auto-detect the phrase "it doesn't work" and ask user if enough info given

Question

I get pretty peeved at questioners who say "it doesn't work" in their questions, without elaboration. Someone has to comment "It didn't work how? How do you know? What error message was given? What did you expect to happen, and what happened instead?" And everyone waits until the asker notices the comment and writes back. By the time that happens, the question has moved quite a bit further from the center of everybody's radar. It will likely be forgotten if the OP doesn't use @ properly to address the person who asked for clarification. And then only that commenter gets reminded about it.

Then someone, maybe the same commenter, provides a solution to the question. And the OP writes back "I tried that and it didn't work." And the cycle starts again.

It's bad for the asker and it's annoying for the answerers.

I think it would be helpful for new users, if SO could auto-detect the phrase "doesn't work" (and variants: don't work, didn't work, did not work etc.) and ask them: "You said something 'didn't work.' Did you tell us what happened, specifically?" with a link to that article about how to ask good questions for further detail.

It seems like this could be implemented on the client side (JavaScript) without taking much time or memory.

Granted, there's only so much you can do to alleviate the problem of people not giving enough info to enable others to answer their questions. We certainly can't solve it completely through automation. But it seems to me that the above would help train new users and would make a worthwhile difference. What do you think?

Update:

Proposed "smell" regexp: d(id|o(es)?)\s*n['o]t\s+work.

Proposed alert text: I smell the phrase 'It didn't work'. Did you tell us specifically what happened? [Yes] [No] [What do you mean?] And the last button can link to Jon Skeet's Writing the perfect question (wish I could link to the Problem Statement section) or some other suitable explanation.

The user can click [Yes], [No] (or anywhere else?) to dismiss the box. It doesn't prevent the question or comment from being posted: it's just a reminder.

And no, I don't mean a JavaScript alert() but something like the boxes we get now, for trying to vote more than once in 5 seconds, for example.

Update:

I just read another question where the asker said "it's not working" and forgot to specify what was actually happening. I know he would have been happy to specify it, because in fact he did so in a later comment. It's just a matter of being reminded.

On the down side, my earlier regex would not have caught the "it's not working" form of the phrase. Maybe the regex should be simplified to

n['o]t\s+work

This would catch "it isn't working", "it's not working", "they're not working", etc., while still matching "it doesn't work" and so on. Yes, it would probably gather some more false positives, but if the reminder is not too obtrusive, I think people who get the reminder when it's not applicable will understand why it appeared and will not be too perturbed about it.

It seems to me that this would be preferable to explicitly trying to match "[i']s not working", "[a']re not working", "isn't working", "aren't working", etc. But the latter is a valid approach.

Update:

Collecting a good idea from way down in the comments:

The warning could also just be dropped at a certain reputation level where it is considerably safe to assume that people will not forget to add this information.

On a related topic: Now that the question has 31 upvotes, and two answers that are varying degrees of negative, what comes next? Does someone decide "yes we will" or "no we won't" implement this? Or is that already decided by the less-than-enthusiastic answers?

Examples

I might collect here examples of questions in which the author used some form of "doesn't work" and didn't specify what that meant. They idea is that they would have benefited from a prompt, asking them (without enforcement) to make sure they specified adequately what actually happened, or what expected behavior failed to happen. (If you follow these links, keep in mind the question may have been edited, so you might have to look at history to see how these are good examples of the problem and the solution.)

