79

I noticed an old question similar to another existing question and voted to close as duplicate.

After this, the system added a comment:

Does this answer your question? Dotnet Windows service setting up Recovery option using Managed code .

It’s completely different to what I meant. I didn’t want to ask the OP, but suggested the community to do a cleanup. See an example, where user interpreted automatically created comment as a real question and replied to it.

I understand that the comment is applicable to a new question, but it’s only one of ways how close of duplicates is used.

Possible fixes:

  1. Restore the original comment with a link to explanation of duplicate (similar to this suggestion: Add clarification link to "Possible duplicate" automated comment).

  2. Make comment time-dependent: if a question is new (e.g. less than three days), show the new phrasing: “Does this answer your question?”. If it is older, show the old phrasing: “Possible duplicate:”.

This bug is a result of recent changes (e.g. see this answer on the Post Notices rollout announcement for another confusing scenario caused by a new wording).

I am aware of five possible scenarios:

  1. New question, similar to existing question with answers. (The current auto-comment is fine in this scenario.)
  2. New question, similar to existing question without answers. (As pointed out in another question, the current auto-comment makes no sense in this case.)
  3. Existing question, similar to another existing question; an example of this scenario is discussed in this question. The existing process for this scenario is described in another question.
  4. A new question, similar to a previous question by the same author
  5. (rare) I’ve created a question and then found that another user asked the same question at the same time. I voted to close my question as a duplicate (I suppose it would probably be better to delete it instead), and see the auto-comment by me asking myself the question “Does this answer your question?”. (The previous “possible duplicate” comment made more sense in such a scenario.)
| improve this question | | | | |
  • 33
    I completely agree that this is often not the message I want to give when flagging as a duplicate. I'm really not so much asking as telling. And this is especially the case when a user posts the same (or very very nearly the same) question multiple times. Even if the prior one wasn't answer, the newer one is still a duplicate. It also implies that the linked duplicate should answer OP's question exactly, but duplicates are often not simple copy/paste answers. – Patrick Q Dec 3 '19 at 19:37
  • 42
    It's part of SO trying to be nice and overshooting the goal. I don't like it one bit, and newbies are responding to the comment with "yes" and the like, as if I had typed that useless question myself. Definitely not a fan. – Chris G Dec 5 '19 at 12:23
  • 10
    @PatrickQ But... but... telling is unwelcoming! Hic. – Lightness Races in Orbit Dec 8 '19 at 19:22
  • 2
    Yes, that was my thought too. Those messages seem to be designed for new questions. However, we can change those messages (or delete them). – P.Mort. - forgot Clay Shirky_q Dec 9 '19 at 5:17
  • 3
    Even if this was not your intention, comments on questions are primarily directed at the OP so it makes sense to use language that's directed at them. If they agree with your suggestion they can mark the question as a duplicate themselves. Also, I don't understand your comment "it’s only one of ways how close of duplicates is used." How else it it used? – ChrisF Dec 10 '19 at 10:11
  • 17
    Taking this and other related posts under consideration. Thanks. – Yaakov Ellis Dec 10 '19 at 10:16
  • 5
  • 2
    @ChrisF , I am aware about at least 4 scenarios: 1. New question, similar to existing question with answers.(supported by the current auto-comment) 2. New question, similar to existing question without answers 3.existing question similar to another existing question( discussed in this thread, existing process is discussed here ) 4. New question, similar to previous question by the same author – Michael Freidgeim Dec 10 '19 at 12:28
25

I am not a fan of users having their comments changed without prior warning.

Personally, I'm ok with the new wording for my own comments. It more-or-less serves the same goal I have when voting to close, (let users know a question may have been answered somewhere else). But this only applies to me personally.

Unlike post notices, comments are inherently displayed as coming from a specific user. Users reasonably expect to have control over the content of their own comments.

When I VTC a question, I almost never want to enter into a conversation with the OP about whether their question is a duplicate. - Raedwald

This change should have been announced officially beforehand, similar to the post notices.

| improve this answer | | | | |
  • 6
    I have zero (or near zero) interest in my dupe vote resulting in a protracted conversation with OP about why their dupe question isn't a dupe. – Richard Jan 1 at 22:13
17

Short answer here is that we're aiming to facilitate what we think is the core utility of these comments... We might be wrong about that utility, or wrong about whether our change helps more than it hinders - but determining that will take time and effort. Yaakov has marked your request as "deferred" to indicate that, while we believe your concerns likely have merit, we can't address them right now; we'll revisit this in the future once we have a better idea of how best to do so. Please see Yaakov's answer for a final update.

