Skip to main content
DA / EN
DA / EN

Search

Menu

Guide

The use of generative AI in teaching and for exams

Digital tools and services based on generative artificial intelligence (generative AI) can be a resource for you as a student at SDU. The following is a guide on how to declare the use of generative AI in exams, as well as how you can use generative AI in your studies in general.

What should I keep in mind when using chatbots

You are responsible for how you use the output of chatbots. The intellectual labour has to come from you. Chatbots should only be seen as a support tool. Like all other tools, methods and aids you use to solve the task, you must be able to clearly account for your use of generative AI.

Think and be critical in your assessment of the answers chatbots provide. They don’t always provide the right answer, and in some cases, they invent a wrong answer. It takes a lot of knowledge about a topic to assess whether the answer is good enough.

Be critical of the sources and quotes referred to in the answer. You cannot place responsibility with the chatbot.

    • Never use personal data and research/business-critical information in prompts. You don’t know where that data ends up or how it will be used afterwards.
      Respect the rights of others if you make use of their material in chatbots.
      Read about and listen to the security and copyright guidelines here (external page).

    • In the context of exams: Comply with the specific rules for the use of generative AI for the exam/test in question and comply with the specific rules for declaring the use of generative AI at the exam/test in question.

  • Using chatbots can be helpful in connection with studies, surveys and assignment writing.
    For example, they may:

    • Provide an answer to both simple and complex questions (based on unimaginable amounts of unnamed sources). 
    • Deliver output in the form of text, images, video, audio, tables, etc. 
    • Provide inspiration for both academic topics and other topics within and outside the programme. 
    • Provide feedback and advice on a given problem. 
    • Be a study assistant who is available around the clock.
    • Support different stages of a research and writing process.

  • Using chatbots can also lead you astray and be a significant source of error: 

    • Chatbots can hallucinate, i.e. make up answers that might sound plausible, but are factual/historically incorrect. 
    • Chatbots sometimes provide biased answers (because their responses rely on the material they are trained on, and (opaque) algorithms that control the calculations.
    • Chatbots cannot think, feel or remember.
      Never attribute human-like qualities to chatbots, even if they sound a lot like a human when communicating. 
    • Chatbots are generally bad at explaining and assessing the nature and adequacy of data used in a given analysis; to account for and assess the validity and adequacy of the research method and to account for and assess value bias and interests – in other words, crucial and essential features of academic work.

 

What are the rules?

At SDU, there are rules in connection to the use of generative AI is allowed in examination and when it is not. You can find these rules on this page: 

Any material from generative AI included (either directly or in edited form) in an assignment as if it were one’s own work, without disclosure or indication, is considered plagiarism. Such an offence will be treated on an equal footing with other forms of plagiarism (opens in a new tab).
As a general principle, the reader must be able to ascertain how generative AI has contributed to the final product. 

If you have used generative AI in your assignment in a way that is not described below, you must always clearly explain why and how generative AI has been used at a relevant point in the assignment (e.g. method sections, footnotes, index, etc.). If in doubt, you can always include prompts (or prompts and answers) as appendices.

  • The prompt is the text or instruction you enter into generative AI tools to achieve the desired response. It can be a simple prompt of, say, one line, or an advanced prompt that describes in detail what response the chatbot should generate. 
    The prompt also sets the tone and direction of the conversation with the chatbot, which can ensure that the chatbot delivers relevant responses. The method of constructing prompts is called prompt engineering.

Here are some examples of how to make it clear to the reader or reviewer of your work that you have used generative AI:

In academic contexts, generative AI is not perceived as an authoritative source. It is therefore not recommended to use generative AI to generate finished text for academic assignments, unless the assignment requires it (e.g. assignments dealing with generative AI or prompt engineering).

If you nevertheless insert text generated by generative AI into your assignment, it must be clearly declared:

  • If you insert passages generated by generative AI directly into the assignment, they must be cited as a citation, and you must cite that generative AI on the list of references. 
    Note: this also applies if you have used AI-generated notes that are included in the assignment. 
    Read more about citation of generative AI here (in Danish). 
  • If you insert passages in which you have paraphrased generative AI output, or if you insert passages in which generative AI has paraphrased the original text, you must cite generative AI.
    Read more about citation of generative AI here (in Danish).