You're not really proposing a flag in the sense that the question gets flagged for moderator attention. Perhaps you could change the title to something a little more accurate/descriptive? — Cody Gray - on strike, Commented May 13, 2011 at 0:56
@Cody: thanks, I wasn't thinking about that sense of "flag". Can you suggest a better term? — LarsH, Commented May 13, 2011 at 1:48
Haha, yeah, I couldn't really think of a better phrase either, or I'd have just edited the title myself! :-) — Cody Gray - on strike, Commented May 13, 2011 at 2:52
@Cody It is exactly what I was asking. I deleted my question. My next feature request should be "improve the search" :/ — Aleadam, Commented May 13, 2011 at 12:25
@LarsH I wrote a feature request exactly like yours a few hours ago. I removed it because Cody linked me to yours. I fully support your request. Reading it, I'm amazed how similar this is to what I wrote: not only the request, but the whole body of the text! — Aleadam, Commented May 13, 2011 at 13:37
@Aleadam, thanks for your support. :-) Interesting that we both wrote such a request at about the same time. Which goes to show, I guess, that this problem bugs SO answerers frequently. — LarsH, Commented May 13, 2011 at 20:32
So this is one case where bad language helps... or were you planning on checking for "Why it no work?" ;) — Lorem Ipsum, Commented May 13, 2011 at 21:07
@yoda: If avoiding a mild question and getting annoyed comments instead of answers is "helping"... then yes. :-) But you bring up a good point... would it be worth checking for a question word order? — LarsH, Commented May 13, 2011 at 21:12
Now you've got me thinking of weird phrases that still match your regex... donot work, do n'twork. Dammit. Those are reasonably likely typos. Your regex is bulletproof! — Nicole, Commented May 13, 2011 at 21:19
@Renesis: :-) Actually the \s+ requires at least one space... but maybe \s* would be better. My first draft of the regex was more exact. I decided a looser one was better, and still wouldn't collect too many false positives. — LarsH, Commented May 13, 2011 at 21:40
@LarsH - Oops. Brain blink. In that case, it actually is slightly wrong, as it won't match doesn't/didn't/don't (no space) — Nicole, Commented May 13, 2011 at 21:42
FYI, another example of a user who just said "it's not working", but once I reminded him that this was insufficient info, he gladly gave more details: stackoverflow.com/questions/6764123/… This supports the idea that askers do want to supply the needed information; they would not just ignore such a reminder (at least if it was applicable). — LarsH, Commented Jul 21, 2011 at 19:56
@Bryan: interesting idea. Length of post could be a heuristic for displaying the message, although I have seen some long posts that never tell what the expected or actual behavior is. I don't think we want to disallow, because that's too hard a consequence for a detection method that will never be perfect. That would just drive people to fill in fluff (as we do for too-short comments). A reminder would be helpful though. — LarsH, Commented Jul 22, 2011 at 14:19

Pollyanna · Accepted Answer · 2011-05-13 02:09:26Z

37

I'm generally opposed to continually adding "trigger phrases" that encourage question askers to change their wording. Primarily because one could think of hundreds of such phrases, and it would be an unending parade. Further, I can post a perfectly good question with that phrase. The phrase alone doesn't correlate strongly enough with a bad question to ban its usage.

The community takes care of them pretty well, and honestly popping up red warning boxes will just encourage them to remove that phrase, not actually improve the question.

answered May 13, 2011 at 2:09

Pollyanna

76.1k41 gold badges270 silver badges479 bronze badges

2

You might be right. I don't think so though... if the warning boxes are capable of training somebody to avoid the phrase, then they have succeeded in (a) refuting @Cody's concern that users don't read; and (b) training people to be aware that "doesn't work" is a bad smell. If this happened to me, I think rather than go to the effort to circumvent smell detection, it would be not much more effort, and in my own best interests, to do what the smell is supposed to remind me to do: give people the info they need in order to solve my problem.
– LarsH
Commented May 13, 2011 at 20:24
7

Also, I agree it would generate some false positives. That's why it should be worded as a gentle question/reminder, and certainly not a "ban" on the phrase. E.g. I detected the phrase 'didn't work.' Did you tell us what happened, specifically? [Yes] [No] [What is this about?] I also agree that we'd never reach 100% accuracy with this detection. But if we got 80% accuracy with say 5 variants on the phrase, that would seem very worthwhile to me. No need to add variants ad nauseum with diminishing returns.
– LarsH
Commented May 13, 2011 at 20:29
5

