Taking a look at the revision history of the question, "Can an Arcane Trickster or Eldritch Knight with the Magic Initiate feat (picking the wizard list) use spell slots to cast the chosen 1st-level spell?, you will notice six gold-badge open/close actions. You might be tempted to say "looks kind of like an edit war, we have a meta for that already", but there is a lot going on there, and I think it beneficial for the community to reflect on the situation and have a discussion about how this situation should be handled.

I will attempt to faithfully reproduce a complete timeline of all the relevant goings-on here. There are three questions involved, they will be labelled so:

A: Can you cast a spell learned from the Magic Initiate feat using spell slots?

B1: Can an Arcane Trickster or Eldritch Knight with the Magic Initiate feat (picking the wizard list) use spell slots to cast the chosen 1st-level spell?

B2: Can an Arcane Trickster or Eldritch Knight who takes the Magic Initiate? [duplicate]

I have used the names of everyone who was involved in the community moderation of this situation, all of whom are over 20k rep. Nothing said here should be construed as anyone having done something with ill intent - I believe everyone named acted in good faith with the best intentions to push everything toward a desirable resolution. If those users named wish their names to be redacted they may do so or leave a comment and I will do so.

The Timeline

  1. B1 is asked. (10/21 3:54)

  2. B1 is dupe hammered by Purple Monkey as a dupe of A. (3:55)

  3. The asker of B1 responds in an answer to A asking about the particular case of AT and EK. (4:23)

  4. User's answer was deleted. (5:13)

  5. V2Blast starts a bounty for updated answers to A. (5:23)

  6. V2Blast edits B1 explaining that the question is specifically concerned with AT and EK, which is something not addressed in any of the answers to A. (5:29)

  7. B2 is asked. (11:32)

  8. Thomas Markov dupe hammers B2 as a duplicate of A. (11:35)

  9. Thomas Markov reopen hammers B1. (11:55)

  10. Thomas Markov answers B1. (11:55)

  11. Thomas Markov answers A. (12:13)

  12. NautArch and Thomas Markov edit A to be asking specifically about the example given in the question. The motivation here was to more clearly differentiate A from B1. (12:44)

  13. V2Blast rolls back Naut's and Thomas' edits to A. (18:41)

  14. NautArch dupe hammers B1 as a dupe of A. (19:11)

  15. Thomas Markov copies his answer to B1 into his answer to A. (19:14)

  16. GcL reopen hammers B1. (19:22)

  17. Naut, Thomas, and GcL discuss the question briefly in chat and come to a concensus that it should be closed. (19:25)

  18. GcL dupe hammers B1. (19:32)

  19. Thomas rolls a natural 20 to leave the question closed. (19:35)

  20. Doppelgreener votes to reopen B1 and Darth Pseudonym sends it home with a reopen hammer. (21:15)

  21. Thomas Markov opens a meta discussion over whether or not B1 should be closed as a dupe of A and answers in favor of closure. (10/22 16:17)

Where did we go wrong?

There is a lot going on here. Too much. The trouble is that B1 seems to be well within the scope of A, but none of the answers to A shed clarity on the particulars of B1 (at the time B1 was asked).

GcL expressed their concerns rather poignantly in chat:

I would assume they find it frustrating to have questions closed as duplicates of something that doesn't answer their question, because I have found that frustrating. But they might not.

It does seem like it's a non-obvious and circuitous path to get to the resolution.

  1. Ask question that isn't currently answered.
  2. Get closed as "answer already exists"
  3. Go back and forth in edits.
  4. Get question opened, answered, and closed again.
  5. A bounty, for which they didn't have the points themselves, added to other question.
  6. The answer to their question gets added to the other question.

I don't think there's a way OP could have gotten to the end state directly. The path seems like, "please cause error to get swat on the nose and service."

What do we do when a question is asked that is obviously covered as a subset of an existing question, but the answers are insufficient for answering the special case?

We should also consider here: what do we expect a new user to do? How can we improve the new user experience here?


3 Answers 3


I think there are many different things going on, each adding adding events to the that time-line, making the discussions more confused and sprawling, and probably why we have two meta discussions for it. I'll briefly identify the different aspects and then try to go through them all in more detail.

  • The answers to the general question needed updating. These updates (specifically the 2018 PHB Errata additions) even make the specific question easier to answer (because more of the rules are defined).

  • The general question asks generally, using a generic example. The specific question uses an example which adds a small additional layer.

