Uploaded on 2025 June 7th
Amended on 2025 Dec. 5th

2025 Dec. 5th: I've decided that I'm mostly going to be writing for myself. If it turns out it's useful for others as well, so be it. This may sound dramatic but mostly what I've done is use "Knowledge distillation" instead of the (admittedly) more fitting "Information distillation", simply because that's what I originally wanted to write. I have also amended the conclusion part 'cause I think I sounded a bit condescending originally and I don't want to be. :)

I've been wanting to write an article with this title for a while - mostly because Knowledge Distillation sounds cool, I admit. It also sounds arcane a bit, like I'm an alchemist trying to create a Philosopher's Stone which will be used along with a couple of dozen brains to create the Juice of Knowledge^TM. Yum.

The first part, the one that makes this sound like this article is the first in a series of articles comes from the fact that at the moment I'm part of the "Professional Services" division at my company. When I get to take my first look at a questionnaire, it's already been deemed "Ready for Professional Services" and assigned to me. So because I'm approx. five years old, I always think:

That is me.
I am Professional Services.
Allow me to service this questionnaire professionally.

Hehehe

Now, this article is supposed to be about Knowledge Distillation - what do I mean by that? I mean that in the above-mentioned questionnaires I sometimes encounter situations or instructions that seem a bit overcomplicated, too wordy. Some are justifiably complicated, because they're describing a complicated situation. Other times I get the feeling that multiple people worked on the same questionnaire, but not necessarily at the same time or with the same information, or with not enough time to get to know the questionnaire.

To better illustrate what I mean, you can find a short excerpt from an imaginary study below.

• ...

• QGEN. What do you identify as?
  Select one

  • 1. Male
  • 2. Female
  • 3. Trans Male
  • 4. Trans Female
  • 5. Non-binary
  • 6. Other

• QCHILD. Do you have children?
  Select one

  • 1. Yes
  • 2. No, but I'm pregnant P.N. Don't show for QGEN options 1, 3, 4, 5, 6
  • 3. No

• ...

What happened in this imaginary study is that new QGEN options kept being added, and they kept being added to the instructions at QCHILD as well. You can note that the above instructions are still correct, but they can be significantly shortened if we switch to a positive affirmation:
P.N. Don't Show only for QGEN option 2

I realise that this might seem like a trivial improvement to make, but that's the point. Someone, somewhere (or someones, somewheres) was rushed and just updated the questionnaire with the first thing that came to mind. Again, it's not a bad thing per se, but unfortunately in order to lessen the mental load in that exact moment, it increased the mental load later on.

I also realise that this crops up more in my line of work than in traditional software engineering. That doesn't mean that it can't help others, though. :) It's just that instead of getting ready-made logic instructions, you will be responsible for the logic of whatever software you're working on. In which case - take a step back, look out the window, swear at passers-by, whatever. Hopefully five minutes of relaxation won't get you fired, because honestly sometimes five extra minutes is all you need to spare you an hour later on.