The warning could also just be dropped at a certain repuation level where it is considerably safe to assume that people will not forget to add this information.
– brunnerh
Commented May 18, 2011 at 21:16
1

Agree with H.B. - IF this was implemented, it would have to be for low rep users only. "doesn't work" is not by itself indication of a bad post, it's impossible to judge without context.
– user159834
Commented May 18, 2011 at 21:21
@H.B. - good idea.
– LarsH
Commented Jul 21, 2011 at 20:03
5

I disagree. The phrase is a very good initial indicator of a poster who has not explained their outcome properly. It happens very frequently, and contains those exact two words 'doesn't work'. It really annoys me.
– Nick.Mc
Commented Mar 20, 2013 at 8:33

Add a comment |

Cody Gray - on strike · Accepted Answer · 2011-05-13 00:53:33Z

26

Hmm, I can't very well oppose this, because the phenomenon you describe certainly exists and is certainly annoying. It most definitely would be nice if we could figure out a way to "fix" this.

But I fear that your proposed solution will be fairly ineffective. The type of user who posts questions like those you describe (without any vaguely useful details, an adequate description of the problem(s), etc.) are the same type of users that will blissfully ignore any prompts to do otherwise. We've learned many times by implementing various features on this site that users don't read. And I can only imagine that the target audience here would be even worse about this.

answered May 13, 2011 at 0:53

Cody Gray - on strike

64.1k23 gold badges195 silver badges310 bronze badges

1

I'm sure there are some people like that, but often I find that when I ask OPs to elaborate on how something "doesn't work", they do... They apparently just didn't think about what information they should have given initially. And I can empathize. When you've been wrestling with a problem and your brain is tired, it's easy to miss information a person new to your problem would need. I think a friendly reminder would often help.
– LarsH
Commented May 13, 2011 at 1:56
3

@Lars: Yeah, people are more likely to read comments from other people than they are automatically generated text. We're trained to click through such messages without reading them, and as programmers, we really only have ourselves to blame ("Spelling check complete", for example). But I agree it probably wouldn't be that hard to implement, and if it helped any, that would be a step in the right direction.
– Cody Gray - on strike
Commented May 13, 2011 at 2:53
if(submitText.indexOf("it doesn't work") != -1){ alert("Nope, try again."); return; (my javascript sucks)
– user1228
Commented May 13, 2011 at 13:09
@Code Gray: I'm relatively new to SE, so please forgive me if that has been asked/tried before: amongst the implemented features you're talking about, has a form like those of bug tracking softwares already been tried? Kind of four textareas: 1. Description, 2. Reproduce Code, 3. Expected result, 4. Actual result. Nothing to read but cannot save w/o filling all four, or s/t like that?
– Jürgen Thelen
Commented May 13, 2011 at 22:25
2

@Jurgen: No, that hasn't specifically been tried. I would also oppose trying it. There are lots of good programming questions that fail to fit into that mold. It's a great fit for basic-level questions that are sort of akin to bug reports, but things start falling apart for more advanced questions likely to be asked by advanced users. While there's no question considered "too basic" for Stack Overflow, we certainly don't want it to become a beginners-only Q&A site. Besides, there's no guarantee that people would fill out those fields correctly. Look at any of your bug reports for proof.
– Cody Gray - on strike
Commented May 14, 2011 at 4:20
@Cody Gray: yeah, I agree, this wouldn't be a general solution fitting all cases. Well, it was just what I thought sometimes when looking at some first questions. Thanks for sharing your opinion.
– Jürgen Thelen
Commented May 14, 2011 at 8:31
1

@Jurgen: That really hits on the key problem with all suggestions to solve this type of thing. None of them that anyone can think of are very good general solutions. They always exclude someone that we want to include, or cause some other sort of problem that we wish to avoid. That being said, a bug report can be a very good model to have in mind when leaving a comment for the asker to clarify their question. Many of the things that a well-written bug report would include, you probably want to ask them to include in their question. So a good thought nevertheless.
– Cody Gray - on strike
Commented May 14, 2011 at 8:47
@Lars you didn't actually take that seriously, did you?
– user1228
Commented May 14, 2011 at 14:21

Add a comment |

casperOne · Accepted Answer · 2013-07-05 15:51:37Z

6

I feel that this request is now obsolete due to the new hold reasons on Stack Overflow.

