MWEs and their wrongly assumed place in the question

Question

Purpose: To improve the user experience of the visitors of TeX.SX (who most likely provide the majority of the traffic), in exchange for the slight detriment of that of the questioners and answerers.

It is a common practice in this community to request MWEs in questions without it. My personal observation is that they are usually taken as the actual question rather than just an example. I think this severely hurts the usefulness of the community as a resource.

Many questions here do not have an answer. The "answers" in them provide solutions targeting only the particular instance in the example, sometimes going completely against the question body and title.

Questioners can use the checkmark to designate which "response" resolved their troubles, not to choose which "answer" answers the question the best. This can create the illusion of answered questions, although in reality they are resolved problems of the individuals, which isn't what the StackExchange communities, and Q&A in general is about.

---

Exhibit A

Coloring a node after its creation

A couple of things are very detrimental for the user experience with this Q&A:

1. The answer is incomprehensible to human eye.
   MWE of the questioner is 48 lines. Accepted answer? 62 lines of code, no text.
2. The "answer" is not an answer to the question.
   The question clearly specifies that the coloring should be done "after creation", even in title. Required me to have a diff to find out that answer actually completely ignores that request.

The questioner is not to blame for their lengthy WE either. A true MWE to that question would be the following, silly as it is:

\documentclass{standalone}
\usepackage{tikz}

\begin{document}
\tikz{\node [draw] {A};}
\end{document}

However, any questioner, due to such abuses of MWE, is forced to provide their entire code without simplifications. Otherwise, they risk receiving workarounds as an answer to the MWE that do not apply even to their own real case!

Moreover, what if the questioner is just curious about something? Well, the they cannot assume that answerers will answer the question in the body. So they have to either make harsh warnings, or provide a bizarrely convoluted MWE so that the answerers are unable to outmaneuver the question with a workaround that applies only to MWE, and are forced to either answer the question, or don't.

These are all backed by the exhibit, and facilitated by the MWEs, and them being socially enforced.

---

Exhibit B

Array indexing does not work for the node label

This one shows how MWEs can harm the generality of Q&A's:

1. The answer is only for a specific case of the question.
   A question with the same body and title now has to be asked, only with a different MWE that instead provides a text/math array to use as the node label. See that question here.

This one is much more mild, as the answer at least actually answers the question. It still does it in a vision narrowed down by the MWE, focusing on arrays with numeric content. This yields an answer simply isn't complete, as the question body does not make such a specification.

Without a MWE, it would be easier for answerer to think about text/math node contents. Perhaps they would even think about those first.

---

Exhibit C

Start aligning from another equals sign after some line of equation

This is a pro MWE example. It was not possible for the answerer to give a generally applicable answer. Thus, they openly stated that their "solution" is just a "trick". It presents an example quirks with the solution, and speculates that there might be others.

^{Of course, it's all posed, as it is a self-answered question. I was just presenting my own workaround that wasn't perfect.}

Unfortunately, MWE's aren't usually respected like examples as in this case. If they were all like this, they would be perfectly harmless points to start for the answerers. I do not object against the idea of MWE. I think it's a great idea, misused.

---

MWEs are taken as if almost the main question, with the rest as secondary to it. This attitude towards MWEs make the questions very individualized, and damages the Q&A's applicableness.

I think every StackExchange community should take Q&A reusability as a performance metric. I think the misuse of MWEs is harmful to TeX.SX's performance.

There isn't just one solution to this. Here are some ideas:

• Discouraging answers that put the example at the center.
• Disallowing "answers" from being marked as answers, if they;
  • fail to address the question title/body (e.g. Exhibit A), or
  • fail to address the question title/body for another MWE (e.g. Exhibit B).

A functionality to let the answerer prevent their "answer" from being marked as answer could also be nice, for people that want to provide an individualized workaround for the questioner, while admitting that theirs is "not an answer to the question asked".

