-1

I have to choose a certain number of data points. The exact number has little effect on my outcome -- it could be as little as 10 or as high as 100. There is no good way to choose this number, and no factor limits me from going higher.

Since 42 is a popular number, I plan on using it. The findings from this will be submitted to a conference for peer review.

My questions:

  1. Do people often use the number 42 in research?
  2. Will this (mostly) be viewed positively or negatively?
  3. Is the number popularly accepted as the "answer to everything," or could I come across a reviewer who has a strong opposing viewpoint, and therefore lead to a biased review for me?
Improve this question
9
  • 2
    42 is the default seed in most if not all my code. More of a habit than anything, and I don't mention the seed value in a paper. I wouldn't think anything of it if you were saying you used 42 as a value. Tbh, it's so common now, I wouldn't realise
    – JackRed
    Sep 25 at 6:22
  • 5
    If it's truly irrelevant, you may use it. But even as someone who really enjoys science culture, I would be massively put off if it were important but chosen for memey reasons.
    – rhombidodecahedron
    Sep 25 at 7:32
  • 21
    "The number (my choice = 42) is the number of data points required for my experiment's result to be reliable. Yes, the number could be as little as 10, but there is no way to assess if 10 would be a better choice than 42, or vice versa." That doesn't make any sense. The number of data points required for a reliable experimental result is not arbitrary. You need to justify it, e.g., by analyzing statistical power.
    – Roland
    Sep 25 at 7:52
  • 1
    @Roland In my case, the number of data points required depends on the underlying distribution of gpt-3.5. As I do not have access to it, I have no way to measure my statistical power.
    – Priyanshul Govil
    Sep 25 at 13:08
  • 1
    @PriyanshulGovil I'm not an expert in AI. But your last statement is certainly incorrect. At the least, you can analyze how increasing n impacts results.
    – Roland
    Sep 26 at 5:18

3 Answers 3

Reset to default
5

One possible concern is to avoid suspects that you chose that specific number because it works better and makes your method look good.

Someone might think that you tested all numbers between 10 and 100, and then you only reported the results for n=42 because they are in favor of your method, while most other numbers would give a worse result.

For this reason, I would rather opt for a simpler-looking number like 50 or 100. If you want to stick with 42, you can include a sentence that the results are similar for all other choices.

But this is a secondary concern, at least in my field; people do not care too much about it.

See also https://en.wikipedia.org/wiki/Nothing-up-my-sleeve_number in crytpography.

Improve this answer
6

Hmmm ...

The exact number has little effect on my outcome

That seems extremely unlikely. How can that be?

it could be as little as 10 or as high as 100. There is no good way to choose this number, and no factor limits me from going higher.

If you are doing any kind of modeling, or statistics (even taking a mean) then there IS a good way of choosing a number. It's called power analysis. And if no factor limits you from going higher (which is also unlikely) then you should go higher.

Do people often use the number 42 in research?

I've never seen it used and I've read lots of articles in many fields (I do statistical reviews for a couple of journals, and I worked as a consultant in lots of areas). But your field may be different.

Will this (mostly) be viewed positively or negatively?

For people who "get it" (those who have read Hitchhiker's Guide) it will be viewed as a joke. For those who don't get it (not everyone will) it will just seem weird.

Academic presentations and articles can be funny. But a) Usually, such papers are from pretty well-known researchers and b) The humor has to serve the purpose of the paper.

Improve this answer
1
  • 4
    I sometimes use it, even here, for things peripheral to the actual argument, or as an illustration, where a joke is fine. But for the actual research, I doubt that people "getting the joke" is a good thing.
    – Buffy
    Sep 25 at 12:14
5

Different fields have different customs. That said, in my experience this style of humor (memes, in-jokes, pop culture references, etc.) is extremely rare in formal publications. Honestly it's rare for research papers to be funny at all, and papers that are funny are usually intended entirely as jokes.

Your idea is liable to give the impression -- perhaps unfairly -- that your research is not worth taking seriously. I'd recommend avoiding it entirely.

I've seen more of it in conference talks. If you're new to academia, I'd recommend you play it straight at first, and not try to be funny. Once you have a few conferences and publications under your belt, you'll get more of a sense of what kinds of humor are typical in your field.

Improve this answer
0

Not the answer you're looking for? Browse other questions tagged .