LLM use policy

These rules apply to all theses with Jakub Krukar acting as the first supervisor and all written assignments within courses taught by Jakub Krukar.

I encourage students to use modern LLMs as tools that support learning, but not replace it. They can help you brainstorm, plan the structure of your work, and navigate large amounts of literature - provided you have actually read and understood the sources you eventually cite. I recommend this method for reading larger amounts of research literature.

However, most students misuse LLMs that leads to multiple shallow, nonsensical, or incorrect phrasings. Therefore, submitting a long text with good quality of English but poor logic or shallow explanation does not guarantee a passing grade. Be careful, because this is what you get "by default" from LLM chatbots. Such texts may look impressive to you, especially if you haven't engaged deeper with the topic of the course, but the quality of the academic argument is often mediocre or unacceptable for a Master-level course. In fact, the average grades of written assignments in my courses seem to have decreased since the widespread availability of LLMs.

The core requirement in all written assignments and theses is this: you must understand, justify, and defend everything you submit. In depth engagement with the topic, and independent thinking are valued over length and sophisticated prose. When in doubt, I may ask you to explain selected paragraphs in person.

Permitted and encouraged use of LLMs in my courses and theses:

  • brainstorming, outlining, structuring your argument
  • exploring literature and identifying relevant papers
  • using LLMs to clarify concepts that you have problems understanding (but don't fully trust the results without verifying at source, e.g. with RAG)
  • improving the fluency of the written text, trying different styles, or connections between arguments

 

Discouraged or unacceptable use:

  • generating long passages of text and submitting them without deep revision
  • relying on LLMs to construct arguments you cannot explain
  • using unchecked or unread references
  • submitting generic, vague, or technically shallow text

 

Quality criteria - Your grade depends on:

  • rigour and conceptual precision
  • depth of understanding
  • correctness of information and references
  • clear connection to course material and in-class discussions
  • originality and independent thought
  • density of ideas relative to the amount of text submitted (higher grade for submitting brief but technically precise text; much worse grade for long shallow texts)

 

Red flags (automatic fail):

  • hallucinated references or mis-citing a reference that is not supporting the given argument
  • logical incoherence between separate fragments of the text
  • arguments that are technically incorrect or cannot be defended
  • lack of understanding of your own submission, e.g. using words or phrases that you cannot explain

 

Required documentation

For each written assignment and thesis that you are submitting as a file (e.g., as a pdf), you are required to attach the LLM Use Checklist generated via the form available here. Your answers to these questions do not affect the grade. Neither does the fact of using or not using LLMs. You do not need to attach the Checklist for those assignments that are submitted inside a text box in Learnweb but the points listed above still apply to the grading policy.

How to write a good MSc thesis at SPARC

A good thesis is dense with content, clearly structured, and reproducible. For this reason, I recommend submitting your thesis in the format of a journal article. This saves you some writing but forces you to be precise and concise. Specific requirements in this thesis format are:

  • Formatted according to the guidelines of the Journal of Enviornmental Psychology (if your thesis contains an experiment with human participants), International Journal of Human-Computer Interaction (if it focuses on evaluating the usability of some user-centred technology), or Environment and Planning B (if it focuses on developing new environmental metrics).
  • All data collected from your study is published online (on GitHub, GitLab, or Open Science Framework), kept tidy, in open file format (such as .csv), and described in a README (what each column means, were there any outliers excluded from the analyses, what other researchers need to know in order to re-use your data).
  • All material used to generate the experiment (e.g., questionnaires, apps, VR environments, videos, and other stimuli shown to participant) is published online together with the data.
  • The oral defense of your thesis must take place one month prior to the submission deadline or later. If your research includes an experiment with human participants, you must complete all data collection and statistical analyses before your defense. While your thesis may be in draft form at the time of defense, it is essential that the key findings are established and that all data and materials are accessible in an online appendix. If there is no institute-wide defense date that meets these requirements, please contact me to arrange an individual defense date.

Open thesis topics

Bachelor


Master




Assigned thesis topics

There are no assigned Bachelor thesis topics to show right now.

There are no assigned Master thesis topics to show right now.





Completed thesis topics

There are no completed Bachelor thesis topics to show right now.

There are no completed Master thesis topics to show right now.