6. Use of Generative AI

If you are planning to use generative AI in your thesis work, keep in mind your responsibilities as a student:

Being accountable and transparent about your work
Investigating when using generative AI is appropriate
Conducting research with academic integrity
Reflecting on when/why generative AI is being used
Refraining from using it to replace learning

In essence, the use of generative AI is permitted only where it supports, but does not replace, the student’s own learning, critical thinking, and original academic contribution.

For instance, this would include students employing GAI to still undertake the following:

Verifying and source-checking: Students must independently verify all factual claims, references, and data produced by generative AI tools using reliable academic sources. AI-generated citations should never be used without confirming their existence and accuracy.

Being academically rigorous: Generative AI may assist in identifying relevant literature, but students are expected to read, interpret, and cite original academic sources themselves. AI-generated summaries cannot replace engagement with primary literature.

Your own intellectual contribution: Generative AI tools cannot be credited as authors. The thesis must reflect the student’s own intellectual contribution, including argumentation, interpretation, and critical analysis.

Using GAI for the technical but not analytical: AI tools may assist with technical implementation (e.g., coding, statistical syntax), but the choice of methods, interpretation of results, and conclusions must be made by the student.

Documentation and reproducibility: When AI tools are used in analysis or content generation, students should retain and, where appropriate, provide documentation of prompts, workflows, and AI-assisted steps sufficient to ensure transparency and reproducibility.

Critical Understanding: Students should critically assess potential biases in AI-generated outputs, including cultural, linguistic, or disciplinary biases, and ensure balanced and academically sound perspectives.

Ensuring Integrity: AI-generated figures, images, or data must not misrepresent empirical findings. Any generated visual or synthetic data must be clearly labeled and justified.

Awareness of consequences: Misuse of generative AI, including failure to disclose or presenting AI-generated content as one’s own, may constitute academic misconduct and be subject to institutional penalties.

General guidelines

The following general guidelines should be followed on how generative artificial intelligence can be applied responsibly when working on your thesis.

Students are responsible for everything included in the thesis
AI as only an aid
Ethics and privacy
Limitations and risks of relying on AI-generated content
Statement and references using AI tools
Referencing when using AI-generated content

These guidelines are expanded upon below:

1. Students are responsible for everything included in the thesis.

You are responsible for everything included in the master's thesis; this applies to text, data, graphics, illustrations, code, analyses, etc. AI-generated text should not be included verbatim as part of the thesis. Any use of AI-generated content must be substantially revised, critically assessed, and fully integrated into the student’s own work. And if AI tools are used in the work on the thesis, you are still responsible for all submitted material.

2. AI as only an aid

AI tools can be helpful in various parts of the work on the master's thesis, such as generating ideas, structuring content, organising data, etc. It is permissible to use such tools as an aid in the work, but it must be emphasized that this is only an aid – you are responsible for assessing how tools and results can be used in a responsible and good way, and for ensuring a sufficient degree of critical reflection in all phases of the work. (See Risks, above)

3. Ethics and privacy

Use of AI tools must always be done in an ethical manner and in accordance with research ethics guidelines. Confidential information and personal data should never be shared with AI tools. This applies to both own information and personal data collected in connection with the work on the master's thesis.

4. Limitations and risks of relying on AI-generated content

Generative AI may provide incorrect or false information, contain text from sources that cannot be included without reference, and which may thereby lead to plagiarism, and/or provide unbalanced, non-academic, or incomplete analyses. It is always the user who is responsible for ensuring that everything that is written is correct and in accordance with norms of research ethics. Pay particular attention to the risk of including text that comes from other sources and which would thereby be considered plagiarism without proper source references.

