8

I (try to) do research in Computer Science/Deep Learning. I recently did some work developing a novel idea and am writing a paper on it. I have only written one paper before, and I'm currently being paralysed by the related works section.

It seems impossible for me to have a thorough understanding of everything that is related. When I started the work, I was working from the understanding I had at the time. The vast majority of the influence on my work was only from ~4 other papers, and what I took from them as background in the area.

In doing the work, I have come to understand other work and the context much better (I find that working on a problem, and attempting to explain my work, demystifies a lot of statements in other people's papers). Now, reading other papers on other related work is more plausible for me. I'm realising that I've made some assumptions based on those core papers, and want to find citations for them.

This appears backwards to how it's supposed to work; the papers that I would cite in this manner I hadn't read or understood before I started working. So, my main question is: is it illegitimate to cite several papers you hadn't read before completing the majority of the experiments?

Follow up questions are: Is it normal to only have a few main influences, and many lesser influences? (Computer-science-specific) How much of a paper's theory/background is worked out after making it work? Is it my fault if I haven't come across some important related idea, or is it generally accepted that you might miss something?

  • 3
    Often you don't really understand just what you did until you write the paper, leading to just this problem as you dig deeper. Yes, if you find more material of greater relevance you should reference it. And, as you go forward it becomes easier since you now have a better grasp of the literature in your field. That first paper is hard because you are still figuring it out. – Jon Custer Apr 17 at 14:31
  • 1
    Please ask only one question. The later questions are mostly opinion based, and would be closed. – Anonymous Physicist Apr 18 at 7:42
  • Citing something that you haven't actually read should be done through an "indirect citation". ( academia.stackexchange.com/a/7776/1622 ; academia.stackexchange.com/a/45374/1622 ) If you go back and actually read the article so you're not relying on a secondary source ... then you should cite both of them directly. – Joe Apr 29 at 13:52
11

An important realization I had while writing the first few papers of my career is that a paper/thesis/any kind of publication should not be a recollection of your path through this research. You will have made many experiments as part of the research that need not actually appear in the paper because they're not important for what you're trying to show.

Instead, a publication is a way for you to explain insights you have gained during this research. In which order you came to these insights, in which you read other papers, or in which you did experiments is -- generally -- not important for the telling of the story.

So, regarding the original question whether one should cite a paper one hadn't read before the research started: it can't be answered with yes/no. Whether a paper should be cited (as asked in the title) has nothing to do with when one read the paper. It has everything to do with whether you think it's a useful reference for readers to have. In the body of the question, there is a variant: whether it's "legitimate", and there the answer clearly is "yes". In particular, we add references to papers all the time that were pointed out by reviewers -- long after the research was concluded for the most part.

| improve this answer | |
  • This does not answer the question. – Anonymous Physicist Apr 18 at 7:38
  • 3
    I think it does: When you read the paper you wonder about citing doesn't matter. What matters is whether that paper can help your reader understand what you're doing. – Wolfgang Bangerth Apr 19 at 2:14
  • 2
    Your comment is a good answer. But your answer does not contain your comment. – Anonymous Physicist Apr 19 at 7:28
  • 2
    As far as I'm concerned it answers the question effectively by getting at something that I didn't ask for explicitly, but should know, and which resolves the question implicitly. My question was predicated on assuming the order was important. – Multihunter Apr 20 at 4:09
  • 1
    Great answer, +1. – 6005 Apr 20 at 16:48
3

Should you cite work you hadn't read before doing the research?

Yes.

You are ethically obligated to cite work that helped you conduct your research. That in no way implies you should not cite work for other reasons.

You should cite every paper that might help your reader. In my opinion, it is also acceptable to cite a relevant paper just because it is traditional to cite it, or because you liked the paper.

| improve this answer | |
2

I strongly believe that one has an obligation to cite "prior art", whether or not one was ignorant of it or not. If one is ignorant of things, they do not directly influence one's work, but to write/publish a paper with no mention of other peoples' work is irresponsible, possibly to the point of being immoral.

Yes, I have heard a few very-well-known mathematicians say, as a joke but not really, that they avoid reading their competitors' papers so they "won't have to cite them".

And, indeed, some people seem to be in some sort of solipsistic mode, in that they feel no obligation of scholarship or helpfulness to other people. Once I myself caught on to the (obvious, really) deficits of this approach, I did indeed start trying to be more conscientious.

(Plus, interestingly, especially in mathematics, an amazing number of ideas were rediscovered over-and-over... in the last 250 years, at least. The supposed "lack of modern " turns out not to mean that people didn't have the ideas!)

So: cite what you used, cite what you looked at and/but didn't use, and cite what you belatedly discovered. I know other peoples' views are different.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.