The rest of the story

It's worth remembering when discussing these things that they are a combination of system behavior and social convention. Duplicates were originally implemented exactly like every other close reason - there was no special behavior, no extra support for recording or displaying links, etc. Most of the behavior we now have either started out as a practice among users, or was created to replace such a practice... In some cases, both. We didn't design this system so much as we paved the "desire paths".

When we're looking at something like this comment, born of the complex maelstrom described above, we gotta ask first: why would folks want a comment? What purpose does it serve in the context of interpersonal interaction? Why is that fence there?

Here's what I see:

For the asker, such a comment...

  • ...provides an inbox notification, letting them know that there is potentially an answer to be had, or perhaps instead a misconception to clear up.
  • ...provides an entrypoint into a conversation regarding the merits of the suggested duplicate (that is, it is possible to reply to a comment; it is not possible to reply to a vote).

For the voter, such a comment...

  • ...implicitly encourages interaction and discussion over the merits of the proposed duplicate, setting the tone for such an interaction.
  • ...ensures that they can be made aware of feedback regarding the proposed duplicate (from the asker or from others) (again, no one can @-notify a voter)

For other readers / voters, such a comment...

  • ...lets them know that the question is possibly a duplicate, and provides them with details on the proposed original (3rd-party readers cannot see this information at all unless they have close privileges; even those with closing privileges can only see this information by opening and navigating through the close dialog). The system automatically upvotes the comment as more voters choose the same option, thus ensuring that it grows more visible as the question moves closer to being closed.
  • ...provides fast access to what is at least a related question, even if the question is never closed.
  • ...ensures that the system cross-links both questions in the sidebars of both posts (such links are not otherwise added to PostLinks unless/until the post is actually closed).

Now. Not all of these effects are equal in their utility, nor are all well-served by a comment. However, it is likely that they have some non-trivial importance, as other approaches were possible and were tried by some or many users in the time prior to the system supporting this one directly (editing body, editing title or tags, answering, etc). IOW: the organic solution may not be optimal, but it likely won out over alternatives due to various practical advantages such as those sketched out above. In order to replace it, we need not replicate its total functionality, but we must improve on its utility to such an extent that users do not just fall back on leaving such comments manually (at which point we would have no influence over their content or tone). And any change short of replacing it should strive to improve one or more of the known utilities without unduly hurting the rest.

Facilitating conversation without hiding information

At last that brings us to the change in wording. Going back to the charter for these changes, where Meg laid out three key priorities:

  • Delivering improved, private feedback to post authors
  • Not putting users who curate content on the spot
  • Giving actionable, understandable information for the vast majority of public viewers

Now, this isn't private feedback; we're working on ideas there, but right now duplicates are public - unless the asker deletes their question, it sits there visible to the world. We don't have a facility for private (or even slightly out of the way) discussion, and even if we did have one putting such a comment there would hurt the utility for other readers and voters (see above) - as long as the question itself is publicly visible, any discussion surrounding it needs to be as well...

...But, perhaps we can do this while reducing friction between the folks identifying duplicates and the folks asking them. The hope is that this wording works toward that end...

...But does it?

Well, we ran an A/B test for about a month with this new wording. Folks from both groups were flagging or voting to close questions as duplicates, so I compared the frequency with which those comments resulted in replies, edits, and flags (on the auto-comment).

...there was effectively no difference. Replies hung around 19%, edits around 27%, flags around .2%.

IOW: we have no hard data suggesting that this is better or worse. All we have to go on is folks like you, and reports from folks asking questions. Those are both useful, but they take time to collect and interpret.

