Academic Life2 min read

What I look for when supervising a first research project

After supervising one doctoral and three master's theses, the thing I select for is not the thing I expected.

ByÖnder Eyecioğlu

I have supervised one PhD and three master's theses to completion, plus a long tail of undergraduate projects. When a student asks to work with me, I am not really evaluating what they think I am evaluating.

What I am not looking for#

Grades. They tell me a student can do assigned work with a known answer. Research has neither of those properties.

Prior knowledge of my field. The literature can be read in a semester. Whatever a first-year student knows about quantum machine learning today will be partly obsolete by the time they submit.

Confidence. I have watched confident students stall for a year because they could not say "I don't understand this."

What I am actually looking for#

Tolerance for a null result#

Most experiments do not work. A student who can run an experiment, get nothing, and report it accurately — without inflating it, and without being crushed by it — will finish. A student who needs every experiment to succeed will start unconsciously steering their analysis, and I will spend the next two years checking their work rather than reading it.

The best answer I have ever heard to "how did the experiment go?" was:

"It didn't work, and I think I know why, and I think the reason is more interesting than the thing I was trying to show."

The ability to say "I don't know"#

Precisely and without embarrassment. There is an enormous difference between "I don't understand the paper" and "I don't understand why they use a Trotter decomposition in equation 7." The second is a research question. The first is a mood.

Something they finished#

Anything. A game, a mod, a small library, a bicycle rebuilt from parts. I do not care what it is. Finishing is a skill, it is rarer than talent, and it does not appear on a transcript.

The one thing I promise in return#

That I will tell them when their work is not good enough, early, while there is still time to fix it — and that I will be specific about why.

A supervisor who is kind at month six and brutal at month twenty-three has not been kind. They have been comfortable.