Identifying and Handling Suspected Cases of Unauthorized Use of AI
At the University of Graz, instructors determine whether and in what form students may use artificial intelligence (AI) in courses and for exams. Whether and how AI may be used in a course must be specified in the course description before the start of the semester. Relevant guidelines and text templates are available in the Resources section.
If instructors - particularly in the case of written assignments or assessment components - suspect that students have used AI applications without authorization, this raises two fundamental questions:
1. How can unauthorized use of AI be identified?
2. What measures can be taken?
Identifying Unauthorized Use of AI
Unlike traditional cases of plagiarism, it is currently not possible to detect AI-generated texts with sufficient accuracy using software applications (e.g., plagiarism detection software). Nor is the use of an AI application suitable for verifying whether a text was created using AI. Additionally, data protection and copyright considerations must be taken into account here. An attempt to deceive by using an AI-generated text is therefore most comparable to ghostwriting.
However, the following indicators, among others, may point to AI-generated text:
- Thematic or factual inconsistencies in the text document
- Insufficient originality of content
- Stylistic inconsistencies throughout the text document
- Frequent use of unusual word choices
- Frequent linguistic redundancies
- Non-existent source citations
While the presence of these indicators may give rise to suspicion, it does not automatically lead to the conclusion that the text in question (or individual parts of it) was in fact created using an unauthorized AI application.
Steps to Take When Unauthorized Use of AI Is Suspected
If there is suspicion of unauthorized use of AI in the production of written exam work, it is recommended to speak with the student. This allows you to get a sense of the situation and better assess whether unauthorized use of AI has occurred.
The following section presents example scenarios for handling suspected cases in the form of FAQs: