7
$\begingroup$

I would like to know whether there is some kind of consensus or policy, what to do.

AFAIK such questions are off-topic here, because this is Q&A and not peer review, many posts fall into "check my proof" or "why this is not solution". I have seen comments, that this is not proper place for scientific advancements and in my opinion there is no chance for breakthrough here.

I am puzzled, because questions "where is error in my attempt" seem to be good, but very similar questions with different wording get downvoted and closed. Posts with hidden attempt to prove get closed.

Is there any hard rule? If not, could we create one?

$\endgroup$
  • 2
    $\begingroup$ The closest we have to current policy (that I know of) is here: cs.meta.stackexchange.com/q/109/755, cs.meta.stackexchange.com/q/647/755. Also interesting, from CSTheory.SE: cstheory.meta.stackexchange.com/q/2720/5038, cstheory.meta.stackexchange.com/a/3023/5038. $\endgroup$ – D.W. Apr 26 at 1:01
  • 3
    $\begingroup$ "where is error in my attempt" -- there's a thin and fuzzy line between "check my attempt" (usually close-worthy) and a focused, specific question about an attempt (the best kind of question there is). P-vs-NP questions usually fall into the former category -- if they are not straight-up peer review requests -- because the askers don't know where their idea is fragile. If they know enough about likely gaps, they are probably able to ask a better question that doesn't mention p-vs-np at all, because they are able to narrow down their question to a detail in a much longer proof attempt. $\endgroup$ – Raphael Apr 26 at 16:21
  • $\begingroup$ FWIW, we've had a canned comment for some flavors of such questions around for a while. Should it be adapted or point to other discussions? $\endgroup$ – Raphael Apr 26 at 23:42
  • $\begingroup$ @Raphael thank you, I personaly think it should be adapted, and comment more widely used (but it isn't). The former discussion pointed in canned comment is to crank heavy topics, but currently it seems that presentation of the question or level of refutation (if it is easy, just answer, if harder then close) is more important than scope, when we take last several questions in this topic. To me it is like saying that Astronomy is off-topic here, but since it asks about distance to the Moon, we could answer it anyway, hence this post. $\endgroup$ – Evil Apr 27 at 1:12
6
$\begingroup$

First of all, if we would like to have a specific rule against such questions, I don't think it is useful to make the topic "P vs NP" part of that rule. I don't believe we are being flooded by questions about that topic and I don't think that that topic is inherently 'problematic', nor that it is within the domain of research. (That is, questions about the open question (and how to deal with the fact that we don't have an answer) are central to the field of complexity, which permeates to all sorts of 'CS experts')

I think I agree mostly with this answer, although that has bit of a different focus. However, the following two reasons to distinguish good from bad questions on this topic mentioned there are relevant:

  1. they are honestly phrased as people trying to understand what they are missing,

  2. they are rather short, have a few points, and are easy to answer for an expert in complexity theory.

I think point 2 is key here. There's quite a difference between someone asking to check their 40 page proof versus someone wondering why an elementary proof of 4 lines doesn't solve this open problem.

Point 1 seems more a case of etiquette than a meaningful property of the question. So what if someone really their crisp and clear 4 line proof of P=NP is correct? As long as the question is clear and they behave themselves (in particular when proven wrong), I do not see why this makes the question bad.

A more relevant criterion, that also focuses more on the question, rather than on the user is

  1. the question contains a cleanly phrased, well-formed proof based on standard terminology

I think this covers all cases of questions we do not want that are excluded by point 1. Point 1 seems to be aimed at 'cranks' that are prone to long, vague proofs littered with (often undefined) non-standard terminology. The long part is already covered by point 2, the vague and non-standard terminology is covered here.

... this is not peer review, [...] this is [the] not proper place for scientific advancements and in my opinion there is no chance for breakthrough here.

Is this not the proper place for scientific advancements? Well..., why not? But more to the point, I do not think 'scientific advancements' is the value of the good questions about these topics. The value is that these wrong proofs often make assumptions that are unfounded, but seem very intuitive. (the mother of all assumptions is of course, "clearly, we cannot do better than brute force" such as in this question) It just so happens to be the case that the central questions in complexity have remained open problems for a long while, despite appearing relatively simple. It is good for e.g. a student to be aware of that.

Put more poignantly, the error in these proofs is often educational, even for others than the one that has written the proof.

I am puzzled, because questions "where is error in my attempt" seem to be good, but very similar questions with different wording get downvoted and closed. Posts with hidden attempt to prove get closed.

I understand your confusion, allowing questions on (complete) P vs. NP proofs seems counter to the policy of not checking peoples answer. I think that point 3 here denotes a clear enough boundary here. If the proof itself is incomprehensible, poorly structured, or has other significant problems unrelated to the subject matter, then this person is better served by improving their basic proving skills before looking into proofs of this complicated matter. These are off-topic, and should be closed.

Note that also, even if the question literally asks to check their proof, it should be interpreted as 'find the mistake'. This is different from grading an assignment, even if we are guaranteed to have a mistake somewhere. Why? An assignment is a measure of the writers competence in their field. An error in an assignment is basically everything that shows incompetence (In practice, grading schemes are of course more lenient and require the student to only show competence on a few well-defined points). We are not here to judge/grade someone's competence. But I see no reason to disallow questions about erroneous arguments for that reason.


To clarify my viewpoint, let me give a few examples.

Naive argument that P ≠ NP: this question gets a full score, passing all three points. I think this is a good question and clearly so.