Explain at a relevant point the assignment (e.g. method sections, footnotes, indexes, etc.) how generative AI has helped organise chapters and sections or other elements of the assignment.

Your programme may require additional documentation.

Describe at a relevant point in the assignment (e.g. method sections, footnotes, indexes, etc.) how generative AI has condensed large data sets. Use source references to document which original data has been summarised.

Your programme may require additional documentation.

At a relevant point in the assignment (e.g. method sections, footnotes, indexes, etc.), include a description of the analyses and interpretations that generative AI has performed, and how they are included in your assignment. Use source references to document which data has been analysed and/or interpreted.

Your programme may require additional documentation.

Explain at a relevant point in the assignment (e.g. method sections, footnotes, indexes, etc.) how generative AI has helped generate new ideas and perspectives.

Describe at a relevant point in the assignment (e.g. method sections, footnotes, indexes, etc.), how generative AI was used to support your assignment, e.g. by identifying relevant sources, literature and materials.

Include examples of search strings and criteria you have used with generative AI.

Describe at a relevant point in the assignment (e.g. method sections, footnotes, indexes, etc.) that generative AI has been used for proofreading (grammar, spelling and punctuation).

If generative AI is used to contribute to the linguistic presentation of your assignment (e.g. the correct use of terminology, wording, linguistic adjustments, feedback on language, translation, etc.), this must be explained at a relevant point in the assignment.

Include the generated multimedia content in the assignment or as appendices (large files) and describe at a relevant point in the assignment (e.g. method sections, footnotes, indexes, etc.) how generative AI was used. 
Explain at a relevant point in the assignment (e.g. method sections, footnotes, indexes, etc.) how generative AI was used to work with programming. 


The Academic Round Table

A guide to your proces with writing academic assignments:  

Academic practice

The library has made this guide:  

Little by little, an increasing number of existing software programs, such as Microsoft Office 365, Adobe Photoshop, etc. are gaining generative AI functionality, and completely new programs have been developed and are being developed based on the technology, which specialises in images, video, transcription work, data processing, etc.

SDU has made a number of generative AI tools available. SDU does not prohibit the use of any programs, but you are responsible for complying with the security guidelines.

Here are examples of tools that SDU makes available:

Copilot

Your SDU Microsoft Office account gives you access to Microsoft's Copilot (external page opens in a new tab). 
Copilot is built on the same language model as ChatGPT combined with Microsoft’s Bing search engine. Find more information here: Copilot chat (external page opens in a new tab). 

Scite.ai

  • The library provides access to the literature search tool Scite.ai, powered by ChatGPT 4 (external page opens in a new tab). Access is available to SDU students and staff as well as staff at the university hospitals in the Region of Southern Denmark and at OUH.
  • The two primary functions of Scite are: 1) you can search for a topic, title or the like, and thus get a list of relevant references as well as data on how many times the listed references are cited and whether they are cited positively or more critically. 2) Scite has an assistant similar to the mainstream generative AI tools; you can ask it questions and seek knowledge. Scite’s responses are composed from extracts of research publications published with a DOI number.
  • To learn more about Scite, read Nicholson et al.’s article from Quantitative Science Studies (2021) (external page opens i a new tab).

What are chatbots and generative AI

Generative AI is a type of artificial intelligence that is designed to generate new content elements, such as text, code, images, music, videos, etc. (Stanford University, Nature, 2016).

Chatbots are computer programs based on generative AI – colloquially referred to as ‘generative language models’ or simply ‘artificial intelligence’. 

A chatbot’s output is based on a series of statistical analyses based on probability calculations. Due to the advanced structure, computing power and large data sets of the language model, it can produce very precise, contextual output in the form of text, images, sound, video, tables, code, etc.

Generative chatbots come in many varieties and are offered by many different companies. Among the most widely used general chatbots are Copilot, ChatGPT, Gemini, Claude and Mistrall.

There are also a number of chatbots on the market that are trained on a more specialised and limited material, such as research papers, and trained for more focused tasks (including literature search and data processing).