On Stack Overflow, we have what some might consider to be a problem. (Many will argue how much of a problem this is, but I'm fairly certain most will at least acknowledge its existence.)

The problem is this: many of the older questions on Stack Overflow have answers that are either no longer applicable to the question (as technology changes and new software is released) or simply don't answer the question. Allow me to provide an example (which is by no means isolated):

How to doxygen comment Qt properties?

The accepted answer has a score of 7 as of this writing. However, a careful reading of the question will reveal that the accepted answer doesn't really answer the question. The OP states that "...I would like to achieve a similar output to this." Yet the accepted answer doesn't even come close.

I recently came up with a solution to the OP's question that actually does generate documentation that is nearly identical in appearance to the page he linked to. However, it is very unlikely that my answer will ever outscore the accepted answer. (It will also never appear above the accepted answer, but that's a different issue.)

The reason for this is simple: the existing answer has been around for over two years and mine has only been around for fourteen hours. Why is this a problem? Anyone viewing the question must scroll to the very bottom of the page to see the "correct" answer - and there's a good chance that they won't since the big green checkmark and many upvotes will likely convince them that this is the best answer. It gets worse if the question has >10 answers.

Here's how I propose to deal with this: decrease the weight of votes over time.

Now I fully realize that there are a lot of implications involved with this. One of the first questions that you are likely asking is whether the weight of votes should correspond with reputation. (In other words, should the user's reputation decrease over time as the weight of votes they've received diminishes.) The answer is: I'm not sure. There is a good case on both sides.

The benefit of a weighted vote should be obvious: as long as the existing answer continues to be the best answer, it will continue to receive new votes and continue to remain at the top of the list. If a new answer is posted and offers a better solution, eventually it will receive more votes and rise to the top.

Why is it so important that the best answer be on top? Well, when Stack Overflow first launched, Joel Spolsky had this to say when describing the presentation of answers:

"Very quickly, the best answers bubble to the top."

This has been the key that has made Stack Overflow so successful.

So... feel free to jump on my proposal and point out flaws, suggestions, or improvements. There's a lot that would need to be discussed before something like this could ever be implemented. I'd love to hear what everyone else thinks of this?

  • What about preserving historical context? Not all technologies progress. – Jason Sturges May 3 '13 at 21:30
  • 4
    This has the underlying (unintended?) assumption, that all that is old is bad. Or at least not as good as that which is new. I think that's too much of a generalisation to be workable. I get the problem you're presenting, but don't think this is the solution. – Bart May 3 '13 at 21:33
  • @Bart: That's fine - I welcome any criticism of my proposal. I did realize that there would be a few drawbacks. – uɐɯsO uɐɥʇɐN May 3 '13 at 21:34
  • 2
    You're 12k on meta so I'm more prone to take this seriously, but can you please make it more skimmable/tl;dr friendly? You know, summary, bold key points, anything to make it less wall-of-text-ish. Immediate reaction was to look at the title then be like, welp, this is going to be overly complex. – djechlin May 3 '13 at 21:37
  • @djechlin: I've made the key point bold. Does that look better? – uɐɯsO uɐɥʇɐN May 3 '13 at 21:39
  • Are you primarily concerned with sort order? – Jon Ericson May 3 '13 at 21:41
  • @JonEricson: Well... yes and no. I realize one can easily change the sort order but that doesn't affect the large numbers displayed next to each post. But I see what you're saying - that might actually be a better solution - the sort order is weighted by the age of the votes each answer has received. – uɐɯsO uɐɥʇɐN May 3 '13 at 21:42
  • With the post again active by your new answer, it has the opportunity to bubble to the top; to which it can compete with the best answer of the time. Personally, I don't have tunnel vision to the top answer only - usually there's more than one way to do something. – Jason Sturges May 3 '13 at 21:46
  • @JasonSturges: But this is not a popular question - aside from my mentioning it here, it would normally receive very few additional views. – uɐɯsO uɐɥʇɐN May 3 '13 at 21:48
  • 4
    In a few years' time, the votes on this request will no longer matter. – BoltClock's a Unicorn May 3 '13 at 21:49
  • I suggested something different to address the same problem once, and the claim was made that the problem doesn't exist at all and I've since come to believe that might be true. Does anybody have any hard data? Or do new correct answers bubble to the top eventually after all? From my anecdotal experience, I'd say it's at possible – Pekka supports GoFundMonica May 3 '13 at 21:57
  • Until someone upvotes your answer, no matter the sorting algorthm used, it cannot float to the top. – Aza May 3 '13 at 22:06
  • @Pekka웃: That only seems to be the case with popular questions, such as this one. – uɐɯsO uɐɥʇɐN May 3 '13 at 22:07
  • @Pekka웃 In the cases that I see, they don't necessarily bubble to the top. But 2nd place is very common. Here's an example: stackoverflow.com/questions/4382178/… The top answer no longer works on x64. The second answer came MUCH later and bubbled into 2nd place in a relatively short amount of time. – Mysticial May 3 '13 at 22:08
  • 1
    should the user's reputation decrease over time as the weight of votes they've received diminishes I certainly hope not. Having one's reputation decrease every day (unless they constantly add new content just to run in place) sounds awfully unpleasant. – David Robinson May 3 '13 at 22:53

