What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 131 Stack Exchange communities.

I suggest that SO uses low votes per view as a reason to to raise questions for review.

Each time someone views a question, there is a chance they will vote it up, or vote it down. Someone that searches for questions on a particular subject and is lead to a question that is irrelevant to what they are interested are unlikely to up-vote the question, but if the question is not inherently bad they are unlikely to down-vote it.

A question with many views but few votes is therefore a question that superficially seems relevant to many people, but in fact is not. I suggest that questions with overly general titles are particular likely to be in this category. Let us call them click-bait: the title in some search-engine results (including a SE "related questions" list) draws the visitor in, to click on the link... but the visitor wastes their time. Those questions are a kind of low quality question.

I suggest that SE automatically searches for questions that have had quite a few views, but a low votes/views ratio, and adds them to the "Low quality posts" queue, so they can be edited to have a better title and/or tagging.

share|improve this question
8  
There's the seed of a good idea in here. It would be better with data to show an algorithm that works (or at least gets close). –  Bill the Lizard Jul 9 '13 at 12:13
    
@BilltheLizard Perhaps someone with access to the back-end could produce a scatter plot of votes against views? Or better still a plot of votes/views against views. I suspect that votes/views tails off for highly voted questions.. –  Raedwald Jul 9 '13 at 12:23
1  
And what about all the anonymous traffic? Some questions might attract more external visitors than SE users, all of which are not able to vote. You'd have to factor in the anonymous feedback here as well. –  slhck Jul 9 '13 at 12:25
    
Votes and views are both available in the data explorer (see the schema sidebar). data.stackexchange.com/stackoverflow/query/new If I have time later I might take a look at this myself, just to see if there's a clear trendline. –  Bill the Lizard Jul 9 '13 at 12:25
1  
Prioritizing the posts in the review/edit pool is, in my opinion, a very urgent addition... This is actually a very good idea! –  Joum Jul 9 '13 at 13:34
1  
data.stackexchange.com/stackoverflow/revision/123511/151171 ... there are 1248 posts with > 10k views and a score < 5 in the last 1.5 years, that are not closed. 10k views seems like rather a lot for a < 5 score post, no? Though I suppose this depends somewhat on the tag? –  Joe Jul 9 '13 at 14:34
    
@Joe Nice. You can use select id as [Post Link],... to display a link to each post in the data explorer result set. data.stackexchange.com/stackoverflow/query/edit/123513 –  Bill the Lizard Jul 9 '13 at 15:05
    
Thanks for the tip! Updated... data.stackexchange.com/stackoverflow/query/edit/123508 –  Joe Jul 9 '13 at 15:18
1  
I would note that I looked at a few of these, and I think it's going to be hard to define the criteria for this appropriately. Too many of these are commonly searched things by people who wouldn't vote - lots of javascript stuff, for example, where the viewcount is going to be insanely high no matter what among non-voters. If we could segregate views by who viewed, that might be more helpful, but I don't see anything obvious I'd change about the top few I looked at. –  Joe Jul 9 '13 at 15:22
1  
I'm with @Bill here; at the moment you have a list of questions, which titles give good SEO but where the questions or answers were bad enough that nothing ever came of it. It seems to be a manual deletion job... (see Bill there was a reason I CC'd you :-). –  ben is uǝq backwards Jul 10 '13 at 6:08

2 Answers 2

The problem with this suggestion is that there is a class of questions that is not likely to be downvoted or upvoted yet is high quality: Questions about specialized or difficult subjects.

They won't be downvoted because they look like effort posts.

But, they won't be upvoted because the viewer doesn't have enough knowledge in that field to determine that it's a meaningful question.

share|improve this answer
4  
but will they have high view count? –  Jan Dvorak Jul 10 '13 at 5:16
    
@JanDvorak I'm assuming so - people will click because 'Ooo, it sounds interesting' or 'Ooo, an unanswered question, time to earn me some rep' –  Patashu Jul 10 '13 at 6:08
1  
Questions about specialist subjects will have a low view count. Specialist questions that have a title that sounds interesting to many non specialists have a bad title and should be edited. –  Raedwald Jul 10 '13 at 7:19
    
