Many users of the site have a need to keep up with current research in specific areas related to computational science. Recently, someone asked

Which journals should I read to keep up on advances in solving PDEs numerically?

The question has been closed as not constructive. A similar question is

Venues for publishing papers that emphasize software

Do we want to allow such questions or not?


I think such questions should be allowed. They have a factual answer (a list of all journals on a given topic) which may not be easy to find anywhere else. They also admit subjective answers, which would be short lists of the most useful/important/whatever journals in some area. The subjective answers are very useful since nobody could read all the journal articles about, say, numerical PDEs.

As discussed in this blog post, subjective questions can be very useful as long as people stick to answers based on references or on personal experience. Many of us have a lot of personal experience reading the computational science literature, and we can use that to give useful answers to newcomers or outsiders.

As Geoff points out, apparently SE staff have an aversion to questions asking for lists. But the stated reason is that such questions are really polls. That's simply not true in this case; take a look at Venues for publishing papers that emphasize software. I learned a lot from answers to that question and I consider myself an expert on that topic. I'm confident that that information doesn't exist in any other single place. Indeed, I've been using that question to "show off" the site to many people, who have reacted positively.

  • $\begingroup$ My eyes are starting to bleed from reading stuff about lists. The gist of your argument, as far as I can tell is: This question is a recommendation question, and it draws the kind of traffic we want. That's fine by me. To play Devil's Advocate, what if we keep getting questions like "What journals should I read to keep up on advances in _________ ?" (e.g., linear algebra, scientific computing at large) Do we want to expand the site in that direction? $\endgroup$ Feb 7 '12 at 5:40
  • $\begingroup$ Yes, that's part of what my question was meant to explore. We'll have to see what the user community thinks. $\endgroup$ Feb 7 '12 at 11:05

I think the Solomonic solution would be to allow list questions when they provoke answers beyond just lists. "What are tools for X, and when should we use them?" goes far enough beyond the "list of X" question that it should be considered a reasonable question for the board.

(Of course, questions that just ask for a generic listing should be edited to encourage better responses!)


My initial response

I think the content of the question DavidKetcheson linked to is certainly on topic, and I have no problem with re-opening that particular question. My issue with that question was not subjectivity at all, because I agree, such answers fall more on the "good subjective" side of the spectrum.

My issue with that question was that it looked like a "List of X" question, and there are several links on Stack Exchange talking about how such questions are bad:

Opinion on "list of X" questions is mixed. These links tend to talk about how list questions might be good in some contexts:

There may indeed be some context in which a list question is okay, but looking at the two responses, one was a list (by JedBrown) with not much in the way of commentary per bullet point, and another was a shorter list (by DavidKetcheson), again with not much in the way of commentary. I believe that the author of the question, Dan, was looking for a list of journals when he asked that question. When the subject of community wiki came up in Meta, we eventually had a comment thread that discussed closing the question DavidKetcheson linked to. Dan and I both felt that the question probably should be closed unless community discussion tilts things the other way.

I am inclined to decide against questions that ask for explicit lists. The question What core skills should every computational scientist have? was well-intentioned and topical, but asked for a list of skills explicitly.

I draw a contrast between the linked question and a questions asking for software recommendations on the grounds that often times, questions asking for software recommendations don't necessarily need a list. If I ask about recommendations for a good numerical integrator or linear algebra package for a specific use case, I may get a list as an answer, but mentioning one or two software packages with sufficient commentary is appropriate and valid. Consequently, I don't consider software recommendation questions to be "list questions" unless the poster explicitly asks for a list.

That was my whole thought process, after doing all of the research; it was originally prompted by dmckee's (seemingly apoplectic) comment about list questions, and I asked about list questions a bit in the Teacher's Lounge to get a feel for it. I thank dmckee for giving me a heads up about the issue of "list questions".

  • $\begingroup$ Your reason for closing says: "We expect answers to generally involve facts, references, or specific expertise..." That is exactly what both answers contain. $\endgroup$ Feb 7 '12 at 4:37
  • $\begingroup$ I believe you're referencing something irrelevant. One of your links above says: The discouraged kind of list is often called an itemized list. This is where all the answers put together comprise a list, with each answer having one item. A singular answer that contains a list is not an itemized list and, in many cases, does not suffer from the same problems. $\endgroup$ Feb 7 '12 at 4:40
  • 1
    $\begingroup$ @DavidKetcheson: I chose that reason for closing the question because the first sentence states "This question is not a good fit to our Q&A format," and a later sentence mentions that the question will likely solicit polling (which is a side effect of lists). If we're going to get into a back-and-forth discussion about it, I'm happy to discuss it in chat. Like I said, I'm fine with reopening the question. The only thing that I care about is that there's some sort of standard or decision rule to move forward so that we don't have to hash things out on Meta or in comments as often. $\endgroup$ Feb 7 '12 at 4:53
  • $\begingroup$ Also, for clarity, the post that has the quote that David is referring to is in meta.literature.stackexchange.com/questions/176/…. $\endgroup$ Feb 7 '12 at 5:05
  • $\begingroup$ Sorry for one more comment, but it's not clear to me if this answer is a "yes" or a "no". So I'm not sure what we'll do if it gets voted up... $\endgroup$ Feb 9 '12 at 8:18

An updated response

In the past couple days, I asked some of the more experienced mods and SE employees about what's going on with journal questions. Opinion was divided, making it sound like these questions are sort of a corner case of list-type questions. The impression I get from the community (based on the current 8-2 total) is that we're going to allow these sorts of questions. Based on that response, here's my take on things:

For starters, DavidKetcheson is right: some questions legitimately have lists as answers. Opinion questions (that ask for lists) for which there are "no" wrong answers (or rather, that allow a wide range of admissible answers) are definitely not suited to the SE format. They allow too much in the way "me too" answers that don't add much to the discussion, even though it's possible to have really good answers. However, questions like, "What languages are most common in computational science?" or "What are good numerical linear algebra books?" do have factual lists as answers, and thus, are reasonable questions on their face.

