Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 158 Stack Exchange communities.

What is meta?
Here's how it works:
  1. Any Stack Exchange user can ask a question
  2. The community provides support, votes on ideas, and reports bugs
  3. Your voice helps shape the way Stack Exchange operates

I was just looking at a question on SO, something along the lines of "how long does it take to multiply two numbers," and observing how most of the answers were variations on the theme: "It depends on your system. You have to measure it." It occurred to me that there are at least a few questions like this, that get asked in various forms over and over again, and are always met with the same set of answers. Wouldn't it be neat if, when a user submitted such a question, SO itself posted a default answer.

For example, if the question contains the words, "parse", "HTML" and "regex", post an automatic "You can't...regular to insane bobince post..." answer.

It would be a good way to let users know when they've asked a question that's not exactly a duplicate, but that's been answered in principle over and over again. Of course the community could up vote or down vote the default answers.

I'm sure it's a ridiculous idea that would be nearly impossible to implement well. And it probably goes against the whole philosophy of the site. But I thought it might be an interesting problem to consider at least.

share|improve this question
We already have a sentient robot. We call him Jon. – Robert Harvey Aug 2 '10 at 18:59
up vote 3 down vote accepted

Automated Question Parsing

Both Richard and Grace dealt fairly well with the troubles of any sort of automated system which would attempt to auto-answer simple questions.

So I'll ignore that particular problem and assume that we (the collective genius of StackOverflow) have solved that problem.

Parsed Question -> Coherent Answer

The problem with common questions is that they are so very... common. Not only are there lots of root questions (how to multiply) there are enough instances of the variation questions (in js, C#, COBOL) to have each of those question flavours also considered common. Except that a common answer would not be appropriate.

If I ask how long does it take to multply two numbers in python, a general answer isn't that helpful, and a specific C# answer is very unhelpful.

The point here is that even truly common questions will have specific details built in which are less common. So rather than 10 common questions with 10 pre-fab answers, we'd have 10 common questions with hundreds of answers, one for each flavour.

But, this is merely an extension of the parsing problem, although as I've illustrated it is still distinct. We'll once again assume that we (the collective genius of StackOverflow) have solved this problem as well.

Coherent General Answer -> Answer For Me

There is a technique known as Rubber Ducking which basically refers to when telling someone about your problem allows you to solve it yourself.

That is an idealized situation, the most common (but related) one is that when you tell someone, you may solve your own immediate problem, but realize or discover a much larger problem.

In these cases, you'll have follow up questions, your problem isn't "solved" even though the naive case has been.

In these cases, I have a common question that is the root, but there is a much more serious problem existing in the question text. So, the automated answer robot (we'll call her Compendia) will post a generic response, completely ignoring the warning signs in the question. The OP will add a comment asking for further details after having edited their question, or more likely, simply saying

"It didn't work, now I'm scared. Hold me." -Sincerely, The OP

At this point, there is no one looking at their question. The regulars will have learned to avoid questions that look 'typical' because Compendia will answer them all and so there's no gain in bothering. The original answerer cannot add further information because she's not real people.

Basically, this will lead to a lot of unsatisfactory answers to "common" questions, and since common questions are the most... common... this would end up having a significant detrimental effect on the experience for nearly everyone on the community.

This answer was an automated response based on keywords in your question. Please do not respond.

share|improve this answer
Please do not respond. — Okay. – detly Aug 3 '10 at 14:39
Excellent answer. The idea is even more absurd than I imagined! – jmgant Aug 10 '10 at 18:38
@jmgant It's wasn't that absurd, it just needs some improvement :) – devinb Aug 10 '10 at 18:49

I have this simple HTML page, it uses regex to parse a DateTime string in a textbox, and...

There is so much potential for error when you try to perform any task related to parsing the text of a question to try and "guess" what it is about. Not everyone uses the same words to describe the same errors, which is why we keep getting a lot of duplicated content in the first place. It becomes impractical to try and account for every possible variation, because doing so will increase the number of catastrophic intersections with legitimate questions.

Even the tags are subject to this error, when the tagging is done by an inexperienced user. Interpreting questions is best done by a human element, not an automated one.

share|improve this answer
+1, you're right, just including a list of words wouldn't help; there would have to be some kind of semantic analysis. (Like I said, probably nearly impossible to implement well.) – jmgant Aug 2 '10 at 19:06

I'm sure it's a ridiculous idea that would be nearly impossible to implement well.

I think auto detection/answering is not feasible; as such, it would have to guess what "common question" you asked; better yet, it could suggest it, perhaps even offer a list of suggestions, and if the asker sees a related question which fits the bill they could check it out.

But SO already has that functionality ;)

share|improve this answer
Good point; I'm sure the existing duplicate-checking logic was very hard to implement, and its success rate is nowhere near 100%. – jmgant Aug 2 '10 at 19:09

You must log in to answer this question.

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