I was offered an opportunity to prepare a 2-page research proposal for a postgraduate research program in computer science.

I am searching for a systematic and step by step technique to select/find a workable research topic.

Is there any systematic procedure/strategy/approach/method that researchers generally use to select/find and narrow down a research topic from an ocean of topics that pops in one's mind?

Is this technique an standard in the academia?

  • 1
    Best thing to do is finding something in the frontier of the state of the art where you can make some progress, not too many people are trying to do so and has some economic interest. E.g. a problem not seen to open, an open problem to close or a limitation to solve. PD: actually, I'm going to make this an answer...
    – Trylks
    Commented Aug 11, 2014 at 18:55
  • 5
    If there were a way to automate this, it wouldn't be legitimate research. Commented Aug 11, 2014 at 19:37
  • 4
    I think that most people just find something that looks like fun, go forth, and do science. Or am I doing it wrong?
    – Moriarty
    Commented Aug 11, 2014 at 19:50
  • 1
    @paulgarrett systematic and automatic are two different things.
    – Trylks
    Commented Aug 11, 2014 at 23:28
  • @Trylks, I understand your point, but/and the question (in my opinion) touches upon the "desire to do research" while not knowing much at all. It's not just about "desire to do research", or "following up on one's ideas". Many people have thought about many things already. Scholarship plays a huge role in competent, useful research, and scholarship is not a quickly-acquired thing. Sensibilities, context, awareness, ... not systematizable except as "study, reflect". That's what I mean to say. Commented Aug 11, 2014 at 23:47

2 Answers 2


I don't know whether it is standard. there may be some standards, but probably not worldwide. I personally had to write a 2-page research proposal on a topic I didn't choose.

IMHO, it's a great opportunity to choose a topic that you can like, if the topic was imposed (or changed after a few years by your supervisor) you would probably hate that, since you have this opportunity, I'd try to make the best use of it.

Systematic approach to select/find a research topic (by trylks):

  1. Make a list of the areas that interest you
  2. Make a systematic literature review (so the approach is systematic) for those areas
  3. Find research problems in the literature review and at least one that:

    1. has not been researched yet (it's "open")
    2. should be done next. It's in the frontier of the state of the art

      a) Without many previous requirements or they will become your thesis

      b) it's interesting from a research perspective

    3. not too many people are working on it (best is zero, but that could be for some reason...). The problem if there are too many people (and you are not a part of their team(s)) is that the frontier of the state of the art may keep moving before you are able to reach it.

    4. you have the necessary knowledge and skills to advance the state of the art in that problem
    5. you have the necessary resources to advance the state of the art in that problem (e.g. don't choose something that requires access to the data from LHC if you won't have access to that data)
    6. There is some economic interest in the problem and the results of your thesis (you don't need this for the thesis, but it will make everything much easier, specially after the thesis)

WRT the frontier of the state of the art, it usually looks like:

a) Some limitation in current systems/techniques that has not been addressed (specially in engineering)

b) Some question that remains open (specially in science)

c) Some question that has not been addressed (in some particular way) (specially in philosophy)

But well, research in those areas is more than that and it will be done in different ways depending on the topic. Computer science is sometimes considered an engineering (software engineering), sometimes a science (and can be empirical with benchmarks or formal with proofs) and it can be as well philosophical (specially in AI, IMHO). The kind of problem should arise from the review of the literature.

BTW: don't try to rush, this is systematic, but it will definitively take very long.

  • Perhaps a sane suggestion is to get advice from an experienced, trustworthy person. Otherwise, one's very perceptions of the supposed issues may be far from the actual substance. Intuition for what is importnat, what is possible, what is impossible, although not "rules", are very, very useful guides, etc. Inexperienced people especially suffer from failing to understand the potential benefits of experience, etc. Commented Aug 12, 2014 at 0:43
  • Indeed. Almost everything is easier with a bit of help from someone experienced. In this case I see two main possibilities, either "a priori" to find your way in an ocean of papers or "a posteriori" to check those two pages and whether they make any actual sense.
    – Trylks
    Commented Aug 12, 2014 at 1:27
  • I'd be concerned with this approach that unless the area(s) chosen in Step 1 are extremely narrow, the following steps will involve iteration over many thousands of published papers. And choosing a sufficiently narrow yet workable area in Step 1 is almost tantamount to choosing a research topic. Commented Aug 13, 2014 at 23:04
  • @NateEldredge the topics can be narrowed down during the literature review, i.e. I'm interested in machine learning, so I check the titles of the papers in main 3-4 conferences in machine learning in the last year. Then based on the titles I continue with the abstracts and based on that I continue with the rest of the paper, if needed. If I'm interested in machine learning and I find a paper in that area that is clearly not interesting (for me, based on the title) then checking that paper in more depth can probably be safely and indefinitely postponed.
    – Trylks
    Commented Aug 14, 2014 at 0:50

I'm not an experienced graduate student or post-grad (yet, to my great consternation), but here's my somewhat-educated guess to expand on Trylks point 2 above, and others, and create a short answer:

A. Do a 2-level review of Review Journals. These are the literature that review the state-of-the-art research done in a prior period, such as the last year. Do a high-level scan to get your bearings and then a deeper-level dig, as you might guess. Here's one for your field, e.g., that was published up until 2011: http://en.wikipedia.org/wiki/Annual_Review_of_Information_Science_and_Technology

B. Combine that with the expert help of people senior to you in the field, as others have suggested, to find gaps in the field.

C. Pick an area in the gaps that interests you in which you can make a contribution appropriate to your level of expertise/skills, and go.

You must log in to answer this question.