Answer 1

While it is definitely true that minimal examples are sometimes asked for reflexively, and it's possible to make questions that shouldn't require them, (see automatic check for MWE for some prior discussion) I don't think that discouraging MWEs is at all the way to address the problem you are raising. Here are some counterpoints to think about (some are the same points raised by Ulrike Fischer in her comments):

1. Questioners generally want to solve specific problems. But much of the time it's hard to tell what they are trying do without a MWE. And removing the need for a MWE doesn't make the specific problem go away and the question more general. Also, things like document class can matter a lot, even for general questions: how to change section titles is a general question, but doing it in the standard classes vs. memoir vs KOMA requires distinct solutions. Similar points can be made for font questions and bibliography questions, among others.

2. Even general questions can benefit from a MWE. What makes them general is not the lack MWE, but the specific question asked along with the MWE.

3. Questions without examples can easily lead to XY problems. MWEs go a long way to clarify issues so that bad solutions don't get encouraged.

4. Answerers are volunteers who give up their time and expertise to answer questions. Providing MWEs as starting code saves time and encourages answers.

5. The "generalization" problem works in both directions. Depending on your level of expertise, applying a general solution to your specific problem may or may not be difficult. This relates back to point 1. Conversely, applying the method used in a specific problem to a more general one may or may not be difficult.

6. The site doesn't operate on rules, generally, but on conventions that people follow. So talk of "abolishing" anything is really not terribly productive.

Some other alternatives

I agree with the main point that answers to very specific questions may be of limited use to others, but there are ways in which answers can be made more general.

1. If you answer a specific problem, think about whether there is a more general version of the code that would be useful to others and include that in your answer.

2. If you think that there is a general question that would help others, ask it yourself, but as with point 2 it's still helpful to have an example to work with.

3. Do what you think is best to help the site, in your own behaviour but don't try to legislate behaviour of others.

Answer 2

Your main claim appears to be,

a MWE forces [queries] to be way too specific.

As @AlanMunn and others have already noted, MWEs are not essential for all queries. However, for a great many queries, they are indeed essential.

• A disturbingly large number of queries starts out vague and unfocused. Postings such as "Ummm, I had some code that was working until last week, but now it no longer does. Why?" are sadly not unusual on TeX.SE. [Maybe you haven't seen many of them because they tend to get deleted if they aren't improved.] Obviously, these OPs need to be encouraged to provide some more, usable information. However, just telling the OPs that they need to provide "more information" often does no good at all. On the other hand, telling them that they need to provide an MWE (and, of course, giving them a pointer as to what "MWE" means) is often all it takes to get them to provide just the information that's required to solve their problems. Of course, sometimes it's only when a semblance of an MWE is produced that one comes to realize that the OP's vague question has no discernible connection to TeX, LaTeX, and friends to begin with. However, without the MWE, it might take a whole lot longer to figure that one out. [Such postings also tend to get deleted!]

• Some queries happen to be of a fairly general nature; e.g., How do I create unnumbered sectioning headers? Most queries are much more specific, though. Without knowledge of the document class that's in use, the main packages that are loaded, and the actual LaTeX or BibTeX code that's causing grief, one simply would never get to the stage of performing an accurate diagnosis of the situation, let along getting to the stage of offering a solution.

• Quite often, there are conflicts between packages. Unless the OP states exactly which packages are loaded, one will hardly ever be able to come up with a correct guess of what's causing the problems. For instance, the real problem in eqref - showonlyrefs was caused by a fundamental conflict between the mathtools package (specifically, one of that package's options) and the cleveref package. Absent the MWE, the OP's valiant attempt at a description as to what might be going would have never allowed me (or others...) to figure out what the nature of the issue was.

• At other times, problems are caused by the OP not using the latest vintages of packages and/or external helper programs. (You'd be surprised to find out how many folks are still using TeXLive2014 in mid-2020 -- and express surprise and even dismay when they're informed that there have been software updates in the meantime.) Unless one can establish which vintages of which packages and programs are in use -- easily done once the OP has provided an MWE -- one will never come up with the correct diagnosis in cases of vintage incompatibilities.

To sum up: If the objective is to keep TeX.SE useful for a great many users, abolishing MWEs would be the wrong way to go.