7
$\begingroup$

So what seems like an age ago now, I asked if we should have a community-run Twitter account. It seemed like the response to that was generally positive.

A few of us have taken on the project of developing a bot to run some aspects of that account. One thing it's planned to do is manage the tweeting of question links. This can be done using suggestions from users in chat, and if that doesn't provide enough material then the site can be scraped to find potential questions.

However, I'd assume that we don't want 'bad' questions tweeted. Therefore, we need to define what 'good' is, in terms that a bot can use. This means in things like

  • score
  • views
  • length
  • close and delete votes
  • existence of post notices
  • number of answers

and any mathematical operations thereon. Some of those are reasonably obvious, some less so. So what should the requirements be?

$\endgroup$
  • 9
    $\begingroup$ I would suggest views are unimportant, many bad questions get many views because they have intriguing titles but the questions are sometimes very bad. $\endgroup$ – Bellerophon Feb 6 '16 at 19:29
  • $\begingroup$ I think I'd agree with that, @Sam - I mentioned it in the question more as an example than because I think it should be a criterion. $\endgroup$ – ArtOfCode Feb 6 '16 at 20:33
  • $\begingroup$ So you essentially wants to improve on the network-wide hotness calculations. Maybe one of the default of the HNQ is that it puts some emphasis on recent questions. In our case, we could require the question to be at least a few days old. To see where it really goes. And if did not reach popularity before getting closed. $\endgroup$ – clem steredenn Feb 6 '16 at 20:47
  • $\begingroup$ Could questions that provoke debate and many different ideas be calculated somehow, these questions could be good for twitter as they will make people inetrested to learn more and maybe visit the site. $\endgroup$ – Bellerophon Feb 6 '16 at 20:55
  • 4
    $\begingroup$ Unfortunately, "questions that provoke debate and many different ideas" are usually the worst questions for the Stack Exchange format. I'm not saying it's always a bad thing, but I would absolutely caution against promoting those as examples to draw in people, especially people who are unfamiliar with the Stack Exchange question and answers format. $\endgroup$ – a CVn Feb 6 '16 at 21:03
  • 1
    $\begingroup$ On Mi Yodeya we sometimes retweet old questions that are timely again. Worldbuilding isn't tied to a calendar like Judaism is, but sometimes questions are connected to current events (science news, related book/movie releases, etc). We'd never be able to automate that, though; this is stuff that human curators should be thinking about. $\endgroup$ – Monica Cellio Feb 7 '16 at 2:18
  • $\begingroup$ @MonicaCellio "We'd never be able to automate that", programmers, prefer to talk about challenges. ;-) $\endgroup$ – clem steredenn Feb 8 '16 at 6:17
  • $\begingroup$ @bilbo_pingouin fair enough. But AI is a hard problem still. :-) $\endgroup$ – Monica Cellio Feb 8 '16 at 13:45

You must log in to answer this question.

Browse other questions tagged .