I was surprised to learn from the latest podcast episode (#60) how serious SE is about predicting good questions as well as about reviewing.

I feel I put quite some effort into my questions, and they still don't always generate many or good answers. How could I get better at this? There are always the help files but those seem to be meant for people already knowing that they should ask a better question and actively seeking help (OK, I am now), instead of getting proactive real-time suggestions when the question is going bad. (The site does recommend other posts, why not recommend question guidelines?)

Where are the lessons from the Vowpal Wabbit experiments, at least the ones that can be shared? (I learnt that e.g. "thanks" predicts bad questions, but there is no point in telling people not to thank, that won't improve the otherwise already bad questions.)

A more operational idea: Why can't we spend some of our reputation to ask for review? I spent some on a bounty on a question, but that did not help.

By the way, many of my questions arise from bugfixing, so they can be very specific, and I am happy to provide detail to help pinpoint the issue. On the other hand, this makes the question less useful for others and posterity. Who will weed out the with-hindsight-irrelevant details?

I was also surprised to learn that some generic tags "worsen" question quality (e.g. simply tagging java). Are they better avoided, then? I thought it is better to use five tags no matter what. Again, the site could gently nudge us away from this, then.


We're looking at quite a few sources of signal when it comes to questions, any one or even two of them are sort of meaningless without the context of the rest. We've spent the last few weeks looking for every possible characteristic of a question, including what users do just prior to asking them that could determine the probable quality of the post.

Think of it not as a series of checks, but a series of dip switches that get programmed as your question passes through various checks. These won't necessarily block your question from being posted, but it will determine the journey your question takes in order to become visible. Some posts will just go right to the front page, some will need to enter a triage where a human being needs to indicate where it should go next, some might go to a special queue where folks interested in improving diamonds in the rough can do so.

Nothing is yet carved in stone, and this is, for lack of a better word, alchemy. David Fullerton, Jay Hanlon, Shog and I have been putting the bulk of our time and effort into it. Bear in mind, this is not a new problem and the metaphorical 'mine' of stuff that we might have overlooked when examining quality was not what one would call abundant. However, we feel that we've got something substantially better than what we have now, and we're working as fast as we can to an implementation (or, more likely, a series of implementations).

There is also the idea of a sort of mentoring program, where folks can make themselves available to other (generally new) programmers. This would not be to teach new programmers how to program, it would be to teach them how to ask the questions that they need to ask in order to become better programmers. It's still a wildly speculative, somewhat controversial idea - but it's on the table. I'm not going to be taking questions about it, because there's no information for me to answer with - just an idea Joel had, that's it - no more.

Finally, Jarrod has implemented a system that allows Shog, Jon or I to put in some expressions that match certain patterns. For instance, questions that begin with "Please write" are universally and always awful - so just-in-time help is going to get quite a bit smarter in the coming weeks. This isn't yet out in the wild, but should be soon.

I wish I had a more coherent answer to your question, but I won't until all of these things have come together and gelled a bit - and we're right on the cusp of that happening now.

  • 1
    This is great, Tim. I just want to add that it can be hard to balance long-term professional/personal development goals and short-term goals. People are not actively discouraged from using the site to solve specific problems. Until this is the case, it would be a mistake to make them only ask "What is wrong with my life? Where did it all go wrong?" questions (exaggeration) instead of "This is not working, help me." (not ideal, but can happen) – László Jul 20 '14 at 14:28

I was also surprised to learn that some generic tags "worsen" question quality

Correlation does not imply causation. During the podcast it was said quite clearly that the things they found correlating with low-quality posts are not the reasons why those posts are low-quality. For example, the tag correlates with low quality, but there is no point in telling users "don't tag your question ". (When this was said during the podcast it was a joke.)

The lessons learned from the data are lessons for SE Team, not for the users. Although some users (like myself) may find the information helpful when using Data Explorer to look for questions that should be edited or closed or deleted.

The guidance is in the Help Center: how to ask (and its much more specific version on Stack Overflow) and how to get out of question ban.

  • Don't you think that generic tags will bring a lot more attention to a question from people who don't understand or appreciate it. Out of those people, who's attention had been brought to the question, who cannot contribute, at least some fraction will unjustly vote or moderate. Those who can contribute are likely to. People have both the power to answer questions as well as to help make sure questions don't get answered. There is a very, very, large quantity of bad moderators and voters in the community; especially who follow certain generic tags. – MVTC Jul 20 '14 at 3:44
  • One more thing: Until visibility is not a more transparent process, it still seems rational that people will use popular tags when desperate for attention (and help). Also, SE seems to be a site built more on the wisdom of the crowds than on algorithms. Trying to hide a question from unworthy moderators, voters and editors seems misguided to me. – László Jul 20 '14 at 14:24
  • "wisdom of the crowds" is a double edged sward. In most instances, crowds are not collectively wise. And it's also kind of the point of SE to hide questions from unworthy moderators, voters and editors. That's why there are tags. When people follow tags, it's because they like to filter out questions they aren't interested in or don't know much about. My main thought though was that it just seams likely that overgeneralizing your audience can result in the question being shot down in some instances. You add a lot more potential close voters for one. – MVTC Jul 20 '14 at 21:25
  • 1
    If your questions are as rambling as your comments, I'm not surprised they get closed. (To both of you). – user259867 Jul 20 '14 at 21:36

You must log in to answer this question.

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