  • Gold badge holders disagreed on whether to mark the question as a duplicate, without there being a clear cut discussion happening. (There's a few comments on the Q, and a conversation in chat).

  • It's a new user asking the specific question.

Updating the old Q&A

The (good) answer has had a note on it regarding the 2018 errata for over a year, but nothing like more attention and bounty to get things updated. This is just our processes working as intended, but it makes it much easier to go "well the generic question should be updated", because it should be updated anyway for other reasons.

Specific question versus general

This is maybe where it gets interesting. Let's try to dissect the nature of these questions with each other. The general question asks how it works for any of the standard spell-casting classes (using the example of a wizard). The specific question adds a layer to that, in that there's a bit more to the spellcasting class — that the class is a Fighter/Rogue, but using the wizard spell list. Thus there's ambiguity between class, spellcasting class, and Magic Initiate class. This is small, but significant and warrants address on some level.

But where to put that address is something we decide as curators. We can either cover it as a separate Q&A, or include it in the general one. And which one we choose will depend on factors such as significance of the layer, scope of the general Q&A, and possibly a little historical lag (we didn't discover it was a partial dupe in time, so it got a good answer and we might as well leave it).

Remember that we thrive on questions and answers being specific. Shorter, more direct answers will more clearly and easily answer the question for (future) readers, so we want to at least minimize bloating general Q&As. This means finding the line at which a subquestion is better of a new question. Those lines aren't always clear.

Users disagreeing

There's a bit to unpack here too, but that doesn't mean anyone involved did anything wrong. Part of it is that many of the voters had 5e gold badges meaning closures and reopening happened much faster than normal, cutting some of the room in which the relevant discussions usually start.

And then the discussions which did happen, occurred in chat. And that isn't wrong either, it's one of the things chat is good for, but it does mean there isn't made readily available for further voters to see and consider. Treating the round through chat as the normal initial process, a meta was opened the first reopening, which seems like fine timing considering how involved the initial process was. This is our processes working, just a one-off with more clatter than normal, but nothing problematic.

New user

So there's a new user asking the new question, and doesn't quite know the processes. Reposting a closed question, nor new question/clarification requests as answer are rare, and we can clean up and explain to our best ability, as was done. The best we can do to help new users is to explain what has happened to their post and what the need to do.

Then it's new user coming in to the improve-old-question process, which normally requires reputation to start (for comments or bounties), so they aren't able to do it "correctly". That always means another user needs to do it for them, so there'll be the mess of whatever post they used to start this, explaining how these processes work, and then hoping they'll stick around to reap the benefits. This isn't a great new-user experience, but is just part of the Stack Exchange new-user experience which is often not great, but it's probably best to leave the rest of that discussion for another meta.

Pulling the strings together

Nothing really out of the extraordinary happened here. Tiny little things compounding, giving a messy time-line, but not really anything to correct. I'll just reiterate some general, applicable encouragements.

  • Identify duplicates and sub-questions, additional layers and aspects. And maybe try to remember what it's like without the system mastery which has earned you the gold badge. And I'll be first to recognize that that's hard. And I didn't really identify the above dissection until I'd rattled it for a while and there was a meta discussion which needed answering. We need to talk it out sometimes.

  • Be careful with unilateral votes on subtle matters. I don't want to discourage anyone from voting, but making your thoughts clear and starting the discussion is generally beneficial to parts.

  • Don't be afraid to show the process. Double that for cases with subtleties, messy processes, and new users. Just clean up when it's done and settled.


If a question should be answered by answers to another question it is likely a duplicate. If a question does not have satisfactory answers, bounty it

From the What is a bounty? page:

[...] If you see a question that has not gotten a satisfactory answer, a bounty may help attract more attention and more answers. [...]

The question about how Arcane Tricksters / Eldritch Knights interact with the Magic Initiate feat is obviously a subset (or subquestion) of asking how the Magic Initiate feat works in general.

Thus, clearly one question is a subset of the other, but the question remains: Should answers to the general case address this specific sub-case?

I have absolutely no idea. However, it doesn't actually matter in this particular case because that sub-question is (now) answered by answers to the more general question.

Whether or not answers to the general case should have covered the specific case of AT/EK is unclear, but at this point, that case is covered. I believe the best option would have been the following:

First, determine whether or not answers to the general case should cover the more nuanced sub-case of AT/EK.

If they should, then the new question is a duplicate and the older question could receive a bounty for not having adequate answers.

If they should not, then it is not a duplicate and can remain open.

