Community Wishlist Survey 2017/Admins and stewards/Automatic detection of admin candidates

Automatic detection of admin candidates

  • Problem: There is a persistent shortage of admins. We need a better process for recruiting new admins.
  • Who would benefit: Everyone.
  • Proposed solution: More of less the reverse of an anti-vandal bot. Run a bot which examines editors' history: contributions, talk page activity, warnings, block logs, etc. etc., and use appropriate algorithms (to be determined) to detect accounts that might be potential good admins. (This might, for example, be a neural network trained on the contributions of current admins before they were made admins.) Make this list public, and send those accounts messages suggesting they put themselves forward for adminship.

    It's also possible that this automatic pre-review might also offer communities the possibility of shortening the current lengthy review/interview process for admin candidates selected this way.

  • More comments:
  • Phabricator tickets:


  • Some current (not-the-best alternatives) include w:WP:List of administrator hopefuls and tools like Enterprisey's admin score/Enterprisey's candidate search (why are these not the same tool--as in, why does the admin score not show up in the candidate search, or at least a cross reference link? @Enterprisey: :D ) as well as the xtools version of admin score. --Izno (talk) 18:34, 17 November 2017 (UTC)Reply[reply]
  • Spamming potential candidates is not a viable (or correct) solution IMO. Eligible candidates typically decline nominations because RfA is synonymous with a trip to the village stocks. -FASTILY 01:44, 18 November 2017 (UTC)Reply[reply]
  • This proposal is very enwiki-centric and doesn't necessarily apply elsewhere. --Rschen7754 18:31, 19 November 2017 (UTC)Reply[reply]
  • A bot is anything but a good solution for this kind of issue. I don't think anything automatic is a good approach in finding potential candidates but, anyway, a SQL query is far more efficient than realtime edit scanning. --Vituzzu (talk) 14:37, 23 November 2017 (UTC)Reply[reply]
  • Ninety percent of any serious RfA voter's criteria are based on trust and an evaluation of the cadidates competence and knowledge, things that no AI or bot can do. This would just produce a lot of failed Requests for Adminship with accompanying disappointment and loss of face for the candidates. Kudpung (talk) 20:01, 6 December 2017 (UTC)Reply[reply]


  •   Oppose This proposal has limited impact as a project-specific problem. MER-C (talk) 01:59, 28 November 2017 (UTC)Reply[reply]
  •   Oppose Per MER-C, too much unnecessary. --Liuxinyu970226 (talk) 12:41, 28 November 2017 (UTC)Reply[reply]
  •   Oppose Further to the above, the English Wikipedia already has more admins than the next 15 Wikipedias combined. Jc86035 (talk) 15:02, 28 November 2017 (UTC)Reply[reply]
  •   Oppose --Gripweed (talk) 21:26, 28 November 2017 (UTC)Reply[reply]
  •   Support Thomas Obermair 4 (talk) 21:33, 28 November 2017 (UTC)Reply[reply]
  •   Oppose IKhitron (talk) 22:55, 28 November 2017 (UTC)Reply[reply]
  •   Support Get the engine working - and then use with extra care. A mandatory opt-in should be enforced. I know, it is somewhat not the idea, as it would only find future admins willing to be found, but the alternative would be listing user's activity without their consent (legal or not, that is not proper, in my book) Nabla (talk) 19:25, 29 November 2017 (UTC)Reply[reply]
  •   Support Andrew (talk) 20:14, 29 November 2017 (UTC)Reply[reply]
  •   Oppose good to have such keen tools to evaluate contribs, but very bad for suggesting candidates (an undeclared bot would score the best records...) --g (talk) 23:43, 29 November 2017 (UTC)Reply[reply]
  •   Oppose I don't see how anything good can come of this. Nihlus 04:48, 30 November 2017 (UTC)Reply[reply]
  •   Oppose In the choice of potentially eligible admins, a key role is the way they approach other users from the human relationship perspective. This cannot be detected by a bot, however "neural" it may be. Furhtermore, there's a consent issue: not all "good contributors" may have the will to become admin. Supporting this tool would mean also, for fairness reasons, developing a kind of "no, thanks" tool to exclude from the algorithm all the users who explicitly decline the idea for becoming admin. But this paves the way to further technical and ethical issues: a way to "step back" from the "no thanks" list (people may change their mind with the time), exposing user names without their permissions, adequate advisment to all registered users (new and existing) that their name could be chosen by an automatic system, an adequate handling of "accept/don't accept" that option, and so on. The balance costs+drawbacks+implications vs. benefits is heavily negative, IMHO. --L736Etell me 07:27, 30 November 2017 (UTC)Reply[reply]
  •   Oppose As the saying goes, "to err is human. To really screw things up takes a computer". The only good I could possibly imagine coming of this would be that for once everybody would agree on what the problem is with our admin-selection process.

