-6

Currently the number of votes to delete a question is varying based on the number of votes on it and its answers. It doesn't appear to take views, favorites or 'links to' into account. These are important measures of how popular a question is, and should be factored into that. There are many old popular questions that are being deleted haphazardly, even though they have a lot of popularity and draw many to the site.

Please add views, favorites and 'links to' into the delete-votes needed algorithm in some way.

Note that I'm not saying just more=more, but that we should do some data-mining to determine the correlation between those and popular, useful questions.

If you want to see examples, just go to this question and look at some of them. I just ran across Getting started with Haskell, with 993 favorites and 91k+ views, but the votes algorithm wasn't enough to save it.

Improve this question
12
  • 3
    So there are many old, great , favoured questions with little upvotes? Never seen that. Have you a few examples of these haphazardly deleted questions?
    – juergen d
    May 27, 2014 at 13:46
  • 7
    Popularity != utility. Throwing a roadblock on site cleanup because a question has a ton of meaningless favorites or views is counterproductive to actually doing the job.
    – fbueckert
    May 27, 2014 at 13:49
  • 1
    @juergend, not off hand, though obviously they're hard to search for. The point is that votes aren't enough, and many of these keep getting deleted.
    – Lance Roberts
    May 27, 2014 at 13:49
  • @fbueckert, and purity doesn't help the site either. If a question is popular, then it was useful to many people, even if it wasn't useful to you.
    – Lance Roberts
    May 27, 2014 at 13:50
  • 2
    How does popular equate to useful? I can ask an absolutely braindead question about a common issue. It will probably be heavily downvoted due to zero research (and rightly so), but will have a ton of views, due to it being such a common problem. I still say that it's not useful.
    – fbueckert
    May 27, 2014 at 13:52
  • 1
    @fbueckert, I'm not saying that mass amount of views should automatically generate higher vote counts. It's an algorithm and should be made after some data mining to determine what does make a question truly useful and popular.
    – Lance Roberts
    May 27, 2014 at 14:04
  • 3
    examples here would help clarifying
    – Sklivvz
    May 27, 2014 at 14:08
  • 1
    And I'm saying, views != utility. Just because it's been looking at eleventy billion times does not automatically mean it should require more effort to delete.
    – fbueckert
    May 27, 2014 at 14:08
  • 6
    I would like to see inbound links added to the algorithm. Views and stars, not so much. I think upvotes and stars are just a function of views above a certain threshold, so there's really no extra information added. You only need one of those parameters.
    – Bill the Lizard
    May 27, 2014 at 14:12
  • @BilltheLizard, that might be all that is needed. A little data mining would show us what really correlates.
    – Lance Roberts
    May 27, 2014 at 14:29
  • And don't forget to add a negative factor, making deletion instantaneous if somebody incidentally mentions biology.
    – jscs
    May 27, 2014 at 18:59
  • 2
    +1 for the suggestion after looking a some data. See my answer below.
    – Bill the Lizard
    May 28, 2014 at 16:04

1 Answer 1

Reset to default
5

After looking at some data, I'm going to have to walk back a statement I made earlier in a comment.

I think upvotes and stars are just a function of views above a certain threshold, so there's really no extra information added.

That might not be the case. Views, votes, and stars are certainly positively correlated, but not as strongly as I would have guessed, and using only one of these parameters might not tell the whole story. (For example, a question might have an inordinately high number of views, but relatively few upvotes and stars. This might indicate that the question is linked heavily from outside the Stack Exchange network, so a lot of viewers were unable to vote on it.)

Here's a Data Explorer query for Views, votes, and stars for Great Questions. If you plot the data, you can see that a positive correlation exists, but not a particularly strong one.

Votes and Stars vs. Views

Correlation
Views and Upvotes: 0.529
Views and Stars: 0.364

Conclusion: It may be worth it after all to include all three parameters in the "delete votes needed" algorithm.

There are caveats.

  • This is a very rudimentary analysis. You might get different results if you looked at questions in different ranges of votes. I'm only looking at those questions that have a Great Question badge, since they would be the most controversial to delete.

  • This data does not include deleted questions. It also ignores whether a question is currently open or closed. These impact whether or not you even can cast a delete vote on a question, so looking at a more specific data set might yield different results.

  • This is not a labelled data set. To do this analysis right, you'd want to compare questions that were deleted (correctly) to those that are not to find which parameters can be used to best determine which questions should be harder to delete.

Improve this answer

You must log in to answer this question.

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