How big a problem are unanswered questions?

Question

Do the powers that be consider 180k unanswered questions a problem that needs addressing? How do they envision the current system clearing out that backlog? (Maybe there are some other mechanisms I'm not aware of as a relative new-comer). I want to get a sense of what the operators as well as the veterans think about this before I start lobbing feature requests around.

To me it seems like there's a lot of junk in SO's unanswered questions tab. A couple times I've tried to go through and clear a few out by answering or up-voting, but the signal-to-noise ratio is pretty low. Even if I was >3k, would enough other people come along to vote to close something before my close vote expired? I think some heavy culling would help answerable unanswered questions stand out and make the whole thing more useful. I know the goal is to be the repository of all programming knowledge, but perhaps the scales could be further tipped from quantity to quality?

One limitation of the current system is that the only people who are really good at answering questions can moderate. Even if those people are willing to spend lots of time cleaning up, I think the Jon Skeet's of the world are probably a lot more valuable to the site when they're answering questions and not playing janitor. But obviously we can't just let anyone come in and close or delete questions, which is why I'm feeling like some sort of automated culling would be beneficial.

I'm also curious about something that maybe one of the gurus can come up with a db query to answer: is the number of unanswered questions growing faster than the pool of people that can moderate them?

---

Did some more digging. Looks like % of questions that have answers of non-zero score has gone from 93% at the start of 2009, to 83% at the start of 2010, to 74% at the start of this month. At some point, that has to become a problem.

% of SO questions with voted answers through a given date

Answer 1

Some stats from Stack Overflow

Questions with no answers:
75,422 / 1m+

Questions with no answers and score > 0:
12,705 / 1m+

Questions with no answers and score < 0:
2,016 / 1m+

Questions with no answers created more than 30 days ago:
64,717 / 1m+

Questions with no answers created more than 120 days ago:
39,915 / 1m+

Questions with no answers and score < 0 created more than 30 days ago:
1,631 / 1m+

oldest zero answer question in the system:
https://stackoverflow.com/questions/87110

Based on this data, I am only comfortable auto-deleting the negatively voted, zero answer questions more than 30 days old. These are basically complete crap, and they're a tiny minority of questions anyway.

Answer 2

It'd be nice if they didn't base that tab on an answer not being accepted, and just had real unanswered questions there. Maybe have two different tabs.

Answer 3

On Ladybuy's suggestion I'm answering myself for voting goodness. This one is "phase one". A minimal proposal with hopefully minimal controversy. It may not be enough to solve the whole problem but it's a start:

The most obvious candidates for auto-deletion are old questions with no answers and score <= 0. If those were deleted after a month, it's extremely unlikely anything of value is lost, and if the question was any good it can always be re-asked, hopefully in a more answerable form.

Perhaps also include questions with score <= 0 where all answer scores are < 0 (not sure if this is a big group)

Edit: After doing some odata exploring, we could probably ignore the question score and just delete anything older than 30 days that has no answers posted. This is currently about 7,000 questions, vs ~5000 with score <= 0. I haven't figured out yet how to inspect the vote scores of answers.

Answer 4

This is meant to be helped by the tumbleweed badge. But it appears that incentive is not strong enough.

Stats

There are 7,700+ questions with no answer, which is less than 1% of the total number of questions on the site.

There are 181,000+ questions with no upvoted answer (or, 174k questions with an answer that has not been upvoted). This is 18% of the total number of questions on the site.

Is it a big problem?

No. They have answers. Fewer than 1% have no answer, which is not bad. The 18% that have answers but are not upvoted - well, that's a consequence of getting 2 thousand new questions a day. There's no great incentive for people to upvote.

If someone wanted to prove the case that this is bad, they would have to show that a significant number of those questions are truly unanswered - in other words that the answers provided are wrong and unhelpful. The few that I've gone through have answers that either answer the question, or at minimum point the author in the right direction for debugging (ie, not enough info given, here's a few things to try, but the author never responds or gives feedback).

Should anything be done?

