# What shall we do with stone soup?

Sometimes, a question is asked that isn't up to the standards of MathOverflow. Maybe it's worded vaguely, maybe it's too open ended, or maybe the asker fundamentally misunderstood something important.

Nevertheless, every so often other users on MathOverflow will "make stone soup": take the unpromising question, and answering it in an unexpectedly interesting way.

What should we do in this situation?

These questions sometimes prompt cycles of closing and reopening --- some people want the question closed (it's not that great a question after all!) and others want to preserve the bad question, for the sake of the good answers is has received or will receive. I think it would be good to have some general guidelines, and hopefully avoid that process.

• Feel free to make points either way; hopefully however we can come up with some constructive suggestions for people on both sides of this debate, and write a 'best practices' answer that I'll accept, then mark this as a FAQ. – Scott Morrison Oct 17 '13 at 4:20
• Another aspect to consider is that such questions are often clarified after criticism but, unfortunately, such clarifications often end up making existing answers, no matter how good, irrelevant to the updated question. Not only is this upsetting for the user who took the time to answer the wrong question, it also makes the answer less likely to be discovered. – François G. Dorais Oct 17 '13 at 4:37
• For reference: en.wikipedia.org/wiki/Stone_soup – Joel Reyes Noche Oct 17 '13 at 5:33
• Scott it is not useful to refer to "not that great questions." We do not close questions for not being great but for being bad. – Gil Kalai Oct 17 '13 at 7:02
• Good point, Gil. (I think, like many from Commonwealth countries, I have an unfortunate tendency to damn with faint praise.) I'll edit. – Scott Morrison Oct 17 '13 at 8:33
• To complement François's comment: it also often happens that the question remains unclarified because, to all appearances, the questioner has left the scene. Whether this is because he or she never really was that interested to begin with, or got frightened by all the ensuing commotion, is hard to say. But then the question becomes something of a Rorschach: answerers have to guess about intent and motivation. In such circumstances, maybe CW (Community Stone Soup?) is the way to go, since there seems to be no one "right answer". – Todd Trimble Oct 17 '13 at 14:58
• The debate is idle since a long time and in view of your first comment I comment as a reminder. – user9072 Feb 12 '14 at 15:36

I'd propose the rule of active editing: if you value a stone soup question, then edit the question to make it better (clearer, broader, better thought out, more answerable) and ensure that it fits the answers. You might lose the original intent of the question, but that horse has already left the barn. By the time we reach the stone soup phase, there's nothing we can do to refocus the existing answers on the original intent, even if the person who asked the question clarifies, so we might as well change the question. If they were hoping for something else, then they can ask another, clearer question. (I'm assuming the OP has seemingly abandoned this question. If they are participating actively and trying to clarify their intentions, then we should not ignore their wishes.)

I see two arguments in favor of editing:

1. The broken windows theory of MO. We want sensible, well thought out questions and relevant answers, while stone soup questions are often just the opposite. For example, you may have a naive question that's unanswerable or trivial as written but on an interesting topic, with a discussion of this topic in the answers. In that case it's better to clean up the question than to leave visitors with the impression that this is what MO is for.

2. Discoverability. If the answers have drifted from the original question or broadened it, then rewriting the question can make the answers easier to find, which is the whole point of posting them on this site. In particular we should avoid question roulette, where you can't predict what sort of answers a question might have until you read them.

Is there any reason not to do this? If one detects the issue at an early stage (before stone soup), then of course it's appropriate to seek clarification from the OP, but I don't see a compelling argument for leaving stone soup questions unchanged forever.

• I worry this encourages people to manipulate questions for the purpose of giving self-aggrandizing answers. By and large I think we need to try to respect the will of the question-asker. Modifications to questions not done by the author should only be to correct basic features to ensure the question is understood: spelling, tags and such. – Ryan Budney Oct 17 '13 at 14:42
• I'm not too worried about deliberate manipulation, since it could be undone, although maybe it could be more of a problem than I foresee. Regarding respecting the will of the question-asker, I think this is an important principle if they are actively participating and clarifying their intentions. However, if someone shows up out of nowhere to ask a question anonymously and then remains silent for several days, I'm not so worried about this. I'll edit my answer to clarify. – Henry Cohn Oct 17 '13 at 14:59
• While in principle I share some of @RyanBudney's concern, I still think that Henry Cohn's suggestion is a good one for the cases where it is too late to take another route. We will not manage to remove the question from the site once it has sufficiently much good content below it in the answer-boxes (and maybe we should not try), so better it is a good or at least alright question than something else. [cont.] – user9072 Oct 17 '13 at 15:14
• But re concern, a worry of me are the 'broken windows' or rather those that break them. While one should focus on the q and not the users asking them there is a risk that some asking the q in the 'stone soup' thread (and possibly others in addition) get away with the idea that their question that was nothing but some ordinary (or even ugly) stone actually was delicious and will offer us ever more of them. I am not sure if this encouraging effect is stronger or weaker when there is substantive editing; likely depends on questioner. (I do not give examples but not for lack of them.) – user9072 Oct 17 '13 at 15:18
• I very much like this suggestion. To help manage Ryan's concern, I think it's good to also add a comment, to the effect "I've substantially revised this question; feel free to revert or change further if you feel appropriate, or start a discussion on meta." – Scott Morrison Oct 17 '13 at 20:43
• For reference: en.wikipedia.org/wiki/Broken_windows – Joel Reyes Noche Oct 18 '13 at 3:20
• "In that case it's better to clean up the question than to leave visitors with the impression that this is what MO is for." This is a kind of common rhetoric on MO but it is fairly unsound. These "visitors" who may visit some MO question and will get a wrong impression, is largely an imaginary notion. – Gil Kalai Oct 18 '13 at 18:04
• @GilKalai we had cases where it was explicitly said by users asking problematic (in some people's opinions) questions that they found some question on the site via a search and then extrapolated from this question what should also be acceptable (just the thing they saw was not really acceptable to begin with). A keyword to help you remember an example "maiden name." And precedences are invoked all the time. So this exists. I would be curious how you can assert it is an 'imaginary notion.' We might not know how common it is and your guess might differ from others, but it exists for a fact. – user9072 Oct 18 '13 at 21:21
• @Ryan Budney I am new here. If I am allowed to put my penny in, I will write a few words. I was quite enthusiastic in my first MO days. Then I faced the case. I admit some users probably committed a kind of fraud, but my euphoria gone away mostly because a priory rude attitude to an OP. – Sasha Anan'in Oct 20 '13 at 19:11

There are plenty of questions that it is uncontroversial that they need to be closed. However, let me propose

(Joel David Hamkins's proposed rule, with quid's addition) Give all questions charitable readings, when doing so leads to interesting mathematical ideas. Make sure, however, that all questions you answer are somewhat in line with the guidelines 'How to write a good MathOverflow question?' (possibly via editing them by yourself)

My view is also that in cases of controversy one should make an effort to keep the question open for the sake of users who are interested. Be aware that a question need not be good for all the community; it is often enough that it is good to some of the community. I also propose to avoid hastily closing questions that attract good answers or that people explicitly endorse. Please note that good answers can make a good question (also outside MO.)

• Regarding the last paragraph: why not agree on respecting the guidelines meta.mathoverflow.net/questions/882/… at least only ever so slightly. Would also save a lot of time arguing. – user9072 Oct 17 '13 at 12:48
• Quid, Sure, I read again the guidelines and they are good. We can add it to Joel's rule: Be charitable and tolerant to others and stricter (but not overly so) on yourself. – Gil Kalai Oct 18 '13 at 6:01
• Since you did not make a precise suggestion what to add to the rule here is one: "Give all questions charitable readings, when doing so leads to interesting mathematical ideas. However, make sure that all questions you answer are somewhat in line with the guidelines 'How to write a good MathOverflow question?' (possibly via editing them by yourself)." – user9072 Oct 18 '13 at 10:41
• Ok, how is it now? Regarding the guideline I will not insist on the OP provide always motivation and background. It is generally a good guideline but there are also various justified reasons people may avoid it, and short questions also have some charm. – Gil Kalai Oct 18 '13 at 16:47
• In other words, if you already have a stone soup, eat the soup and let the stone rest in piece on the bottom of the pot ignoring occasional curious passer-byes who wonder what the stone is doing there. I agree. – fedja Oct 18 '13 at 18:55
• Thanks for the additions. I still feel you pay too little attention to the interests of those that find certain uses of MO disruptive and problematic. Specifically, 2 and 4 as well as 3 and 5 are somewhat in opposition, and then 1 trumps everything to push through your extremely permissive ideas. – user9072 Oct 18 '13 at 20:56
• Gil, this is the third or so time I saw you using 1) instead of 1. for the numbered list. If you use the latter then the software will automatically make this into a numbered list with appropriate spacing and whatnot. – Asaf Karagila Oct 18 '13 at 21:12
• Quid, what was the point to suggest modifying my third point (which I did by adding 4. and 5.) if you feel that my first point trumps everything and promote my "extremely permissive ideas?" Apparently, we simply disagree and we can leave it at that. – Gil Kalai Oct 19 '13 at 1:16
• "A question need not be good for all the community it is enough that it is good to some of the community." I am a bit confused now, Gil. Is there ever a case where a question can be justifiably closed, in your opinion? (Note I am not arguing for or against your suggestion, I genuinely do not understand how to interpret your first point.) – Andrés E. Caicedo Oct 19 '13 at 16:55
• Thank you for now taking in the modification in 3. Personally I might prefer something slightly stricter in the general tone, but also if people would really follow it, it were alright in my opnion. Thus I revoke my downvote and deleted two obsolte comments. Perhaps one could add a remark that 1. cannot override the principle that there must be no content that could (reasonably) be perceived as offensive. (Perhaps this is anyway understood as obvious.) – user9072 Oct 19 '13 at 17:31
• I think this is quite close to what would be desirable. I would suggest having 3 as the main point, really (don't you think?), and explaining somewhat what you mean in 1 by "some of the community". – Andrés E. Caicedo Oct 19 '13 at 18:29
• Andres, the cases of controversial closing are fairly rare. (I suppose that I regard more than 90% of the closing justifiable.) Cases of controversial closing are also quite anticipated, so people can often realize that a certain problem that they regard uninteresting and even unsuitable to MO is appreciated by others. In any case, if people actively express the opinion that the question is good, or supply good answers to the question this should be a strong signal against closing. – Gil Kalai Oct 20 '13 at 2:32
• OK, that clarifies the meaning, thanks. I like this suggestion, as I said (and again, maybe having 3 be the key point would even strengthen it). – Andrés E. Caicedo Oct 20 '13 at 3:09
• Quid, Andres, what about this relaxed version: 1) In case of controversy, make an effort to keep the question open for the sake of users who are interested. A question need not be good for all the community; it is often enough that it is good to some of the community. 2) Avoid hastily closing questions that attract good answers or that people explicitly endorse. Good answers can sometime make a good question (also outside MO.) – Gil Kalai Oct 20 '13 at 5:55
• I like this wording. But do not lose 3). My current rule of thumb is that even if I don't like a question personally, I will try to keep it open or vote to reopen (even if originally I voted to close) if good answers appear or are hinted at. It ends up being rewarding, so why not? – Andrés E. Caicedo Oct 20 '13 at 6:29

Note: This answer refers to an earlier version of Gil Kalai's answer and does not really make sense for the current version. I preserved it "as is" to keep the developement of opinions visible; as written in an old comment, I essentially agree with the current version of his answer.

The points made by Gil Kalai seem very good, but then they seem based on a wide spread misconception, namely that answering a question as is and closing it forever are the only possibilities.

Yes, of course, let us be charitable in reading question, but why not confirm before answering via a quick comment conversation whether ones charitable reading even is what the questioner was after?

Fine, good answers perhaps make a good question, but why not still trying to improve the question in the process in addition? Won't this make the package still better?

So why is this done so relatively little?

Perhaps because it is a lot of work without much gratification. By contrast writing some high-level essay vaguely related to a couple of sentence the so-called 'question' is fun. One can write on a subject one likes, show-off a bit ones knowledge and typically such general interest material is very well received.

Don't get me wrong I did/do this on occasion, too. So, if you (here I mean an abstract 'you') do it sometimes, too, I won't blame you. But at least don't fool yourself into thinking you are most helpful in doing so.

The actually helpful thing to do, and better for the quality of the site, is to work together with the questioner to figure out first what they are actually after and to answer only then.

• I think a reason for this misconception is that there's a significant percentage of mathematicians (this is a seat-of-the-pants computation) that are afraid of discussions that lean towards motivation and intent. Asking people what exactly they're interested in and why -- what is the greater purpose -- these questions can be frightening, especially to a grad student that takes it as an axiom that certain things are important. Enough mathematicians have a hard time explaining why their own research important on their grant applications. – Ryan Budney Oct 17 '13 at 12:33
• This is more of a math-sociology rant now but I also suspect that the evolution towards "moralistic language" in mathematics, when you see questions like "what is the right formalism for..." are symptoms of more or less the same thing. Some people explain what they mean by "right" but far too often people won't, or can't. – Ryan Budney Oct 17 '13 at 12:35
• Sure, I see no conflict between my suggestions above and also trying to improve the question itself: by editing it directly or by gently (and in a friendly way) guiding the OP toward improving his own question. One thing I don't like is trying to figure out the precise level of researcher the OP is or nudging him in a pressing or patronizing way. I dont think I ever wrote an answer just vaguely related to a few sentences of the question. But it is true that when you write an answer you don't aim just to the OP but to a wider audience of people who may be interested. – Gil Kalai Oct 18 '13 at 5:03
• @GilKalai from the top of my head I do not have an example of an answer of the form I have in mind that you gave and it might be pointless or even counter-productive to search for one now. (The closest I could think of now, but this was more joke/playful so is still in a different category is your answer where you mixed two diacritics on an o). However, there are many such examples (whether you contribute one or not is not really relevant), and Ryan Budney, Henry Cohn, François G. Dorais, Todd Trimble all mention in some form the same phenomenon in this thread. (c.) – user9072 Oct 18 '13 at 10:50
• (c.) Also, what I am talking about has nothing much to do with figuring out some precise level of OP but merely with insisting that there is an actual question being asked at all that then gets answered, as opposed to people writing short essays inspired by the keywords in the question; an example would be mathoverflow.net/revisions/101420/1 (I link to the original version of q, that thankfully got somewhat improved). But there are many more. And re "patronizing" I do not like this either but then tp guide the asker can be useful. (c.) – user9072 Oct 18 '13 at 11:01
• (c.) But on the subject of "patronizing": what annoys me considerably in some of these debates on questions actually is also certain patronizing attitude of some in these discussions, namely those in favor in some questions talking like those against the question (in its current form) are just against it as they do not get it. (Added: I originally gave two examples, but then deleted them [for now].) – user9072 Oct 18 '13 at 11:22
• One more remark on the 'aim just to the OP': as always it is a question of balance, and the question is what this means in practice. And again this is not an issue that only I observe, let me quote Greg Kuperberg from a long time ago who it seems saw some problem in this direction too, while actually I saw less of a problem then, but well I was rather new on the site then: – user9072 Oct 18 '13 at 15:42
• Quoting Greg Kuperberg (emphasis mine, source middle of thread): "[...] I have always thought that the ratio of score for accepting an answer to getting an upvote, which is only 3/2, is ungenerous. I think that the ratio should be at least 2; and if it had been 3, I would never have objected to that either. (I know that the OP can also upvote, but still.) Anton once made fun of me for wanting higher scores in general, but really that's not it. Rather, this ratio encourages people to play to the crowd more than to help people with questions." – user9072 Oct 18 '13 at 15:43
• Quid, Regarding the music question: This is a question that the short unedited version was, in my opinion, much better then the long edited version (with the motivation). – Gil Kalai Oct 18 '13 at 16:59
• @GilKalai you do not know any exmples of it? The ABC question was completely void of any actual content and just contained some bloated pseudo-informed prose. It was nothing but a dirty stone, best ignored or kicked away. "Tell me something around the recent (claimed) proof of ABC" would have at least been a clean pot, but what we had orginally was worse. Still some people provide some information related to it. But the question(er) contribute nothing at all to it. And we had closed some (or at least one) question just like it before that also could have played the role of the stone. – user9072 Oct 18 '13 at 18:07
• The edited ABC question is good. And the answers are superb! Overall, this is one of the most important successes in the history of MO. (And it was a pity that there was a close/open war in this case.) The question of the exact credit to the questioner is secondary. (But he was at the right place at the right time.) If there were other questions on other issues that could have similar impact but that were closed this is very unfortunate. – Gil Kalai Oct 18 '13 at 18:40
• @GilKalai Oh, so the edited version is good. (And please note the 'what we had originally' in my post.) You do realize that a main point of this entire debate is how to proceed in such case and the proposals by Henry Cohn and me, against which you are arguing a bit, is precisely to encourage such editing. Also, if one rereads the meta thread one might come to the conclusion that this editing was a consequence of the critical reception of the original version. (As I said there were other users that asked or meant to ask an analogous question, so right time right place not even this.) – user9072 Oct 18 '13 at 20:20
• Quid, for ABC, my view is that the editing was fine and closing the question was not. Generally speaking I dont think that improving asking the OP to improve the question should be carried out by the action or threat of closing the question. – Gil Kalai Oct 19 '13 at 1:20
• This is veering way off the main topic, but since this ABC question keeps getting brought up (not just here), I want to express my disagreement with quid (and apparently many others). I think the original version of the question was fine. The OP was asking for a description of Mochizuki's main idea at a particular (rough) level of abstraction. To demonstrate the level of abstraction he was looking for, he gave an example. His example invoked the Weil conjectures. Some interpreted this as showing off. I interpreted it as helping us understand what he was looking for --- and quite usefully. – Steven Landsburg Oct 19 '13 at 18:21
• @StevenLandsburg I disagree (unsurprisingly) and I could and would go to great length over this, as this question does not live in isolation but OP was well-known (as MO user) to me. If you [or somebody else] want this, please, open another thread here or perhaps still better on "tea," if you want a quick idea search "tea" for 'ELIZA' (and let me add, while hard to reconstruct now if you do not know, OP of the relevant "tea" thread knows very well what they are talking about mathematically there, a lot better than me). – user9072 Oct 19 '13 at 21:56