I've noticed recently that when I write a new question such as this one, the "Similar questions" box just above the Body text entry box shows results that are similar, but they're similar for the wrong reasons. I format my (self-study) questions the same way: problem statement (including which question and which book I'm asking), my work so far, and my question. You can see that if you look at my last 10 questions, at least.

Here's the problem: the "Similar questions" box, which I rely heavily on to know whether I'm asking something that's already been asked, has lately been returning almost all questions that I've asked before! If the question I'm currently writing is that similar to previous problems, I wouldn't be very likely to ask it. I'm guessing that the ML matching algorithm is picking up on my formatting and keywords like the textbook I'm studying.

Here's a proposed solution: the "Similar questions" box should at least limit results authored by the same author as who is writing the current question, if not exclude them entirely.

[EDIT] Per whuber's suggestion, I have re-posted this to meta.so.

  • 8
    $\begingroup$ This is an interesting observation and potentially useful proposal. But since it applies to how all of SE works, the stats Meta isn't really a suitable place to post it. Consider reposting it on the main Meta site. $\endgroup$
    – whuber Mod
    Oct 1, 2021 at 16:50
  • 2
    $\begingroup$ @whuber: Well, perhaps. The thing is, though, that part of the issue is how few questions there are on stats.se versus, say, math.se or so. It wouldn't surprise me if this effect is negligible on the really big sites, and much more useful on smaller sites. I could see how it's site-wide, though. Will repost. $\endgroup$ Oct 1, 2021 at 16:54
  • $\begingroup$ "If the question I'm currently writing is that similar to previous problems, I wouldn't be very likely to ask it." The question might be different enough in order to be asked, but does that make the questions not similar? $\endgroup$ Oct 2, 2021 at 18:50
  • $\begingroup$ @SextusEmpiricus It means that even if they are similar, if the answers to those other questions had been helpful enough to answer the current question, I wouldn't be asking the current question. The Similar Questions box, to my mind, is about being able to easily find out if someone else has asked my question before so I can avoid duplicates. I'm not going to duplicate my own question, so I want to know other people's questions that are similar to mine. $\endgroup$ Oct 3, 2021 at 1:02
  • $\begingroup$ That makes sense. Of course you are not gonna ask a question you that is a duplicate of something you got already answered. However, I was thinking about the similar questions box also being useful to formulate your question based on the similar questions and for instance, draw relationships with other questions as an introduction (this is like question A bit here we do B). In that case, the reminder of ones own questions might be useful. $\endgroup$ Oct 4, 2021 at 19:40


You must log in to answer this question.

Browse other questions tagged .