@Raedwald - in effect, you are proposing to make titles less clear, aren't you? –  Deer Hunter Jul 10 '13 at 8:09
1  
@DeerHunter No, the opposite: have the title accurately reflect what the question is about. –  Raedwald Jul 10 '13 at 10:41
2  
But we're only talking about putting them in a review queue, not closing or deleting them, so no harm will come to good questions like these. –  Bill the Lizard Jul 10 '13 at 11:17

I'm emphatically against that. Here's why:

  • There are popular questions for problems that occur every now and then. They are fine by your definition.
  • There are also long fat tail questions that fit into narrow niches on 'esoteric' topics. While many may have a look at them, not all (an understatement!) may be qualified to judge the quality of either the question or the answers. Your metric effectively discriminates against those. If anybody searches for key terms, she will be led to this site and will have her problem solved, without being forced to register and gain enough reputation to upvote. Your metric kills this part of service to the general public provided by Stack Exchange.
  • Vote per view ratios are going to be significantly different on different SE sites, and they also may evolve during sites' life cycle (alas! have no hard data to support that handy, and too lazy to concoct a query to SE data). How will your algorithm handle these difficulties?
  • Another review queue will simply clog mental pipelines of sites' reviewers. End result - reviewers' fatigue and mass apathy.
share|improve this answer
    
How is it possible that an esoteric question gets many views, unless the question has a bad title that tempts people who are not really interested in what it is about? –  Raedwald Jul 10 '13 at 7:23
    
A new review queue: the existing low quality posts queue could be used. –  Raedwald Jul 10 '13 at 7:24
    
Different vote/view ratios: if the criteria is "in lowest x% of votes/views AND ..." that should be OK –  Raedwald Jul 10 '13 at 7:26
    
@Raedwald - re: lowest x% criterion is not objective, nor is it absolute. Just imagine you've reviewed the lowest x% of the questions - what's next? You'll have to do the same again and again. –  Deer Hunter Jul 10 '13 at 13:57
    
Is it possible to identify views by someone who is logged in, versus not logged in? I'm thinking if someone is logged in, but takes no action on a topic, that might be informative. –  ToolmakerSteve Jul 10 '13 at 14:02
    
@ToolmakerSteve - we cannot look inside the brains of readers. You'll have to account for those who are able to upvote (15) but wouldn't, and separately (125) for those who can downvote. A stagnating, apathetical site or a high-turnover site will have more lazy viewers, an enthusiastic community will upvote more easily. A rat's nest of latent variables, if you ask me. Practically unsolvable as far as one-size-fits-all formulae are concerned. –  Deer Hunter Jul 10 '13 at 14:34
    
As a practical matter, I think @Raedwald has the right idea. Why does the criteria need to be either objective, or absolute? If all it is doing is putting the questions on a review list? That has no AUTOMATIC effect on the visibility of those questions, to the ordinary user, right? Just periodically, it goes thru all the questions, and makes a list of those at the "bottom", by this criteria. Why isn't that useful? If specific questions appear in the list, and reviewers think those questions are worthwhile, then their MANUAL indication in favor of the question would take precedence. –  ToolmakerSteve Jul 10 '13 at 15:00
    
@ToolmakerSteve - all right. What we need is hard data from a couple of Data SE queries. Could you collect stats on the distribution of Raedwald's metric for all SE sites accessible through Data SE (and publish the query links here)? After that, we can go through a random sample of questions with low Raedwald's metric to see if it makes sense. –  Deer Hunter Jul 10 '13 at 15:06
    
"what's next? You'll have to do the same again and again" not if one of the actions that a review is encouraged to take is "upvote this question if it is good", and another is "edit the title of this question" to make it more specific, and recently edited questions are not listed as needing review. –  Raedwald Jul 10 '13 at 15:45
    
@Raedwald - you see, there are already several variables here: age of latest edit, vote per view percentile (I assume SE staff gets the internal data on eligible populations of up- and downvoters), possibly numbers of up- and downvotes (aren't some questions controversial?), number of answers... I hope you flesh out your formula, run it on Data SE, and provide the links for us as beta testers. –  Deer Hunter Jul 10 '13 at 16:02

You must log in to answer this question.

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