Specifically, the following off topic close reason (emphasis mine):

Questions must demonstrate a minimal understanding of the problem being solved. Tell us what you've tried to do, why it didn't work, and how it should work.

Namely, if you or another person comes across someone who doesn't explain why it doesn't work, you should vote to put on hold using the above reason. This clearly asks the user "how exactly does it not work?"

answered Jul 5, 2013 at 15:51

community wiki

casperOne

Good to know about this. Not sure about that wording though. "Why it didn't work" could easily (most plainly?) be interpreted as "what the root cause of the problem was." And that's usually what the OP came to ask about. "How it should work" could be clearer too. I would rather see "Tell us what you tried to do, what the result was (how do you know), and how that differs from what the result you expected."
– LarsH
Commented Jul 5, 2013 at 18:20
Oops, I meant "... and how that differs from the result you expected."
– LarsH
Commented Jul 5, 2013 at 18:28
4

Moreover, it would be an improvement for both asker and reader if some such problems could be avoided at the start, instead of waiting for a round of intervention by other people. So yeah, the above helps, but I don't think it makes the request obsolete.
– LarsH
Commented Aug 28, 2013 at 13:47

Add a comment |

djechlin · Accepted Answer · 2013-04-29 14:58:40Z

3

As an experienced user it's really annoying when I have to kludge up the wording of a question I am asking that meets community standards because it tripped some regex that correlates with poor questions. Your solution is particularly bad because it's very reasonable to say something "didn't work" at the top of the question then elaborate below.

answered Apr 29, 2013 at 14:58

djechlin

14.6k5 gold badges42 silver badges67 bronze badges

3

Thanks for your feedback. I agree with your first sentence; I recently had to change a question title because "problem" was forbidden, and the linked page gave no help as to why that was intolerable or how to improve it. But notice that with my proposed solution, you don't have to change the wording of the question. I'm only proposing a suggestion/reminder. Regarding your last sentence, yes it is reasonable. That's why only "asking the user", not forcing the user, is proposed.
– LarsH
Commented Apr 29, 2013 at 15:07
@LarsH which is as annoying as a popup and no one will read. I didn't go into that because it's covered in Code Gray's answer.
– djechlin
Commented Apr 29, 2013 at 15:08
1

The question is not whether there will sometimes be false positives. That's already been established. The question is whether there are enough true positives, and enough benefit, to outweigh the inconvenience of the false positives. I've tried to demonstrate that there are.
– LarsH
Commented Apr 29, 2013 at 15:09
4

"no one will read" - can you substantiate that somehow? I read those things when they appear.
– LarsH
Commented Apr 29, 2013 at 15:10
@LarsH if someone skips all the FAQ, about, including the one we show them before they ask a question, then we give them more text that they may optionally click to see, before plugging their question into the magic Q&A machine, then no, they won't follow that link.
– djechlin
Commented Apr 29, 2013 at 15:10
@LarsH honestly no, I consider this common sense in UX and SE has been through this enough times to know better.
– djechlin
Commented Apr 29, 2013 at 15:12
Unfortunately if not everyone shares your "common sense" intuition, it doesn't help us figure out whether it's right or not.
– LarsH
Commented May 24, 2018 at 0:12

Add a comment |

bjb568 · Accepted Answer · 2020-06-03 13:30:57Z

2

The problem: Bad questions are getting asked, human intervention is expensive.