  • 2
    \$\begingroup\$ Downvoting on the conviction that a new user should not need to depend on the kindness and willingness-to-bounty of strangers to get an answer on a question like this. \$\endgroup\$
    – Novak
    Commented Oct 23, 2020 at 17:19
  • \$\begingroup\$ @Novak I linked some related posts under the question here in a comment. That seems to be the general expectation. I also never said that that had to happen, only in the case that the new question should have already been answered and hasn't been answered could that occur at all. Even then, I'll gladly bounty such questions the exceedingly few times this is likely to occur. Sure, it may not be an ideal situation for the new user, but it seems to be what the site expects \$\endgroup\$ Commented Oct 23, 2020 at 17:34
  • 1
    \$\begingroup\$ Additionally, getting your question closed a duplicate that doesn't answer your question while you don't have the reputation to bounty that question does not mean your question won't get answered. I myself was more than willing to answer their unanswered question, bounty or not, and I do genuinely believe others would be as well (that those here are kind enough and/or willing enough to place bounties where this isn't a problem) \$\endgroup\$ Commented Oct 23, 2020 at 17:39
  • 1
    \$\begingroup\$ @Novak Just because a new user doesn't understand the duplicate system doesn't mean we should ignore it. If we don't want to close questions as dupes at all (because why not close for a new user, but closer for an old one?), then that would be an interesting meta to cover. But having a duplicate rule that depends on a user's status seems even more convoluted. \$\endgroup\$
    – NotArch
    Commented Oct 23, 2020 at 18:16
  • \$\begingroup\$ @Medix2 nevertheless, I dissent. If this answer gets the highest number of votes, it will inevitably be pointed to as policy and then actually be the policy. And regardless what "seems to be the general expectation" I consider a policy telling users to just hope someone is paying attention and willing to bounty for them to be an actively bad policy. Your promise to bounty does not withstand this dissent because you might run out of points or experience some external change in circumstance that sees you posting less frequently. \$\endgroup\$
    – Novak
    Commented Oct 23, 2020 at 19:38
  • \$\begingroup\$ @Novak In all honesty, if you believe the good nature of those on here is not enough, and that meaningful harm will be caused by questions with inadequate answers not receiving the bounties that they should while users nonetheless mark the newer question as a duplicate: a Meta post calling for a policy change sounds in order (though I'm unsure that this even is a policy, moreso just how things have been done) And I can certainly see the use in something along the lines of "Don't close questions from (lower-rep) users as dupes if the question isn't aswered in its dupe target" \$\endgroup\$ Commented Oct 23, 2020 at 19:40
  • \$\begingroup\$ @medix2 Also, brushing away this particular dispute as "obvious" is just begging the question. We often get into intricate discussions over obvious things, but we usually don't take it to the level of edit wars or this... edit-war-like tangle of competing answers, deletions, closures and re-opens. That's why I specified "on a question like this." \$\endgroup\$
    – Novak
    Commented Oct 23, 2020 at 19:41
  • \$\begingroup\$ @NautArch no one is proposing to ignore the duplicate system. Please do not imply that I am. See Someone_Evil's answer for details on why the specificity of this question may make it not a duplicate, and the more general notion that there is some subtlety and flexibility on the notion of what makes for a duplicate question. \$\endgroup\$
    – Novak
    Commented Oct 23, 2020 at 19:44
  • 1
    \$\begingroup\$ That all said, I'm unsure such a proposal is needed as one of our dupe criteria is that the older question obviously answers the newer one. And what you're proposing seems to only apply if duplicates are made where that criterion is failed; in other words, where questions are wrongly marked as duplicates. I believe the question should have remained in some state until a Meta was asked about whether the general question should address the specific case; this didn't happen. \$\endgroup\$ Commented Oct 23, 2020 at 19:45

What do we do when a question is asked that is obviously covered as a subset of an existing question, but the answers are insufficient for answering the special case?

Answer it. In terms of acting as a knowledge base, having specific questioned subsets of broad questions answered separately vastly improves readability and searchability of the entire information set.

If you have any flexibility in how broad a question will be, you will have questions at both ends of the spectrum. If you push all possible answers to be collated under broader questions by deleting/marking as duplicate any more specific instances, you make a less searchable knowledge base that also requires more reading (of layered answers) to answer questions. This is the worst of both worlds.

In effect, by deleting more specific instances of broader questions you are deeply undermining the far more important rule of be specific, while also lowering usability of the site in general.


You must log in to answer this question.

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