42

Due to its massive (and growing) size, there are a lot of low-quality questions and answers appearing on Stack Overflow.

I'm concerned that an influx of low-quality questions and answers create a lot of work for our users and moderators and can potentially damage our expert Q&A ecosystem.

Is there some way we can prevent these low-quality questions and answers from appearing on the site?

So:

  1. What current processes are used to block these low-quality questions and answers?

  2. If the current processes are inadequate, how can they be improved?

1

6 Answers 6

24

We already do this for answers.

We now have some basic heuristics in place to check for the short answer, and the short “thank you” answer. If the answer is by a new user and matches enough of the heuristics, we provide automatic guidance — as originally suggested on meta by Kop — in the form of a How to Answer page .

But, don't take my word for it. I STRONGLY recommend that you fire up Google Chrome in Incognito Mode and try posting this as an answer on Stack Overflow. Right now. On the live production website. No, I'm not kidding -- try it. G'wan. Go. Do it. Seriously, do it!

Here's what I want you to post as a new user "answering" a question. And remember you must be in Google Chrome Incognito (aka new user) mode:

i have the same problem!!! can anyone help?

Not enough? How about some visual proof; I'll do it for you. Here's me entering this as a new user ..

Here's what happens after I click Post Your Answer; I am required to read and agree with the How to Answer page.

So, as you can see -- this already works for answers.

This setting is fairly safe, since it's triggered by a heuristic on answers -- we have a pretty high confidence level that when this heuristic is triggered, the odds of the answer being a bad one is high. To get an idea of what things trigger the heuristic, click on the "low quality posts" tab on the review section:

https://stackoverflow.com/review/low-quality-posts?pagesize=15&filter=day

The heuristic value is indicated at the bottom left of each post. Lower is worse, higher is better.

This answer quality heuristic check is currently enabled on all our "large" sites like Super User, Server Fault, Programmers, Ask Ubuntu, etc.

After reaching some threshold, an account may be blocked from answering, showing "Sorry, we are no longer accepting answers from this account".

20

We already do this for questions.

Every new user on Stack Overflow is REQUIRED to click through the How to Ask EULA before posting a question.

Don't take my word for it! Enter Google Chrome incognito mode and try clicking "Ask Question" on Stack Overflow and see for yourself what happens:

How to Ask help page

This setting is fairly invasive; it is triggered universally on clicking the Ask button for all new users. That's why it is only enabled on Stack Overflow for now -- when you get 3k+ questions a day, you can afford to throw 10% of the worst ones away.

Also, there are certain quality filters applied to questions to try to ensure a clear title, a reasonable explanation of the question and correct use of English and actual sentences. When the tests fail, the post is rejected.

Finally, there's an automated question ban for IP addresses or accounts with a history of extremely poor questions, partly based on question votes. See What can I do when getting "We are no longer accepting questions/answers from this account"? for some more details.

8
  • 6
    Someone wrote an excellent blog post on the effectiveness of forcing users to click "Yes" ;)
    – Andomar
    Commented Mar 24, 2011 at 23:11
  • @andomar if only 10% of users read that, it's still a win over not doing it. That said, we only enable this on sites that get ABSURDLY LARGE numbers of questions -- it is a game of statistics. Commented Mar 25, 2011 at 0:10
  • @Jeff Atwood: Won't users who read that have 5 minutes less to spend on their question? Did you measure the effect of the page, in terms of how long it took for the user to check the box, and what impact that had on their questions?
    – Andomar
    Commented Mar 25, 2011 at 11:28
  • 1
    @andomar well, for the answer check it is Answer Eulas Invoked: 24709 Answer Eulas Ignored: 11595 Commented Mar 25, 2011 at 21:33
  • 7
    @Jeff Atwood: Well, the useful query would be flags/downvotes/upvotes related to invoked/ignored/noeulaatall. I've added "If anyone reads this, mail me, and I'll give you ten euro" to one eula-like popup. It took a year before someone noticed it (and that was a fellow developer.)
    – Andomar
    Commented Mar 26, 2011 at 12:23
  • @andomar look up "learned helplessness" Commented Apr 10, 2011 at 7:55
  • 2
    @Arjan: The "is rejected" link leads to a deleted question. Commented May 14, 2011 at 6:20
  • we all saw how well that went in the famous Southpark episode
    – vsync
    Commented May 23, 2016 at 15:07
11

We now have dynamic answer help that forcibly pops up for new users (< 100 rep) when they focus the answer box.

enter image description here

Now with a proper link to /help/how-to-answer as well

1
  • This I like. I remember having read a post somewhere that claimed that prompting users on the specific thing at the right time is more effective than having them to read a huge list of guidelines in one place.
    – John Red
    Commented Jul 29, 2017 at 10:12
6

How about breaking up the asking a question process?

Step 1

Enter question title

Step 2

Review a list of similar questions, press button that says "None of these answer my question". You can make the page a lot more visually appealing than the current iframe style box.

Step 3

Enter question body

2

There are two possible approaches to dealing with poor quality questions:

  1. Prevent the poor questions/answers from being entered
  2. Don't show the poor questions/answers unless we ask for them

As Jeff Atwood's posts show, a lot of the former is already done, so let's talk about the latter:

Ideally, a poor question would get less attention on the site. In practice, most of the standard question views do not appear to discriminate against poor questions, even though Stack Overflow has a couple of indicators that ought to correlate with question quality, such as the history of the asker, size of the bounty, number of upvotes and downvotes (corrected for the number of question views).

0

Taking a page from acadamia.

What about an option that allows you to flag an answer for peer review? That allowed you to flag question’s for review (maybe better from a technical perspective). Adding some specialized voting for questions/topics-area's by senior people may help offset the tendency of vote's only originating from unskilled/less-aware people (inherently, that's the idea, you don’t know, that's why your here).

Most of the time, sure, the high voted question is the correct/ideal one. But for some domain's (deeply low-level) the specialized knowledge is not common and is very rooted in fine details. A general, “almost” there answer may be good enough to satisfy the 80%, it may be average quality, but not high or the fabled 20% great quality.

Having some mechanism to request peer review rather than simple helpful votes would go a long way to separate to good from the great or fluffy/squishy from the hard and true.

Peer reviews could take into consideration objective criteria to designate a preferred approach; the first answer, the most technical, the best/most complete examples, new/novelty...

A peer system seems a good supplement to the current, neophyte voting system (only those that do not know are the one's voting). This enable make an effective way (points & medals) to counteract the snowball effect (high voted questions gaining momentum are hard to sway) and the presentation phenomenon.

2
  • Think about the case where several answers come in in quick succession, the one at the top has a natural advantage from day 0, that advantage, over time, means that it will rate higher (Vegas anyone?). This has happened to me by a ratio of 4:1 or more for answers with fluff, assumed fact's and such do amazingly well despite being of low-quality. If peer review is too hard to implment, how about randomizing equivielently rated answers (all 0 voted ansers are randomly shuffeled instead of time basd)? Commented May 14, 2011 at 10:43
  • If peer review is too hard to implment, how about randomizing equivielently rated answers, all 0 voted ansers are randomly shuffeled instead of time basd, shouldn't the newest answer be on the bottom, the late answer allready, the last answer allready has the advantage of being able to read all the other answers (which is big), why amplify their status by giving them the top page rank (rank=order)? (p.s. I edited this after the 5 minute mark, sorry for the 1 sentance dup). Commented May 14, 2011 at 10:49

You must log in to answer this question.

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