I think that an incentive to go through those answers and upvote/downvote/close would be appropriate. Tumbleweed provides incentive to answer, so if a question has no good answer, but is otherwise reasonable, I'd say leave it alone.

A badge for getting questions off the unanswered list would probably cause problems (people upvoting answers regardless of their quality).

I can't think of a good way to give people more incentive (other than tumbleweed) without actually making things worse.

But the problem isn't bad. It's not ideal, but so far no one has convinced me that the issue is a real big problem.

Answer 5

Alternative approach: create a set of "Garbage Collector" tools for people with rep>=3k that shows which questions have active close votes so that other people can find them and help close them within the 4-day window if they think that's appropriate. Maybe it would just be another tab on the main page for people who have the necessary rep. Being able to see questions by new users might also be helpful so that questions that need to be migrated an be found and moved quickly.

Answer 6

You make some valid points here. But if automation can solve them, is questionable.

What should be the rules for this automatism and how can we figure them out? You can do that yourself. When scanning the unanswered questions, look at the posts, which should be "culled", think about why they should be culled and if you see a pattern. After that flag for moderator attention.

Yeah, you've read correctly. I doubt, that you come up with a useful pattern. But if you have found some of these questions, it would be a shame to leave them there. The mods can care about them, that's why we have elected these lazy bastards. To clean up the site.

Answer 7

If I've understood this correctly, bemace's problem is that the "unanswered question" list is full of old specific questions with no upvotes nor answers. And that this makes it difficult to find questions to answer in the unanswered questions list.

Instead of automating deletions of such questions, I propose it should be possible to ignore/hide specific question threads, in addition to tags. Introduce a "hide" button to questions, and show the hidden questions as a list under your profile settings.

This might be difficult to implement though. I'm just tossing ideas around, since I think deleting posts is slightly intrusive, when the question is simply unpopular, and not necessarily badly phrased or impossible to solve.

Bemace has another good point though; StackOverflow needs more moderators! And not necessarily all with the same powers/privileges. I'm 100% behind this.

Answer 8

Possible phase 2:

Do something with questions with score <= 0, no accepted answer and no answers with score >= 0 after three months

If I did my query right, this is about 44,000 questions, which seems like too high to me (as in, not an accurate count. Are things really that bad?).

http://odata.stackexchange.com/stackoverflow/s/609/qs-with-no-as-0-after-three-months

This query will let you browse the affected questions that actually have answers so you can judge what sort of treatment they need: http://odata.stackexchange.com/stackoverflow/s/610/view-deadweight-phase-2-questions

These questions could simply be removed from the Unanswered tab into a new status called "Expired". Wouldn't need it's own tab, questions could just be accessable by searching.

Answer 9

I think these unanswered questions fall into different categories and should be looked at separately:

1. Bad questions
2. Complicated, specific questions

While no one cares about getting rid of the first category, it would be nice to have some extra motivation to get answers for the second category. The complicated questions can probably only be answered by a bunch of experts and/or take some major to answer. They probably only got low attention, because they're only interesting and relevant to a small group and not the broad mass. And as the site grows, chances are, they just were not seen by their interest group.

So they ended up where they are now, and get digged out and answered only rarely. Why that? The chances are high, that the big effort put into answering them won't even be rewarded at all. The author itself is not expecting an answer anymore, probably not actively taking part in stackoverflow anymore, and so the chances that he accepts the answer and the reputation is granted are low. The question is also hard to find for other users, so upvotes for the answer will stay rare.

Suggested solution

• Questions older than one month should also be accepted automatically. Every new answer to a question with no answers after one month should trigger an auto-accept feature. If there are no other upvoted answers within - suggestion - one month, the answer with the most upvotes gets accepted automatically, maybe even with no upvotes at all. This makes sure, that no efforts on these questions go unrewarded, while still keeping quality of answers high by allowing competition for upvotes.
• Questions with active auto-accept triggered should probably be visible in some separate category to increase competition for an answer.
• There should also be no punishment (or at least further reduced punishment) for down voting those old stuff. This allows people searching for some material to answer, to sort out the trash at the same time. This should be enough to solve the problem with the bad questions itself, when they are autodeleted later.