Bad results of the last two site self-evaluations: what to do about it?

Question

So... um... if nobody is going to react or write anything, let me. What should we do about this?

When I came to this site, most of the questions were well written, interesting, and enticed me to research even the things I was not an expert in. I sometimes even felt like reading questions completely different from my area (which is Computer vision and Image processing). Not any more...

I've also actively participated in the reviewing process up until a few months ago... but as the quality of questions visibly declined, the questions started lacking details, the language got worse, no effort was visible in the questions and the requests for details from established community members went unanswered, I stopped. Back when I review would take me 2-3-5 minutes per question, it was okay. But now, if I need to spend 10 minutes reviewing a question related to my field, and 20+ (or impossible) for other questions... I don't do it.

I used to come here for some stimulative problems when taking a break from my own work. They're slowly but steadily becoming less and less stimulating.

Some facts to support my claims:

• this and the last self-evaluation not only produced bad results but barely got people participating
• the last time I remember really enjoying the site (and, coincidentally, the last good self-eval was 6 months ago)
• this site used to be at the brink on putting all stats in green on Area 51 (I remember), while now we're completely in orange except the number of users
• I actively participate on another beta side, academia.SE. I remember when Academia used to be a baby compared to DSP. But, a site a year younger has managed to preserve the quality of the questions, increase all the stats, and basically, "grow up" while we are stagnating, or even declining.
• I remember when we had above 90% answered questions. We tried to do something about it, but I don't think it really worked.

I care about this site. I think it's amazing to have a SE site in my topic, it helps me with my work, I use it for procrastination instead of facebookish things cause it re-directs my mind to other problems and keeps it sharp, I enjoy helping but... I feel like the site has to get back on track.

I am not a very experienced SE user (not very inexperienced either, but well..), so I don't really have any suggestions to improving the site, but it does seem to me that something has to be done. Does anybody else feel that way? Does anybody smarter than me have some suggestions?

Answer 1

Noise in the data

Let's start with the most obvious thing: we don't have a lot of data. The most discouraging thing about the September evaluation was that the number of evaluators dropped from 7 to 3. The December eval was completed by 4 people. With so few people looking at questions, it's not surprising that voting will be noisy.

Since April, there have been about 600 questions/quarter. Of those, only 10 are represented in each evaluation. (The sample does not include unanswered, downvoted, or closed questions.) So an especially good or bad sample won't tell the full story. It's possible we got unlucky with the questions selected.

Too few reviewers

We collect data on how many people are active on the site 5 times or more in a month as a rough estimate of user engagement. In the last year, that number has hovered around 45. These are the people who are most likely to participate in self-evaluations. We also measure people who post on the site at least once. These more casual visitors have more or less doubled (to about 400) in the last 12 months. Naturally, the number of posts has also roughly doubled in that period.

One thing that can happen on a site is that contributions from newer users can overwhelm site regulars. When that happens basic cleanup tasks can suffer. It's one of the reasons we shut down sites:

This is one of the biggest problems with “quiet” sites already—not enough activity for the posts to be reviewed, ranked, and updated. We saw this with the SE 1.0 sites, where eventually the task wore out even [the] tiny handful of users who remained.

I hasten to add that we aren't looking to close down Signal Processing. But I'm concerned that there aren't enough people voting on and reviewing content.

Vote Early, Vote Often

I don't mean to vote like a Chicagoan, but to make use of your daily quota. There are two huge benefits:

1. Answers will get properly sorted.
2. Good contributors will be motivated and empowered to play a more prominent role on the site.

Feedback of any sort (upvotes, comments, and even downvotes) encourage engagement with the site. When exceptional posts come along, feel free to encourage others to vote via a comment.

---

I took some time graphing votes per post by month:

Stack Overflow saw a similar drop from 2009 to 2011. We aren't yet sure what changed things there, but it should be a lot easier to reverse the trend on a smaller site, such as this one.

Answer 2

I'll give you my perspective as a semi-frequent user who isn't part of the community. The site seems kind of dead, which is surprising for a beta with 4,500 visits per day and 260 'avid' users. The two questions I've asked (here and here) both received 1 answer (both fine answers) from users with relatively low reputation. Where are all the 1k+ users? Why does the front page have 50% or more questions with 0 votes? You guys need to wake up.

At physics.SE we have ~15 truly avid users who combined answer 50% of the questions (my impression, not actual data). I'm not one of them, and I think some of them are crazy, but they make up a significant portion of the community activity. We also make an effort to answer even the questions which seem borderline infantile to a professional physicist.

I like this SE because it is a place for me to learn about a field which I use but have never had the time to study in detail. You guys just need to start doing things; anything. Vote up, vote down, mark duplicates, delete low quality posts, edit questions, answer questions, just do something.

Answer 3

A possibility: May be there aren't that many interesting new DSP questions. People keep asking the same newbe/broad questions and asking more deep/to-the-point questions takes much more effort than on other sites.

What might help - although I don't think it's practical in SE - is to have an online DSP IDE which will make the question askers and answerer's job easier. It can be a huge library of DSP routines and will help in quickly coding and showing something by example.

Answer 4

If I may bring up one thing for image processing as I am mostly interested in the area in this site, many people seem to misunderstand that single script or single method of image processing would work for their problem and thus is generally a target they look for. But proper image processing should not be treated that way. Like data analysis, proper procedure of image processing might be

1. you should know circumstance, assumptions, restrictions, and any other useful information on the images,
2. simplify the problems into smaller problems or tasks
3. appropriately preprocess the images for each task
4. build algorithm to resolve those smaller problems with various methods and models
5. validate your algorithm with more similar images
6. check the result to see if there is anything to improve algorithm

Many of questioners on image processing, however, post a single image or its description and ask how to achieve their goal without mentioning facts, assumptions & restrictions on image acquisition. They simply ask for a script or a method to solve their problems so it's very hard to answer because there can't no right answer and at best an opinion in those cases. As I don't know how to handle those questions, I am currently avoiding them because I cannot learn anything from answering them.

Answer 5

Biggest turnoff on the site right now is the recent flood of homework questions. They are usually put in here without any effort on both, trying to solve it and trying a good question style. Sometimes it's just a bad phone photo of the question from the textbook.

Due to their academic nature, the questions are more often than not very uninteresting. You can rarely learn something new from their answer and answering them is also not very challenging (and would just encourage more of them anyways).