in scientific research

Amid rising concerns about academic integrity, the Chinese government is also sounding the alarm on the use of AI in scientific research. In December 2023, the Ministry of Science and Technology published Guidelines for Responsible Research Conduct, officially banning the direct use of generative AI in applications for research funding and approvals.

It also asks that authors make declarations and mark research content where AI has been used, especially when it involves facts and opinions.

In recent years, China’s higher education sector has reported some incidents of academic misconduct, including ghostwriting, AI-assisted ghostwriting and fraud.

In March, researchers were surprised to find common prompts of ChatGPT in a SCI (Science Citation Index) paper authored by a professor at the China University of Geosciences. The article was published in the scientific journal Surfaces and Interfaces and subsequently raised suspicions that it was ‘ghostwritten’ by AI. The university has since launched an investigation.