Trying to do the right things in academia

Grad students often tries to do the right thing, but sometimes the right thing is unclear. Such things are still hard to find online even in 2013, so here is a collection of things I've encountered.

Submitting the same paper to multiple classes or to a class and a conference

I've done this several times as a student and seen it done many times as a TA (my field is computer science). There is a big debate about this matter at the NYtimes. here and discussions It was not mentioned there that one good thing to do is to check with the prof teaching the class and see if it is okay. In one case, I checked my research adviser who also taught the class, and he said submitting my research to the class is definitely the thing I should do.

While TAing a relatively advanced class, I found this to be quite common and most people submitted their research to the class. At least one did not even bother shorten his write up to the class specification and just submitted a conference paper he was working on. The instructor just responded by giving a still good grade that is slightly worse than if the work has been done solely for the class. I know of one other similar case in another class. When a student just used his research project on some different data, the instructor objected and gave the student a slightly lower grade. One cannot go wrong with checking first, but this is expected to be generally acceptable for grad students. I've never had the chance to this in my first 3 years of undergrad.

Correcting mistakes in submitted papers

Days after submitting a paper, a mistake is found in the paper. This also seems to happen quite often to computer scientists who often submit to conferences under time pressure. In one case, we tried to write the conference chair with a correction. The particular mistake is in a table in the supplementary material. It is very straightforward to correct and it's unlikely for the reviewers to notice. The conference chair responded saying that no changes will be accepted after the submission deadline as it would be unfair to others who do not change their papers afterwards.

Build on unpublished work

Recently, I had the need to build on some unpublished work. My adviser outlined 3 options:

  • Do not acknowledge the unpublished work, and repeat the necessary background in the new paper. This is fine if the necessary background does not take long to repeat. The advantage is the new paper can be read standalone, and we still have anonymity. The unpublished work would still be cited in the final version, but not citing might be acceptable for submission. Still, this approach can be unethical if the reviewer did not realize the unpublished background material is not the contribution of the new paper. So one needs to be quite explicit.
  • Attach the unpublished work as supplementary material. This wold technically work for careful reviewers who really want to get to the bottom. But otherwise, the reviewer might just ignore it, fail to really understand the new paper and then make an arbitrary decision.
  • Put the unpublished work on arXiv. This is not frequently done in the more applied areas of CS, almost totally breaks anonymity, and the reviewer still probably won't read it carefully.

So we ended up writing it so the paper can be read standalone by a knowledgeable reviewer, and put the rest on arXiv. It's still technically anonymous, but there really is no anonymity if the reviewer felt like spending 2 minutes finding out who the authors are.

Including new results in rebuttal

Dealing with questionable baselines