Rules and Guidelines for Education

Page contents

Rules set by the Examining Boards

Unless students have explicit permission from the examiner as written in the course guide, using AI to generate the content of an assignment is considered fraud (Article 16, paragraph 3 of the 2023-2024 Rules & Regulations). Examiners can request an additional oral verification of the assignment in case of doubt. Suspicions of AI fraud will be investigated by the Examining Boards in a similar way as they investigate plagiarism. Besides prohibiting this specific usage of AI tools, the Examining Boards stress a few important points:

  • Students may use AI only to support essential academic skills like critical reflection, literature research, and scientific writing, not to replace those skills.
  • Students are always held accountable for the correctness, completeness, and coherence of their (written) assignments.
  • When using AI in an assignment, students should acknowledge this usage and reflect on how it impacts the assignment. See “How to document your GenAI use?” for more information on how to cite AI tools in (academic) texts.

If you have questions about these rules, or require assistance in the use of AI in your course, contact us at [email protected].

Guidelines for MSc Theses

The following text is included in the MSc Thesis Course Guidelines (Part A). Please note that these are minimum requirements, and that more thorough documentation is possible. For example, linking to a conversation in ChatGPT is less reliable than a hard copy of the full conversation in the appendix, as web-based links expire when the original content is deleted.

Taken from The July 13th 2023 version of the MSc Thesis Course Guidelines:

3.4 Use of generative artificial intelligence (e.g. ChatGPT, BARD, DALL-E, Elicit) 

The use of generative artificial intelligence to create ready-made content in assignments is considered fraud, so it is not allowed to copy-and-paste the output of AI. However, you are allowed to use AI as a sparring partner, and as a feedback tool for the quality of your text (e.g. as a spell checker or grammar checker). However, the use of AI is always subject to the following rules: 

• Acquiring active writing, designing and reflection skills is an important part of your internship. The use of AI should only be in support of, not as a replacement for these skills. 

• You will always be held accountable for the correctness, completeness, and coherence of all your texts. The correctness of the output of AI is never guaranteed. AI chatbots have been known to confidently assert false claims as true. You should always critically evaluate the output. 

• When you use AI for your work, acknowledge your use and report how it affects your products. 

Your thesis report should contain an appendix on the use of AI. In this appendix, you state whether you used AI for your research and report, and if so, how. In case you did not use AI, this appendix can be one sentence in which you state that you did not use AI. In all the other cases you have acknowledge your use and report how it affects your assignment. The appendix should contain a list of the prompts you used, a link to the conversation and an explanation of how you used the output of AI (i.e. in what way did the output of AI affect your text).

Detection of AI use

Within Wageningen University we have access to the AI-detection tool developed by Turnitin. When uploading a document in Turnitin, after the plagiarism check has been finalized, in addition an AI score becomes visible (upon inspecting the document). Clicking on the score shows the full output of the AI-detection tool. This tool uses the same technology as is employed by LLMs to generate texts, only in reverse. The tool uses statistics to determine the chance that certain words occur in a text, and in a specific pattern. Based on this information the tool indicates with a score how much of the document it determines to be potentially written by an AI-tool. This detector has several specific limitations, however:

  1. It does not look at the correctness of the references. This still needs to be checked through conventional means.
  2. The tool does not function for Dutch texts, only English.
  3. The texts should be longer than 300 words and shorter than 30000 words. Headers, captions, and references are excluded from this word count.

The use of the Turnitin AI detection software should never be a first step for a teacher. Instead, it may only serve as a potential confirmation of ones own suspicion. It is upon the Examining Board to determine whether fraud has been committed through the use of AI. The report generated by the Turninin AI detection software is not valid as evidence of potential AI misuse.

As LLMs are being trained to resemble human-written texts more and more, the accuracy of these detection tools becomes more difficult to trust. Hence we recommend not to rely on these tools.

Never use other online AI detection services, as this would be a violation of privacy and data security regulations.