Wikipedia talk:Deferred changes/Archive 1

Stats on the reviewer group
Just curious, on the talkpage of Flagged revisions there where many requests for statistics, I see Pending_changes/Metrics has been made by the WMF statistics team for its use on BLP, thats progress, are there any statistics on the reviewers ? Lets say, collect all reverts in 6 months and show the amount of names that do 80% of all reverts in the reviewer group, to compare it, collect all sightings in 6 months and show the amount of names that do 80% of all sightings in the reviewer group, to compare it.
 * In relation to Deferred changes, did the statistical team make a baseline for future statistical testing ? The current supply of proof is not much more than "Hey, i found a vandal that made edits for 1 hour!", did you ask the Cluebot NG people if the bot was down in that period ? Mion (talk) 01:04, 9 December 2014 (UTC)
 * Regarding your specific question about Cluebot NG, I know that in some of the cases the vandal was repeatedly vandalizing the same article, and Cluebot follows an 1-revert-rule, meaning it cannot repeatedly undo edits by the same user on the same article. Tony Tan98 ·  talk  04:50, 9 December 2014 (UTC)

This one shouldn't be too difficult, one just needs to extract the data from the advanced review log, but I don't know the API well enough, and then the analysis is simple. Cenarium (talk) 09:45, 13 December 2014 (UTC)

Stats on vandalism
All the stats that you mention are very old, probably not much relevant to deferred changes. A recent and helpful study is at WP:Wikipedia Signpost/2013-02-04/Special report, which I've discovered very recently (I'll add it), where as you see a substantial number of people can see even obvious vandalism. So if those damaging edits had been deferred by the edit filter or a bot, they would not have been seen by the public. In the draft proposal, those are just random examples, not statistics. I can't make the kind of stats that you ask myself, and I don't know the statistical team. However, it may be a good idea to ask at the signpost if they are aware of any similar studies. Cenarium (talk) 17:07, 10 December 2014 (UTC)
 * Hoi Cenarium, thanks for the link, the signpost stats are aready 3-4 years old, because BOTS improve at such a fast rate at the moment, see Artificial intelligence, the numbers mentioned in the article are not correct in this year. We need better stats on the reviewer group anyway, so pls contact Analytics/Research_and_Data--direct link and ask them to deliver proper stats for the section Impact analysis.Mion (talk) 20:51, 10 December 2014 (UTC)
 * ClueBot NG's neural network is certainly impressive and in constant improvement, but it still lets a lot of vandalism through in order to achieve a very low false positive rate. That's why I was thinking of having ClueBot NG defer the edits that are very likely vandalism but not over the treshold for rollback. It would probably take a while to get results from the team, so I'll see if there aren't more existing stats. I'd like to inform the authors of the article, this subject may interest you, and maybe you're aware of new or related stats ? Thanks, Cenarium (talk) 10:15, 13 December 2014 (UTC)

Suggested change of wording
Hi I hope that you don't mind a few of my copyedits on your draft. (They are minor.) However, do you think it would be better if we make the following change to the second paragraph of the rationale section:

Original:

First, the excessive amount of time it takes for administrators to block the reported users, a concern that is only amplified by the widely held view of a shortage in administrators. Among users reported at WP:AIV, both the total number of edits performed between last warning and AIV report, and performed between AIV report and block, are significant.

Suggested change:

First, it takes an excessive amount of time for administrators to block the reported users, a concern that is only amplified by the widely held view of a shortage in active administrators. Among users reported at WP:AIV, both the total number of edits performed between last warning and AIV report, and performed between AIV report and block, are often significant.

If you do not mind me joining in and collaborating with you on this proposal, please let me know.

Thank you, Tony Tan98 ·  talk  21:24, 10 December 2014 (UTC)
 * Yes, feel free to edit the draft. This change is good. Thanks, Cenarium (talk) 00:08, 11 December 2014 (UTC)

Let rollbackers put vandals on deferred editing
I like the idea of reducing the amount of vandalism that gets seen, but I'm not convinced you could screen out much with the proposal as is. What would make a difference would be if anyone with Rollback rights was able to put editors on deferred editing after even a single act of vandalism. The current system involves vandals doing a series of vandalisms to go through four warnings and a block, if Rollbackers could put vandals on deferred editing then a significant proportion of vandalism would be less visible. OK we might need to recruit a few more reviewers, but so few vandals mend their ways that I wouldn't expect many of the deferred edits would be good ones.  Ϣere Spiel  Chequers  21:21, 19 December 2014 (UTC)
 * And anyone who wants nothing to do with Pending Changes or Flagged Revisions at all will respond by surrendering the Rollback tool. We're not your fucking rookies, CRASH. Deal with Rampart yourself. — Jeremy  v^_^v  Bori! 05:42, 7 January 2015 (UTC)
 * Maybe a bot could automatically put users on moderation if they have been rolled back "too much" (exact meaning of "too much" to be defined in further discussion) across multiple pages? Since moderated edits would be listed on AIV, this could also act as a check on abuse of rollback, ensuring that admins will notice if rollback is abused when the target of abusive use of rollback gets moderated.  Vandals will, of course, remain on moderation until they either mend their ways or get blocked entirely.  Pathore (talk) 21:03, 31 January 2015 (UTC)


 * If we give this to all rollbackers, are they actually going to use it, and if they do, will they actually know what it means and how to use it properly ? I don't think so, rollbacker is a single purpose usergroup with little vetting. A new usergroup is cheap, and ensures that the community has a say on who uses this brand new permission. As for when to apply moderation, it may be a good idea to revisit the issue once we have tried the suggested "one step before block" standard for a while, but using a broader approach immediately may be too risky, as we need to keep an eye on the backlog of unreviewed changes, among other things. Cenarium (talk) 00:26, 1 February 2015 (UTC)

If deferred edits are to be a lesser restriction than Pending Changes, why require a Pending Changes reviewer to accept them?
I suggest allowing any other autoconfirmed user to accept a deferred edit. Since, in this proposal, moderation cannot be applied to autoconfirmed users, a greater level of trust for autoconfirmed users is already implied in the current proposal. I say "other" autoconfirmed user because users should not be able to accept their own edits that are flagged by an abuse filter. Pathore (talk) 21:11, 31 January 2015 (UTC)
 * No. — Jeremy  v^_^v  Bori! 21:59, 31 January 2015 (UTC)
 * We could always extend the pool of reviewers if it turns out that there is a shortage, i.e. revisions take too long to be accepted. But it's granted quite liberally already, and even if we give it to all autoconfirmed users, it doesn't imply they'll actually use it (in fact, the vast majority of reviewers don't review). In addition, having two different types of reviewers would be quite difficult to implement technically (it would require another flag, another special page, to duplicate mediawiki messages, etc), it would really complicate everything a lot for little gain. Cenarium (talk) 00:15, 1 February 2015 (UTC)