Use of GenAI in the capstones

UvA policy

For the most current University of Amsterdam (UvA) policies, see the student information page^[1] or, if you want more details, the Policy Framework GenAI in Education ^[2]. There is also an e-learning module for students ^[3].

The UvA recognises the growing role of Generative AI (GenAI) in society and academia. While GenAI can be a valuable tool, its use in your research project and thesis must adhere to principles of academic integrity, transparency, and responsible scholarship. Misuse of GenAI can be considered plagiarism under the Regulations Governing Fraud and Plagiarism for UvA Students ^[4]

General Principles for GenAI Use

• Human in the lead: You, the student, are at all times responsible for the work you submit. The knowledge, insight, and skills demonstrated in your research project and thesis must be your own.
• Critical engagement: GenAI tools are not reliable scientific sources. Always critically evaluate any output from GenAI for factual inaccuracies, non-existent references, biases, and stereotypes. You are responsible for the accuracy and quality of your work, even if GenAI was used as an aid.
• Transparency of use: You must be transparent about any GenAI use and you can expect supervisors, assessors, examiners and MBCS to be transparent about which use is permissible.
• Privacy and Data Security: Do not input any privacy-sensitive data, personal information, or confidential research data into GenAI tools. Many providers are not transparent about data processing, and your input might be stored and used for other purposes, such as training their models. Make use of the UvA-managed tool UvA AI Chat if you are sharing any personal or sensitive data from UvA labs. If in doubt, do not submit the data.
• No obligation to use non-UvA licensed tools: You cannot be required to create a personal account for or purchase a GenAI tool for which the UvA does not have a license.

Transparency

• Explicit permission: You may only use GenAI for a graded assignment or work product if it is explicitly permitted.
• Sharing use cases: If you use GenAI for any assignment, you must be transparent about how and why you applied it. This includes clearly stating:
  • Which GenAI tool(s) you used (e.g., Mistral, Gemini, ChatGPT, UvA AI Chat)
  • For what specific purpose(s) you used the tool(s) (e.g., brainstorming, rephrasing, generating ideas, structuring text, spell-checking, creating practice questions)

For each of the capstones, the requirements for GenAI disclosure are codified in a grading rubric. Note that even if you made no use of GenAI at all, a section disclosing this is mandatory.

Ethical use

The fact that UvA allows you to use GenAI does not exempt you from the responsibility to consider whether it is ethically acceptable to do so. In a recent open letter, researchers (including, notably, many cognitive scientists) expressed concern about the adoption of GenAI tools by higher education ^[5]. It is important to consider what the consequences of GenAI use are for society, the environment, your learning, as well as the integrity of academia.

Ethical sensemaking is based on continuous reflection and discussion and we strongly encourage you to participate in this discussion. If you are thinking about using GenAI in your studies, consider the following questions and discuss any concerns you have with supervisors, peers, teaching staff or anyone else.

• [ ] Is my prompt/query worth the energy cost of this GenAI system?
• [ ] Would I be depriving myself of deep thinking by using GenAI?
• [ ] Can I (and will I) critically engage with the generated output?
• [ ] Could I feel ownership of the work for which I plan to use GenAI?
• [ ] How do I verify whether information from the GenAI system is accurate?
• [ ] Would I be providing sensitive information to a third party? If so, what are the possible consequences of that?
• [ ] Am I becoming dependent on GenAI systems? What are the social and political consequences of such dependency?

Use within the research projects

During the research process

The research projects are by design filled with a broad range of tasks, responsibilities and opportunities to gain skills, knowledge or dispositions. Many of these can in principle be supported by GenAI use. If you are considering such use, ask yourself the following two questions:

• What would I learn if I were to perform this task without using GenAI?
• What would I learn if I were to perform this task while using GenAI?

If you are depriving yourself of any learning by using GenAI, consider whether that particular loss is relevant to the objectives of the research project. For this, you can use the project rubrics as a guide. Your supervisor can also help you with this question.

Once you choose to use GenAI, be transparent about your use case towards your supervisor, so that they can give you feedback on what they consider responsible use and so that they remain aware of your level of performance. You can achieve this transparency by informing them about your GenAI use during a progress meeting or, if you submit any documentation, by including a paragraph on which tool you used for which purpose.

Note that during the research project, much assessment takes place on the basis of your experimental work and performance during the overall research process. This means you should help your supervisor see the process and that includes transparency about GenAI use. Especially if you employ GenAI for uses that you are assessed on in the rubrics, it is important that you inform your supervisor accordingly.

Creating report and presentation

LLMs are masters of linguistic form. This means that you can, in principle, delegate the process of creating a report or presentation to them. This approach, however, bypasses your cognitive engagement with the content of your project. It is through drafting, reflecting, revising and deleting that you think deeply about the meaning of your project and its value to the research body. Indeed, it has been argued that responsible AI use includes only delegating tasks to AI after trying, analysing and deliberating yourself. ^[6].

Permissible use

This is why the only permissible use of GenAI for writing your research project report is:

• Writing sections independently and asking for feedback on your writing in order to learn from the feedback. No LLM output may be pasted into your report and you may not use embedded LLMs in your writing software of choice.
• Supporting your literature research by requesting LLMs to provide an overview of your research topic. This can offer you citations that you can explore further. Note that you can only cite work you have actually read and that your submitted work has you as the author. Any false citation is considered to be fraud.
• Brainstorming ideas for how to visualise data or structure your report, on the basis of your own best effort.

Note that submitting parts of your draft report can constitute a data policy breach. If your project is hosted at UvA, make sure to use UvA AI Chat (https://aichat.uva.nl/chat) for feedback on submitted sections. If you are hosted elsewhere, check the policies of your hosting institute.

Transparency and attribution

Regardless of whether you use GenAI or not, your report should close off with a section declaring your actual GenAI use. That means listing the tool(s) you used, as well as the purpose(s) they served. The grading rubrics for the research project list different levels of performance when disclosing such information. Note that disclosure of GenAI use is also necessary if you did not use GenAI at all.

Use within the Literature Thesis

The Literature Thesis assesses whether you can collect and synthesise existing research findings and use them to address a problem statement or answer a research question. The writing process is to a large extent also the thinking process, and should therefore not be delegated to any GenAI system.

Permissible use

The only permissible use of GenAI for writing your Literature Thesis is:

• Writing sections independently and asking for feedback on your writing in order to learn from the feedback. No LLM output may be pasted into your report and you may not use embedded LLMs in your writing software of choice.
• Refine a problem statement or research question that you yourself wrote.
• Supporting your literature research by requesting LLMs to provide an overview of your research topic. This can offer you citations that you can explore further. Note that you can only cite work you have actually read and that your submitted work has you as the author. Any false citation is considered to be fraud.

Transparency and attribution

If you use GenAI during your writing process, inform your supervisor about such use before your progress meeting. Share the key prompts you used so that you and your supervisor can jointly decide whether your use was responsible.

Regardless of whether you use GenAI or not, your Literature Thesis should close off with a section declaring your actual GenAI use. That means listing the tool(s) you used, as well as the purpose(s) they served. The grading rubric describes different levels of performance for this section. Note that this section is also mandatory if you did not make use of GenAI.

Teacher use of GenAI and AI detectors

Quality assessment depends on the expertise of teachers and may therefore not be automated via LLMs. LLMs can be used, however, to formulate feedback on drafts. This should be done without submitting personal information to the LLM.

As there are no reliable tools to detect the use of GenAI, teachers are kindly asked to refrain from using instruments that promise to recognise GenAI Use. Submission of the final work via the Datanose project page will automatically lead to a similarity-based plagiarism check.

Further resources