Community Wishlist Survey 2021/Admins and patrollers/Automatically detect edit-warring

Automatically detect edit-warring

  • Problem: Bringing edit-warring to the admins' attention currently requires editors that have the time to go through a fairly bureaucratic process.
  • Who would benefit: Editors who wish edit-warring to be more strictly enforced against, and administrators/other editors who wish to enforce it more strictly or just arbitrate conflicts.
  • Proposed solution: Add a function that automatically compares every new revision against all those of the last 24 hours, and then flags to the administrators or other editors if it detects the same text being added or removed more than three times (as per the three revert rule).
  • More comments : There could still be a separate process for making formal complaints, and this could simply be a way to get attention that there is a conflict brewing, preventing things from getting out of hand earlier. Further comment: it seems that the current process for edit-warring complaints is extremely inefficient, with complaints being archived without any attention having been given to them. Therefore, perhaps this function should automatically give short blocks for first time breakers (24hrs) and longer ones for repeat offenders (72hrs and more). I think having complaints not addressed at all is extremely disappointing, and perhaps a reason for editors to stop contributing.
  • Phabricator tickets:
  • Proposer: Keepcalmandchill (talk) 04:19, 17 November 2020 (UTC)[reply]


  • This seems like something that could be easily automated and would make it a lot simpler to detect edit-warring and hopefully keep it from festering. Jessamyn (talk) 04:36, 17 November 2020 (UTC)[reply]
  • This should be a lot easier to build now thanks to the reverted edit tag. You might even be able to write a Quarry query for it. MusikAnimal talk 05:36, 17 November 2020 (UTC)[reply]
  • I've seen some research on this topic that is interesting (unfortunately the results of which are predominantly not used by wiki editors). --Izno (talk) 19:51, 17 November 2020 (UTC)[reply]
  • This shouldn't be too hard, but I do think you'd need to take some steps to avoid major false positives. I don't know how general the auto-tagging system is, but with that en-centric bias noted, you'd want to have edits with "potential vandalism/blanking/etc" tags discounted, so edits reverting them wouldn't count. The pages are already being handled and so pulling further attention to it is unneeded Nosebagbear (talk) 15:03, 18 November 2020 (UTC)[reply]
    • There are various other issues with this, as well. For an enwiki example, reverting against consensus doesn't really need to be reported to ANEW. Bots wouldn't really need to report for 3RRNO exceptions either. As such, I think it's best this be done on a per-project level, with a bot, if needed. ProcrastinatingReader (talk) 06:46, 1 December 2020 (UTC)[reply]
  • I'm going to move this to the Admins & patrollers category, since the proposal is more about moderation than it is about improvements to editing. Hope this is okay, MusikAnimal (WMF) (talk) 20:22, 24 November 2020 (UTC)[reply]
  • This should probably be implemented at the community/bot level, at the discretion of each local community. Now that we have the reverted edit tag, each local community can action automatically-detected edit warring as it sees fit without further WMF technical effort. Best, KevinL (aka L235 · t) 06:39, 1 December 2020 (UTC)[reply]
  • Support with reservation. I agree with the problem, but not sure with solution. It's not difficulty to create automatic detection but once it's implemented, it's also very easy to bypass and circumvent. It motives warriors to change their behavior and make it hard to detect. Xinbenlv (talk) 22:24, 8 December 2020 (UTC)[reply]