5. Statement and references using AI tools

If AI tools are used in the work on the thesis, this must be clearly stated in the thesis. Such information shall consist of two parts:

5.1. A statement on the use of AI tools when working on the thesis

This should be placed in conjunction with the reference list (see the templates for master theses) and have the following content:

AI tools that have been used in the work on this master's thesis:

- - Name (and version) of the AI tool: e.g. ChatGPT, 3.5 (or 4.0). (Write “None” if you have not used any AI-tools.)
  - The purpose of the use of the tool: e.g. used to organize data, data analysis, idea generation, etc. If multiple tools have been used, they should all be listed in the same way. (If you are in doubt about which tools to include, it is better to include too much rather than too little.)
  - The list is followed by this text: [I am/we are] aware that [I am/we are] responsible for all content of this master's thesis, including the parts where AI tools are used. [I am/we are] responsible for ensuring that the thesis complies with ethical rules for privacy and publication.

5.2 Referencing when using AI-generated content

Several referencing styles have now included templates for referencing AI-generated content. Recommendations regarding how to refer to AI-generated content when using the APA reference style can be seen here for more detail. But, in short, referencing should be as follows:

	In the text	In the reference list
Rule	(Company, year)	Company (year). Title of software (version) [Explanation], link.
Example	(Sikt, 2023)	Sikt (2023). Sikt KI-chat (version GPT-4) [Large language model], https://ki-chat.sikt.no/

6. The role and responsibilities of the supervisor

The supervisor can be a good help for the students in assessing permitted and good use of AI-based tools, and it is strongly recommended to discuss such questions with your supervisor. However, it is the students' responsibility to follow these guidelines; the supervisor can give advice but cannot take responsibility for compliance with the guidelines.

Risk levels

The different uses of generative AI can be divided into three risk categories, highlighting the increasing need for prudence and care as AI-use becomes a detriment to learning and original research.

Low-to-moderate risk: AI being used to aid in doing the work rather than to eliminate important steps of the process, thereby supporting the work the student is already doing (e.g. improving the quality of writing or expanding the search for relevant literature)
Moderate-to-high risk: AI being used to produce work the student is expected to do themselves, to the detriment of the student’s input, learning, and ownership over the thesis.
Very high risk: AI being used in a manner that is not permissible.

The table below offers examples for each of these categories.

Risk level	Low-to-moderate	Moderate-to-high	Very high
Use of GAI	GAI as a tool	GAI to generate content for one’s own use	GAI to generate content and presenting it as one’s own work
Examples	Proofreading: Spell check, grammar check Editing: Lightly rephrasing objectives in grant writing or tightening language in a draft paragraph, while remaining cautious of not losing one’s voice Translation: translation of existing text to another language (e.g. translating one’s English abstract into French). The permissibility of this use is dependent on the field of study and the purpose of the translation Programming: Debugging existing code Search engines: Using generative AI to sift through and identify relevant research articles for a project References: Formatting the reference list	Learning: Using generative AI to break down complex topics or processes, for one’s learning purposes. Literature review: Producing summaries of published research on a given topic for one’s own consumption Programming: Writing programs or generating code for one’s own use that will not be distributed widely Research writing: Generating suggestions for potential reviewers (suggested names may not be the most relevant for a given topic) Research design: Suggesting additional steps/tasks for experimental design (e.g. using generative AI to identify steps for one to consider, adapt and/or modify)	Literature review: Submitting summaries of published research on a given topic as one’s own work Programming: Sharing entire programs or large amounts of code produced by generative AI without proper attribution Research design/writing: Generating objectives, arguments, perspectives, research ideas, figures/images, etc. Research design/writing: Autonomously conducting analyses on a data set, where the student does not understand, justify, and take full responsibility for the methods, results, and interpretations. Research writing: Writing entire sections or the entire proposal/paper Research writing: Generating research quotations

For further information, including tips and signing into Copilot (Microsoft's AI chatbot), please check out LNU's "Guidance and recommendations for teachers on generative AI".

NB: Most content for this page was taken/adapted from McGill University's Guidelines on using generative AI in research and the Norwegian School of Economics' Guidelines for use of AI-tools in the work on master theses...as well as suggestions from ChatGPT (GPT-5.3)!