MG as your master thesis supervisor?

Drawing by Paolo Ghelfi


This page is for the students who are considering having Michele Govenatori (MG) as their master thesis supervisor

What the aim of a dissertation is

Writing an essay, be it a literature review or an empirical research, requires adopting critical thinking based on the knowledge acquired throughout your study career so far.

Although in some cases the final essay might be useful as a reference to further analysis or applications, its rationale is checking the student’s ability of critical, consistent, creative thinking.

Using AI to write an essay

Large Language Models (LLMs) are and will be increasingly available. They can be an effective tool for literature research and even for drafting or correcting a text. However, in my experience the only way of mastering such tools is knowing more than they know, at least in terms of general understanding of main trends and implications.

Although LLMs are quickly getting better, they still tend to write in a shallow, optimistic, predictable, recursive, bombastic way. Phrases like “pivotal importance” sound to me as typical benchmarks of use of LLM. It looks like current LLMs (at least the cheapest ones) tend somehow to imitate the way lousy human authors write: repeating clichés to avoid risks.

I cannot do anything but give low grades to essays suspected of being written with LLMs unless I have substantial evidence of critical or creative thinking when interacting with the student. Sadly, such interaction is typically very limited (see next chapter).

In simpler terms: why should I allocate my time in reading your text if you delegated a machine to write it?

How much time MG can allocate to your thesis supervision

Very little. I am sorry to write this, but I get way more supervision requests than I can manage in terms of accurate interaction. (Not to mention the co-supervisory jobs). I am grateful for the trust, but I must be honest: you will need to be extremely autonomous.

Here is a list of things I cannot afford providing to a supervised student:

1) Recursive text reading. I can only read once and make remarks when I see relevant issues. I expect issues to be dealt in deep and critically by the student by amending the text and addressing my points.

2) Comprehensive editing. I will only signal once any category of issues. It is up to you to apply the suggestions analogically throughout the whole work.

3) Comprehensive literature check. I cannot verify all citations. Beware some LLM make up literature. I do not mind how many papers you mention. I mind how meaningfully you do so. 

Automatic checks on authenticity and non-LLM use can fail. I will base the assessment on my (limited, fallible) experience. The more you write content or style clichés, the more you are recursive, obvious, generic, the higher the risk you get a low grade, including by mistake. So simply don’t: be creative, be different, use your own language, make choices, take risks.

The research question (RQ): set it, follow it.

No essay is suitable to test your proficiency if your research question (RQ) is not clearly and precisely set at the beginning. You should allocate to such preliminary job more effort than you might think. If the RQ is vague or too broad, it is also ineffective.

Please do not let me read long introductory chapters. Go straight to the point. There is plenty of introductory stuff everywhere. The more I get bored in reading, the lower my grade proposal to your graduation commission will be.