So... We'll keep watching this, and make adjustments as the long-term effects become more clear.

| improve this answer | | | | |
  • 7
    You well described that changing of a text of a single comment is a complex decision. Why it wasn’t posted as a separate Meta question prior to implementation to allow proper discussions with both answers and comments? Many bugs/wrong decisions could be avoided if they would be brainstormed on Meta on the early stage and all possible usage scenarios would be considered. – Michael Freidgeim Dec 17 '19 at 20:59
  • 2
    First: I appreciate the time and thought you put into all your posts, and I know you must struggle with decisions on just how to phrase things, frame outlooks, etc. Kudos. But I will say both bullets in your “for the voter” section are not at all for the voter, they’re the opposite of what the voter wants. They’re transparently for the asker, the precise frustration the OP here expressed in his question. I really wish SE would outright say “we are making things easier for askers at the expense of curators”. – Dan Bron Dec 17 '19 at 21:34
  • Both of those bullets have corresponding bullets in the "asker" section, @Dan - they're mirrored because they correspond to the same form of interaction, but from different roles. Both are things I've heard voters express in the past, but of course they don't correspond to the desires of all voters. Ironically, I've heard far fewer askers express such a desire, at least in so many words - I suspect most voters have more familiarity with the system than most askers, and this lends itself more readily to seeing the opportunity for such interaction, and also seeing the utility in it. – Shog9 Dec 17 '19 at 21:39
  • 3
    @Shog9 This ignores the practical reality that most such interactions are fruitless and lead not to improved questions but to mutual enervation. As expressed in the comments under the question, most curators are fully aware of this dynamic and are leaving the comment for the community at large, not the OP. They don’t want to invite (highly likely contentious) interaction with the OP. – Dan Bron Dec 17 '19 at 21:42
  • 4
    Short answer there is time, @MiFreidgeimSO-stopbeingevil. As hinted at by the blog post I linked, the set of changes that rolled out over the last month or so was both massive in scope while also representing just a tiny corner of a larger plan for future work; ideally, we would be a lot more proactive in our communication here, but for various reasons (cough) that's been lagging - so we do the best we can with follow-up work instead. – Shog9 Dec 17 '19 at 21:42
  • 2
    I think you're describing most interactions period, @Dan. Some folks have more of a stomach for that than others. There are people whom I've watched patiently answering the same questions, patiently explaining the same concepts, for decades; I wish I had that kind of patience. I suspect most of us gradually pull away over time, optimizing our efforts for the areas where we see them do the most good. But somehow, the system has to work for all of us. – Shog9 Dec 17 '19 at 21:46
  • 2
    Thank you for this wonderfull insight into what drives the decissions behind something so trivial as a comment. – Luuklag Dec 17 '19 at 22:41
  • 1
    Nice answer @Shog9 – Meg Risdal Dec 19 '19 at 6:50
  • 7
    So far this change has resulted in my being drawn into two unwanted conversations. Now I have to remember to redact the auto-comment each time I vote to close something as a dupe. – Richard Jan 4 at 19:31
8

Some more follow-up on this: Our data team helped us in looking at if the change in comment language for duplicate comments has had any effect on the close rates and the comments that follow.

  • We haven’t seen a change in the closed-as-duplicate rate for the past two years, including during the deployment of the new duplicate text.
  • We measure a 3.8% increase in the chance that someone “gets in a conversation” with the new duplicate text. (The old text had a 42% chance of no further discussion, and the new text has a 40.4% chance of no further discussion following the duplicate comment)
  • We measure an 11.8% drop in unfriendliness in the comment conversation with the new duplicate text versus the old text
    • Unfriendly Comments from the Question Owner have dropped 6.3%
    • Unfriendly Comments from Other People have dropped 11.4%

So overall the new comment seems to be fulfilling its purpose: informing the asker of the duplicate, and doing so in a way that is improving welcomeness (or decreasing negative interactions) compared to the old text. Given this, this request will be .

| improve this answer | | | | |
  • Re "3.8% increase": Shouldn't that be percentage points ("3.8% points increase")? Perhaps also for 11.8%, 6.3%, and 11.4%. – P.Mort. - forgot Clay Shirky_q Feb 5 at 13:57
  • 3
    It is 3.8% relative increase. The percentage itself moved 1.6% (from 40.4% -> 42%). The others are relative as well. – Jason Punyon Feb 7 at 13:10
  • 7
    My question is about the scenario “Existing question similar to another existing question”, for which the new message does NOT MAKE SENSE. Your stats are not relevant to the bug, as they are likely work for “New question, similar to existing question with answers” scenario, for which the current auto-comment makes sense. I’ve updated the question to list the scenarios that I am aware about. Strange that your QA didn’t verify all these test cases. – Michael Freidgeim Feb 14 at 13:45

You must log in to answer this question.

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