Artificial intelligence (AI) detection tools cannot determine on their own whether a student has committed academic fraud, the group representing Dutch universities has said following controversy over an investigation that claimed one in five recent master’s theses were mostly written by chatbots.

Dutch broadcaster said it scanned “hundreds” of recent theses by students at 12 universities using the AI detection tool Pangram and found that “the contribution of ChatGPT, Claude, or Gemini is greater than that of the student” in a fifth of them. 

Responding to the findings, the Dutch education minister Rianne Letschert said using AI to write a thesis amounts to “fraud”, adding that text generated by chatbots is “not a student’s own work”. 

But Universiteiten van Nederland (UNL) warned AI detection tools are not always accurate in identifying misconduct.

ߣߣ��Ƶ

“Based on a specific score from a detection tool, it cannot yet be determined whether a student is guilty of fraud,” it said  “However, these findings raise the question of how universities structure their final assignments to ensure that learning objectives are effectively assessed, and how graduates can demonstrate that they meet the knowledge and skills associated with a university degree.”

The group, which represents 14 universities in the country, added that changes in higher education were “well under way” to adapt to the growing use of AI in higher education. 

ߣߣ��Ƶ

It stressed that the handling of AI can vary across universities and programmes. “AI plays a different role in a medical programme than in computer science or a language study. Universities therefore have rules that align with the learning objectives of a study programme,” it said.  

The controversy has put a renewed focus on AI detection tools like Pangram and their reliability. In the Netherlands, many universities have opted not to use such tools. Delft University of Technology (TU Delft) has said that their conclusions are pointing to research that shows that the tools are often “inaccurate” and “can drive students to adopt basic ‘evasion’ prompting techniques that further reduce their accuracy”. 

Utrecht University, among the biggest in the country, has to scan students’ work. Leiden University, another major institution in the country, also professors to use such tools. “There’s a growing number of tools aimed at detecting GenAI. Unfortunately at present, and probably also in the near future, these tools do not meet a standard that justifies confronting the student with the result,” it said on its website. 

seher.asaf@timeshighereducation.com