I was wondering if (let's say for example) someone come out with a certain problem found in a certain domain. The problem is divided into subproblems. Each of them is given as a stackexchange question. The combination of solutions will give us a solution to our big problem. This solution can be then published in a journal or conference.

Was there something done like this before ? -- What problems would occur from such collaboration ? Would the paper be accepted if it offers an innovation ?

What if someone has already a review and wanted to be reviewed by stackexhange users before conference submission ?

  • $\begingroup$ If the questions are suitable for SE separately then the fact that you are attacking a bigger problem using them as subproblems would not make them unsuitable, and if they are not that fact will not make them suitable for SE. (But be aware of the policy about crank-friendly topics.) $\endgroup$
    – Kaveh
    Nov 11 '12 at 5:27
  • 1
    $\begingroup$ ps: You may want to check this meta thread on MO. $\endgroup$
    – Kaveh
    Nov 11 '12 at 5:28
  • $\begingroup$ I don't see a problem with doing this. We have our own cstheory meta thread on original results proved on the site, and there's at least one paper that came out of discussions on the site. $\endgroup$ Nov 11 '12 at 6:31
  • 1
    $\begingroup$ there are some stackexchange responses that could be used possibly as lemmas or research leads in a larger paper. however, the stack exchange medium is inherently different than papers and its interesting to contrast the two as you have. papers require much more effort for previously unresolved questions and stackexchange excels at short answers to resolved questions. there is some disconnected talk about using stackexchange for a polymath-like attack but it hasnt caught on. polymath is the best model for cyber collaboration invented so far. $\endgroup$
    – vzn
    Nov 12 '12 at 17:56
  • 1
    $\begingroup$ @vzn: There have been plenty of previously unresolved questions answered on this site. It's actually pretty good at it, what it is not good for is answering famously unresolved questions, because these tend to be much harder than the garden variety unresolved question. $\endgroup$ Nov 15 '12 at 7:47
  • 3
    $\begingroup$ @AJed: I imagine some people would get very annoyed if you wrote a paper centred around their answer without them as an author. $\endgroup$ Nov 15 '12 at 7:48
  • 1
    $\begingroup$ 1. Polymath is usually lead by a leading expert in the field, and is assumed to have definite progress. 2. The major problems are unsolved, probably, because there is no clear way to attack it. 3. Although probably many approaches have been used to attack them, there is no resource discussing approaches in a single place, nor giving intuition why people assume certain approach is not good. 4. The community wiki here is a good place to review them, and may actually be a good place where particular problems/questions can arise, and be solved through the main site, with updates in community wiki. $\endgroup$
    – mkatkov
    Nov 15 '12 at 8:59
  • $\begingroup$ ok, concede poor phrasing in the prior limited comment. its more about the scale of the result. tcs.se is indeed being used for (sometimes novel) "bite-size" results. papers are for "chunk-size" results. but lets face it, even while highly prized by some, & a definite priority, new research results on here are neither big or common. & have seen at least one prominent person (once highly active, now not) complain in meta that sophisticated new results taking some work maybe dont get much votes. agree with that, have experienced it myself! research is hard, stackexchange doesnt change that... $\endgroup$
    – vzn
    Nov 15 '12 at 20:13
  • $\begingroup$ @mkatkov agree this is a useful medium for some actual research approaches but its limitations should be understood/recognized... what do you mean re polymath "is assumed to have definite progress"? imho there are fairly clear ways/paths to attack all open problems, even the very hard ones, what is not clear is whether those ways will ever succeed... stackexchange has some definitely research- and collaborative- unfriendly features eg the ease with which comments/questions/answers can be deleted/vaporized into ether in numerous ways! $\endgroup$
    – vzn
    Nov 16 '12 at 16:42
  • $\begingroup$ @mkatkov, a question can be made CW only by a moderator and there is a reason for that: some users were abusing CW to post non-suitable questions. In principle what you suggest can be useful to be done in some place (e.g. Richard Lipton and Ken Regan do so on their blog), but that doesn't mean such discussions need to take place on cstheory. Putting aside other issues, I think on cstheory it will be mostly abused by non-experts who want to draw attention to their ideas/opinions and that is not a good use of cstheory. $\endgroup$
    – Kaveh
    Nov 17 '12 at 19:44
  • 1
    $\begingroup$ [cont.] cstheory is mainly for researchers to get help with their own research, not for non-experts to check their ideas/opinions about how to solve/attack famous long-standing open problems in a topic they do not have even basic knowledgeable/familiarity. $\endgroup$
    – Kaveh
    Nov 17 '12 at 19:48
  • $\begingroup$ @Kaveh. So let's say I have an issue with a proof of a theorem. I did well in most of the proof, but I found an issue at some little point. Is it ok for you to post this problem as a stackexchange ? if i get a solution and use it my paper, should I use the stackexchange contributer as a co-author ? or reference ? or wat exactly ? $\endgroup$
    – AJed
    Nov 17 '12 at 22:29
  • 1
    $\begingroup$ You seem to be asking about ethics of publishing and acknowledgement, not cstheory. I think this was answered somewhere on Academia. There was also a meta post here about what should one do if one finds an error in a paper. $\endgroup$
    – Kaveh
    Nov 17 '12 at 22:38
  • $\begingroup$ @vzn I think you well answered to you question re polymath. My understanding of intention of polymath is following. Post problems there that have clear path of attack with very high chances to succeed. And projects there lead by very credible researches exactly by this reason - it is believed there are very high chances they are right in their intuition. $\endgroup$
    – mkatkov
    Nov 18 '12 at 9:35
  • 8
    $\begingroup$ Is there a way to define our own set of badges on this site, and have the moderators award them for off-site activity? Because I think "publish a research paper based on a cstheory.stackexchange question and answer" should be worth a gold badge, "write a cstheory.stackexchange question or answer that is cited by a published research paper" should be worth a silver badge, and "write a research paper that cites a cstheory.stackexchange question or answer" should be worth a bronze badge. But maybe that conflicts with the pseudonymity of some of the identities here... $\endgroup$ Nov 24 '12 at 5:55