I think I get what you are saying in terms of presentation.

Basically, at least, both of us want a repository of questions and answers that accurately reflects the current available information to those who see votes as a high indicator of correctness.

Say currently it's set at 7 votes (A) vs 0 (B)

Current time

Then at time 0 + t, we have three more votes for (B) but no extra votes for (A), a simplistic and not very well thought out initial draft would be that the 7 votes lose some kind of nominal value over time.

Several months later with three votes added to second answer

Simplistic weighting

Assuming we did get a good weighting going on, it still isn't so good in numerous ways

  • Presentation wise, the community has gotten used to the idea of linear voting over time, so this is going to throw everyone
  • Does this mean that that presentation should carry on to actual representation? Does these 7 votes become of less value to a user's overall score
  • How do we determine when to make the switch over, for example if user C now comes along with an answer two years which trumps both. Does the weighting scale stay the same?
  • Maybe we should be introducing a separate factor to denote votes on questions that scales appropriately to reputation for users

I'm agreeing I just don't have the statistical/math background to show the proper pros and cons to this idea. I think if it was given the proper research at the very least if the idea is rejected, it is because statistically and reputation wise it becomes a lose-lose situation.

It really does seem like an interesting idea, we could at least think this out analytically without subjectively declining it.

  • What if my proposal changed a bit and instead became exclusively about sort order - the number and value of votes displayed remained the same but the order of answers reflected the age of the votes? – uɐɯsO uɐɥʇɐN May 3 '13 at 22:05
  • Maybe though the ordering is not going to look natural against the votes. Because of how the votes are seen I'm sure they play a big factor in what users think are "correct" – phwd May 3 '13 at 22:23
  • 1
    Precisely why I'm concerned - votes don't necessarily reflect a quality answer. – uɐɯsO uɐɥʇɐN May 3 '13 at 22:24

I think your idea of the fundamental issue is incorrect. We can categorize all questions under this focus as questions with incorrect upvoted or accepted answers, regardless of age. Effectively, the age of a question is irrelevant to this matter - except that an older question will be viewed with less frequency (on average).

Still, the set of all questions under this domain is a subset of the set of all questions whose upvoted/accepted answers are incorrect. I offer that incorrect acceptances and upvotes are the problem, not the age of the question with respect to a new answer.

However, this is not a broad subset, and it doesn't seem to me like this occurs very often. Fact of the matter is, any question which is explicit enough and which meets good question criteria will receive an answer which is relevant to the question itself. Even if the question isn't very explicit, the probability that it will become heavily sought after is low.

Additionally, a decay in upvotes wouldn't be very useful. It would allow potentially wrong answers to jump above very good answers, given enough time. It would result in strange sorting, and there wouldn't be all that much benefit, as the set of questions this would be useful for is so small.

So, to summarize: The issue here is not old upvoted/accepted answers. That issue is a very small subset of many broader issues. It's fallacious to assume that addressing this issue will help address any other issue, since the impact of this would be minimal. Finally, anyone actually using a question to help them is going to look at all of the answers, not just the first one, so if the first answer isn't helpful, the order probably won't matter.

You must log in to answer this question.

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