3
$\begingroup$

If you take a look at our states on Area51, the two big areas that are still a problem for the beta are the rate of unanswered questions (which is close to 25%), and the average number of answers per question (which is about 1). These are related statistics.

Looking through our queue of unanswered questions, there are some that might be answerable, but that I cannot answer, and many that look unanswerable or very low quality. Many of the questions are very old.

Should we make a more concerted effort to close questions that are old and unanswerable? What criteria should we use to decide if a question meets that standard?

| |
$\endgroup$
  • $\begingroup$ Note that "unanswered questions" in SE context is "questions with no upvoted or accepted answers", so if there are good enough answers without upvote, consider upvote them (if you think they worth it) to remove them from "unanswered" state. $\endgroup$ – Andrew T. Aug 30 '18 at 9:47
  • $\begingroup$ @AndrewT. That makes sense, but we have some 350-400 questions with no answers at all. $\endgroup$ – John Doucette Aug 30 '18 at 11:21
  • $\begingroup$ @JohnDoucette The exact number is 393. And the fastest way to clear the queue is to search for topic in which you're interested right now. For example someone could enter in the searchbox “answers:0 chatbot” $\endgroup$ – Manuel Rodriguez Sep 2 '18 at 8:51
  • $\begingroup$ @ManuelRodriguez That also makes sense, and I have done so. Many of the remaining questions look difficult or impossible to answer however, hence this question. $\endgroup$ – John Doucette Sep 2 '18 at 11:23
2
$\begingroup$

If a question is unanswerable, it should be closed, be it old or new. This is more or less what closing is for.

But don't do it for the sake of Area 51 statistics. Those statistics outlived their usefulness, as did Area 51 itself. The post Graduation, site closure, and a clearer outlook on the health of SE sites explains that already in 2015, those stats did not really matter for site graduation or closure.

| |
$\endgroup$
  • 1
    $\begingroup$ That's interesting and makes sense, but what criteria do you think should be used to decide if a question is unanswerable? For example, if a question is new and I don't know how to answer it, I am sometimes pleasantly surprised to see someone else give a useful answer soon after. If it's 6 months old, if I don't know the answer, it starts to seem more reasonable to guess that the question is too poorly worded, or too exotic, to ever receive a good answer. $\endgroup$ – John Doucette Aug 29 '18 at 13:59
2
$\begingroup$

Speaking as a pro tem mod, we see a lot of single close votes, but tend to give the OP the benefit of the doubt, and err on the side of caution.

My feeling is the best method to increase closure of these "grey area" questions is to keep attracting knowledgeable contributors, and supporting those contributors by upvoting good questions and answers, so that more users have informal moderator privileges. (i.e. I'm personally more comfortable with closures being consensus-based because, as JD notes, it can be a difficult determination, even for qualified individuals.)

That said, I'd like to prune away as much of the noise and dead-weight as possible to improve our stats. I'm wondering if we might start a chat thread to address questions in limbo, so that if contributors make a strong case for closure, the moderators can more confidently take action.

| |
$\endgroup$
  • 3
    $\begingroup$ Some users don't Know they can actually accept an answer...I also didn't know when I was new. $\endgroup$ – DuttaA Aug 30 '18 at 4:24
  • 1
    $\begingroup$ @DuttaA agreed. (i try to gently remind new OPs to upvote if the answer is useful, and formally accept if it answers the question, when I notice comment discussion but no upvotes.) we've definitely got to support each other to build this community. $\endgroup$ – DukeZhou Aug 30 '18 at 17:08
1
$\begingroup$

I make a point of visiting the unanswered queue on all sites that I am active on. It's possible to earn an Explainer, Revival, Necromancer or other badge available to new questions.

We should run through the queue when we visit here.

The site Interpersonal.SE has an unanswered queue style similar to ours (single tab), while LifeHacks.SE has the advantage of a multi-level Unanswered Questions queue; with additional tabs for "my tags", "newest", "votes" and "no answers" permitting better differentiation. Both those sites have a similar total number of questions as we do, yet the number of unanswered questions is near zero.

The remaining question is do we want a multi-level queue like LifeHacks has? I'm new to AI.SE, so I'd prefer a senior member put in a feature request over at meta.SE.

Be certain to improve and better these similar requests that became ignored or status-declined:

Fortunately, automatic deletions are performed on old questions:

If the question is more than 365 days old, and ...

  • has a score of 0 or less, or a score of 1 or less in case the owner's account is deleted

  • has no answers

  • is not locked

  • has view count <= the age of the question in days times 1.5

  • has 1 or 0 comments

  • isn't on a meta site

... it will be automatically deleted.

| |
$\endgroup$
0
$\begingroup$

A good starting point is, that a statistics about the facts is available. That means, we can predict how long it will take for a new question until it get answered and how many answers it will generate. The bad news is a high rate of unanswered question is equal to a low service quality.

SE.AI is the frontend of AI in the internet. That means, there is no other forum out there which is better. If SE.AI has a low quality, AI in general has a problem. It is some kind of bottleneck. Bypassing the bottleneck is simple, it is only a question of resources. I'm with you that in the current status the resources are not here. The number of moderators is limited, the number of users who answering questions is limited. On the other hand, people who are familiar with Artificial Intelligence are available in the world, they are only not here in the forum. Why?

I think it is time to ask people who are familiar with neural networks, machine learning, NLP and robotics what they are doing with their time, if they are not contributing to this website here. Do they have holiday? Are they are interested anymore in AI? Are they are involved in projects with greater importance? We need them here at the frontline.

| |
$\endgroup$
  • $\begingroup$ I think you're partially right: there is an extreme shortage of AI expertise worldwide, and demand is very high. Anyone with the expertise to answer these kinds of questions can work as a highly paid consultant. The other half of the issue though, is that many of the questions are malformed, and unanswerable, especially the ones that don't have any upvotes. This isn't a question of not having Neural Networks folks to answer them: the questions have no answer, because they are founded on deeply incorrect assumptions. Others are really asking programming questions. $\endgroup$ – John Doucette Aug 31 '18 at 12:10
  • $\begingroup$ A clarifying example: The entire US education system produces about 100 new AI PhDs per year. A colleague of mine joined a startup that wants to hire 2-3,000 AI PhDs. Once they work there, they have exciting problems, great pay, and often longer work hours. They're unlikely to answer questions for free online. $\endgroup$ – John Doucette Aug 31 '18 at 12:13
  • $\begingroup$ @JohnDoucette I don't like Phd students, they are troublemakers. The better approach is to invite Wikipedia-Admins to SE.AI, they are easy to handle ... $\endgroup$ – Manuel Rodriguez Aug 31 '18 at 13:15
  • $\begingroup$ My point wasn't that we should invite PhD students. It was that the reason we don't have a lot of knowledgable contributors is because there is extreme demand for the skills, and most people don't want to work for free after spending all day working on the same thing for pay. Wikipedia articles on AI do not exactly inspire confidence in the knowledge of the editors or admins on those topics, even if they might be easier to work with. $\endgroup$ – John Doucette Aug 31 '18 at 14:17
  • $\begingroup$ @JohnDoucette Easier to handle means, that Wikipedia Admins are trained for a special food source. They are not interested in money, but they are hungry for lulz. Lulz is some kind of value somebody gets, if he can block a user or delete content. $\endgroup$ – Manuel Rodriguez Aug 31 '18 at 14:23

You must log in to answer this question.

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