Your solution: Warn the asker before they ask, based on a list of specific phrases (in regexes).

Pitfalls:

Telling users that a phrase is "bad" will make them change their wording more than improve the question, changing which phrases are "bad"
Creating and updating a list of a bunch of phrases is a lot of work
Popups are ignored, especially by people who post bad questions
Popups are annoying, especially to people who post good questions who trigger these accidentally

Counter proposal!

Use a machine-learning system listening to specific close reasons, voting, and user information (rep, join time, net helpful flags, whatever that's statistically significant) to guess what problems a question has before posting. If the expected outcome is close with downvotes, block the question. If there's a chance the question will be closed, warn with a specific reason in a popup (without mentioning a phrase). This solves the pitfalls of your approach:

Doesn't tell users that a specific phrase is "bad"
List is updated by the system
Blocks aren't ignored
If you're an experienced user, the system will take that into consideration (besides, using statistics is more accurate so there won't be as many false positives)

edited Jun 3, 2020 at 13:30

community wiki

bjb568

Didn’t we try that already?
– icktoofay
Commented Jul 29, 2014 at 4:06
2

Developing a machine-learning system is easier than coming up with an approximate regexp? Blocking a question for unexaminable reasons is preferable to a warning? I just don't have the confidence in machine learning that you do.
– LarsH
Commented Jul 29, 2014 at 4:08
@icktoofay Didn't know about that, seems interesting. Can't download data, weird wabsite requires account.
– bjb568
Commented Jul 29, 2014 at 4:37
@LarsH Statistics, bro. "Blocking a question for unexaminable reasons is preferable to a warning?" Not unless you're sure (i.e. significant correlation). But my proposal is going to mention which close reasons based on which reason it is likely (i.e. significant correlation).
– bjb568
Commented Jul 29, 2014 at 4:40
1

I disagree with your premises. The proposal isn't to tell the user a phrase was 'bad,' but to remind them to check whether their question met good criteria. The work of creating a regexp is about done; maybe some tweaking is needed. When I get a popup like that, I pay attention to it to see what it's about. I don't find a popup annoying if the motivation is explained well, if it isn't coercive, and if it doesn't pop up unnecessarily very often. Some people will react differently, but I don't think a broad-brush dismissal without evidence is very credible.
– LarsH
Commented Jul 29, 2014 at 5:05
@LarsH I know what the proposal is trying to do, the negative consequences are just a side-effect. You may argue that the side-effects are rare, but they are still there. Also note that for you, this means something very different than for the crappy new users who ask poor questions - they think differently (not like I fully understand them either).
– bjb568
Commented Jul 29, 2014 at 5:35
I agree there will probably be non-zero negative consequences. Besides being rare, I argue they are very mild. I would like to see evidence that negative consequences from the machine-learning proposal would be both rare and mild. It sounds like it would be very coercive when it decides a post is likely a bad one.
– LarsH
Commented Jul 29, 2014 at 5:39
@LarsH Well, a human one is "damn, that's annoying, let's discourage/block it". A machine-learning one is "hmm, this pattern looks interesting... Oh! Questions with only one tag are downvoted an expected 3 times more, there is a 1% chance it is less than 1.5 times more and 3 times more was just by chance" - much more accurate.
– bjb568
Commented Jul 29, 2014 at 5:47

Add a comment |

Stack Exchange Network

auto-detect the phrase "it doesn't work" and ask user if enough info given

Update:

Update:

Examples

5 Answers 5

Counter proposal!

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged
feature-request
question-quality
.

Linked

Hot Network Questions

auto-detect the phrase "it doesn't work" and ask user if enough info given

Update:

Update:

Examples

5 Answers 5

Counter proposal!

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged feature-requestquestion-quality.

Linked

Related

Hot Network Questions

Not the answer you're looking for? Browse other questions tagged
feature-request
question-quality
.