However, I believe we as a community need to be careful. A good answer to a recommendation question has some of the following characteristics:

  • It states reasonable recommendations (e.g., "Green Eggs and Ham" is not an answer to "What are good numerical linear algebra books?", but Trefethen and Bau is a reasonable recommendation).
  • It justifies each recommendation in some way. (e.g., Trefethen and Bau is a good linear algebra book because it gives a broad overview of numerical algebra in a clear, expository format. Breaking the book up into lectures makes it easy for self-study, and the examples they use often shed a new light on familiar concepts.) Sometimes, grouping recommendations can be okay; it depends.
  • It provides links where appropriate

What I liked about the answers to the software question was that the better answers of that question had links, as well as paragraphs justifying the response with quotes from the aims and scope of the journals, or personal experience (either reading the journal, or experience submitting to the journal), and those justifications really put the recommendations into context. What I didn't like about some of the answers to that question was that they were a sentence or two stating that software was in the aims and scope for the journal recommended (in whatever subfield).

In that vein, I'm more inclined to think that a journal question is not a list question if answers tell me:

  • Again, reasonable recommendations.
  • Something about the quality of the journal. Is it prestigious? Is it terrible? (I think it's fine, although perhaps impolitic, to recommend NOT to submit to a particular journal in a given subfield.) Is it widely-known, or is it more obscure?
  • Who reads the journal? Is it a general audience journal, or limited to specialists? What disciplines do they come from?
  • What kinds of articles are in the journal? Are they more applications-based? Are they more methods-based? Do the journals require computational results? Do the journals allow proofs? Are articles shorter (a few pages) or longer (10-20 pages, or more)?
  • Does the journal allow you to post pre-prints or post-prints on the arXiv? Do they allow supplemental material?
  • If I want to submit something, how long will it take them to turn around a review?

I say all of this because I too want to know what journals to read, and I've learned that not all journals are equal. Merely getting a list of journals doesn't help me (and it probably doesn't help anyone else), because my time (like everyone else's) is limited. If an expert tells me in what situations and why I should read a journal, I'm much more likely to actually try and read it than if they don't, and their answer is profoundly more valuable. If they give me a list without any of that information, it's virtually worthless to me, because they've given me no context. With minimal effort (a Google search), I could probably cobble together a list that answers the question. With some effort, I could probably answer some of the questions I posed about journals (read the aims and scope, skim some articles, hunt down the impact factor). To actually get an expert opinion on a journal I initially know nothing about would take considerable effort, and I think those expert opinions are most likely to be valuable and save people time.

If a lot of the answers are short, or lists without justification, I'm much more likely to comment that I think the question isn't constructive, because it isn't a good example of what the SE sites can do when they reach their full potential. I think the answers to the PDE question could be better, and I think for these questions to have useful answers, the moderators (myself included) are going to have to keep an eye on them to make sure that we have good answers that are valuable for people.

I'll also move to wikify any journal question.

  • 1
    $\begingroup$ Geoff's comments go to something I wanted to put in an answer, namely how does one go about making the list (rather than giving a list per se). This how-to knowledge/guidance is high-level stuff, and obviously quality can be good or bad, but IMHO ought to be allowed. $\endgroup$
    – hardmath
    Feb 10 '12 at 16:04
  • 1
    $\begingroup$ Based on what I've seen, and from talking to other mods, a recommendation-type question is a becomes more of a candidate for closing if it attracts answers that are basically just lists without any explanation, because the site is about offering informed opinions rather than data dumps. One of the exceptions to the rule would be a question like, "What books has Gil Strang written about computational science?", in which case you could simply list the books he's written without any commentary whatsoever. $\endgroup$ Feb 10 '12 at 17:11
  • $\begingroup$ I think I like what you're suggesting, but a potential downside is that it asks for more subjective answers. To experiment, I've tried to do what you ask for in my answer here: scicomp.stackexchange.com/a/1118/123 $\endgroup$ Feb 11 '12 at 14:34
  • $\begingroup$ @DavidKetcheson: I agree, the guidelines I'm suggesting invite more subjective responses, but they're inspired by the "good subjective, bad subjective" blog post by Robert Cartaino that you yourself linked to. Done correctly, they should welcome more informative answers; I found your revised answer significantly more informative than its previous version. Done badly, the opinions are either misleading (in which case, I expect downvotes and backlash), or we continue to get list answers (again, downvotes and backlash). $\endgroup$ Feb 11 '12 at 17:12

You must log in to answer this question.

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