    If you like computer dating works out for you, you'll love computer admin-recruitment. Daniel Case (talk) 00:45, 1 December 2017 (UTC)Reply[reply]

  •   Oppose this cannot be done in automatic or even semi-automatic mode. --Superchilum(talk to me!) 16:14, 1 December 2017 (UTC)Reply[reply]
  •   Oppose Wostr (talk) 10:07, 2 December 2017 (UTC)Reply[reply]
  •   Oppose --Termininja (talk) 15:32, 2 December 2017 (UTC)Reply[reply]
  •   Support Otapka (talk) 18:41, 2 December 2017 (UTC)Reply[reply]
  •   Oppose Admins should be recruited/nominated by the community, not a bot. Matěj Suchánek (talk) 21:04, 2 December 2017 (UTC)Reply[reply]
  •   Oppose per the opposes above. Boing! said Zebedee (talk) 21:22, 2 December 2017 (UTC)Reply[reply]
  •   Support WikiMasterGhibif (talk) 23:42, 2 December 2017 (UTC)Reply[reply]
  •   Oppose When it comes to selecting admins ultimately I would trust the judgement of human editors over any algorithm. -- OlEnglish (Talk) 00:07, 3 December 2017 (UTC)Reply[reply]
  •   Oppose It is easy to detect many kind of bad edits. But it's technically impossible in a sufficient way to detect good people. This will hold for the next few years. So don't let us waste time in such topics. -- seth (talk) 10:35, 3 December 2017 (UTC)Reply[reply]
  •   Oppose--Per above.Winged Blades of Godric (talk) 16:23, 3 December 2017 (UTC)Reply[reply]
  •   Oppose - IMHO humans should pick and select the suitable candidates - Not bots and all that, (If they run for RFA and it's instantly SNOWCLOSED then that really wont look good on you or the project). –Davey2010Talk 15:09, 4 December 2017 (UTC)Reply[reply]
  •   Oppose Most likely it only identifies dull candidates, not the good ones. The Banner (talk) 16:22, 4 December 2017 (UTC)Reply[reply]
  •   Oppose Humans need to do this job.-Bryanrutherford0 (talk) 17:32, 5 December 2017 (UTC)Reply[reply]
  •   Oppose There will indeed come a time when the number of candidates no longer compensates the attrition, but not for anytime soon. Kudpung (talk) 20:08, 6 December 2017 (UTC)Reply[reply]
  •   Weak support Support as long as we use machine learning technology and it only gets a list of a few BLATANTLY satisfactory Admin candidates. Mr. Guye (talk) 23:17, 7 December 2017 (UTC)Reply[reply]
  •   Oppose, unfortunately. Anything simple would not work as simple definition of what is a good admin does not exist. I can imagine a very complex solution where interactions (messages, edits/reverts etc.) of each user are studied and compared to patterns of previously successful admins, but this would be probably too complex to develop for a rather limited impact. In the end most editors have a few names of good potential admin candidates, and this would be way more impactful than a complex tool — NickK (talk) 20:13, 8 December 2017 (UTC)Reply[reply]
  •   Support Zppix (talk) 20:55, 10 December 2017 (UTC)Reply[reply]