Contradiction proof for inequality of P and NP: this one passes 2 and 3, but it is not clear to me whether it passes 1. This is one of the reasons that I prefer to judge only on points 2 and 3. I think this question is fine and well within the lines, although not as clear as the previous one. (The difference in votes is better explained by the HNQ exposure of this question, rather than the inherent quality)

Solving Diophantine equations: does having a bound on the size of the solution help?: Note that this question has been rather radically edited to solve various issues. In its original form, this is a question that I would consider a bad one. It is long, and littered with irrelevant or mostly unhelpful digressions. On the surface, it fails all points, and in its original form, it should have been closed. Still, it is based on an interesting assumption, which is not clearly false, but unproven (and likely even harder than resolving P vs. NP). This is not obvious at first, however. Unless someone is able to 'extract' a good question out of these, these questions should be closed. While it is nice if someone does this, we have no obligation to the author to do so.


In summary, I think that asking about such proofs should be acceptable, if

  1. The proof is short and has only a few points; and
  2. The proof is cleanly phrased, well-formed and based on standard terminology.
  3. The question has no other problems that would apply to any type of question. See e.g. here or the help center for more on what to do and what to avoid when asking a question.
$\endgroup$
  • $\begingroup$ Ok, this is very elaborated and nice answer, what should we do now, to ensure that this is reference policy? I mean, for now there are two agreeing people. $\endgroup$ – Evil Apr 26 at 9:29
  • 1
    $\begingroup$ @Evil First, we should wait until other people show their agreement, disagreement or anything else they want to add. If e.g. there is a single answer with high votes, we can put a summary in front of the post, for ease of reference. If there is no clear consensus, then we can resort to voting, which should be done in another meta thread. But for now, we wait. $\endgroup$ – Discrete lizard Apr 26 at 9:34
  • $\begingroup$ The former discussion was independent of topic, this one exactly asks for such insight like level. If you could propose policy that would directly say "dependent of level do..." it would be perfect. $\endgroup$ – Evil Apr 27 at 1:16
-3
$\begingroup$

thx for asking about one of my favorite CS topics of all time, but alas also at times a controversial "lightning rod". there is huge interest in the P vs NP problem by both amateurs and professionals but alas its been described as a "crank magnet" topic, but then theres also some concern on SE about avoiding that characterization as not tactful. somewhat involved in that theres a $1M prize by Claymath institute that has gone unclaimed for nearly 2 decades now.

as for site policy, my feeling, and others have expressed similar sentiments, is this site could function positively as an occasional peer review system for papers on the subject. there is a history of 2 attempts that went viral in cyberspace over 2 decades: Deolalikar and Blum, and some experts have had favorable response to those review "efforts". the emphasis would be on trying to educate people in answers in a professional/ public relations/ pop science way why particular attempts typically fail and try to focus more on "higher quality" attempts. think there is value in this as a public service instead of shunning all such effort.

suggest that we use upvotes or downvotes to express enthusiasm or unenthusiasm respectively for the specific proof attempt in the question but not immediately close questions on the subject unless the attacks/ attempts are of very low quality. its also helpful to refer to the papers as attacks or attempts or claimed proofs instead of proofs. suggest that questions refer to published papers and not try to be self-contained proofs. and everyone can start to understand that advancement of science is a collaborative process that involves peer review and in the 21st century sometimes cyberspace systems/ social media can play a positive role in that effort.

$\endgroup$
  • $\begingroup$ So your policy is to maintain hard references and discard self-claimed solutions? $\endgroup$ – Evil Apr 27 at 1:18
  • $\begingroup$ eg am picturing encouraging site members to learn latex or other math typesetting language, write up their claims, incl diagrams, try to communicate clearly as possible, learn prior work in area, cite relevant references, use notation/ terminology correctly, etc as in science (as found in some undergrad programs eg senior theses, project writeups etc) and then "return the favor" with some expert attn/ feedback to such work. sort of like the advisor-student relationship. and moreover, maybe not try to solve the problem but instead make some kind of nearby contribution in the area. etc $\endgroup$ – vzn Apr 27 at 2:01
  • 5
    $\begingroup$ I don't think SE is a suitable platform for peer reviewing large papers, or any argument that spans more than a page. Users are of course free to organize something like this, but I don't think this would make a suitable question, unless the 'key' of the argument or an specific short part could be extracted first. $\endgroup$ – Discrete lizard Apr 27 at 18:01
  • $\begingroup$ It essentially goes to square one, close on random, from time to time indulge advisor-student discussion, which is not statutory goal of Q&A. $\endgroup$ – Evil May 1 at 1:55
  • $\begingroup$ DL afaik academic peer reviews in professional journals are typically not very long even for long papers. my idea is to "find the 1st (few) mixed up/ glaring issue(s)" in any paper and not try to disentangle/ deconstruct the entire thing piece-by-piece. peer review is a crucial element of science. for the record heres the mini review of the 2017 blum attack on Theoretical Computer Science 233 votes Is Norbert Blum's 2017 proof that P≠NP correct? (ofc "high power" attacks such as this are rare) $\endgroup$ – vzn May 1 at 15:26
  • 1
    $\begingroup$ @vzn Reviews are usually not that long, but the main problem is that this the content of the question (which is long) is mostly off-site. So the question isn't self-contained, unless it already picks at a particular part of the proof. In my opinion, Is Norbert Blum's 2017 proof that P≠NP correct? is not a good fit for this platform, while Tardos Function Counterexample to Blum's P≠NP Claim is, as it stabs at a particular part of the proof. (or rather, a refutation of that part) $\endgroup$ – Discrete lizard May 2 at 11:31

You must log in to answer this question.

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