The short answer: yes. (It's not actually published in a peer-reviewed venue yet but who cares.)

  • $\begingroup$ +1 out of respect, but it doesnt really answer key parts of the question & is not really matching the scenario sketched out.... $\endgroup$
    – vzn
    Mar 16 '13 at 2:49

will take a stab at answering this based on comments & culture here & sketch something out that can be used as a point for further discussion because its a direction/approach Ive been thinking about & looking into & even veering towards myself & its key to this group.

this question is probably a little too hypothetical. in principle you could build a paper by stringing together some stackexchange answers but it would be unusual. as mod kaveh writes, each question must stand on its own separate merit and also note, few people regularly post questions here or with high frequency esp those with high rep, and there is some possibility that frequent questions are actually informally frowned on via the voting mechanisms & current collective voting tendencies.

the pattern is more in general that high rep users answer random newcomer questions. also, it is unlikely that you would get solid answers to all the separate (presumably nontrivial) questions.

basically yes, that is part of the raison d'etre of the group, to occasionally collaboratively generate scientific research to theoretical computer science in general & complexity theory in particular (complexity theory entries even were once highlighted here in css formatting). however the conventions around this are new & still being established as we speak.

1st, use the stackexchange search engine highly extensively and also the tag system, both can be very effective. find every relevant post to your paper or subject. figure out the best knowledge that is posted on the group. figure out the border between the known and unknown and what is posted vs not posted. this also requires close familiarity with nearby papers. it helps to bookmark or "favorite" all related posts on here & elsewhere. use the related posts column on the right also. mix this with heavy cross use of google. in some cases you may find an expert on here posting on the subj. you can look at their own answers & look at their papers.

next, try posting a single carefully crafted question totally centered on your subject and your paper and what you're thinking of establishing. its a delicate mix between the general (background) and the specific (the new result/concept you're aiming for). as if you have a single chance to ask the experts what you'd most want to know but what you also most think they'd be able to answer.

responses will help you figure out what is known by the crowd but also how relevant it is to this group based on reaction in form of comments, enthusiasm, voting, etc. it has to be at a "just right difficulty" between too easy [not research] and too hard [you wont get replies/answers]. it takes some time to find that balance & its an ongoing challenge. dont assume that if there is a good answer, someone will post it. dont assume that if people post answers, those are the best possible.

you will prob find if you are very focused that there are two gaps. one is the gap between what you want to establish in your paper and what is established in related literature. the other gap is what you can actually find out by using this group (both by searching it and posting questions). these two gaps are interrelated. you can judge whether you can build on all this with meaningful new results. but those key results will basically have to be generated by you, nobody else, unless you have found offline collaborators in your own professional circles. you can use other results as a "jumping off" point.

so finally, leverage everything youve found and write a paper. if the answers on stackexchange are very notable and innovative, you can cite them in your paper using the stackexchange urls/"cite" feature. this is really not much different than citing blogs, which is occasionally done in real scientific papers. however note that this is somewhat problematic because content on here can disappear very easily due to voting, moderators, users [who can delete stuff] etc...

probably the best and most successful model of all this invented so far is polymath[3] with a difficult problem and published arxiv paper to its credit but keep in mind this is a project started by a field medalist Gowers and promoted by his friends eg Tao and Kalai who are also very elite/accomplished theorists.

[1] How will you cite a discussion on this site in your paper?

[2] How can we collaboratively investigate open problems?

[3] Massively collaborative mathematics/Polymath

[4] Highlights of cstheory.stackexchange.com

[5] What interesting open problems have been asked on this stackexchange?

[6] Attacking Open Questions

[7] References to cstheory in the literature

[8] Original proofs generated on the parent site

[9] Questions answered. in theory. (early tcs.se ad & highlights reel for the beta period by mod Suresh)

[10] How to ask a good question


You must log in to answer this question.

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