Talk:Auto-Edit

From MusicBrainz Wiki
Status: This page describes a failed proposal. It is not official, and should only be used, if at all, as the basis for a new proposal.



Proposal number: RFC-Unassigned
Champion: None
Status: Failed, due to Unknown
This proposal was not tracked in Trac.


Discussion of an enhanced AutoMod system

  • The following discussion is only of historical interest, since it predates even the elections conducted on the musicbrainz-users MailingList; it was on the AutoMod page, where it really didn't belong, but I didn't want to delete it either. So I've moved it here, updating the interjection style to be a bit more Wiki. @alex


Moderators now spend a considerable amount of time voting on obvious correct moderations. Moderators with, for example, > 250 approved votes should be given automatic moderation approval (AutoMod).

  • I note that before we have more experience with this feature, it should probably remain a manually managed class. In addition, no matter what such an automatic threshold, a determined troublemaker could always stay just above it with trivial mods, and then introduce some really bad errors. Perhaps the best suggestion would be to require a peer-vote component for automod approval. In addition, the rate of reversed to non-reversed votes should stay below some strict limit. Self-deleted (auto)moderations would obiviously be excluded from this determination. --DonWulff
    • Good idea; if moderators with AutoMod pass the threshold of 20% bad AutoMods (determined with the reversal vote) they should have been more careful and their AutoMod status is revoked automatically. --Johan
    In addition, moderations which are determined obiviously correct on technical grounds (Such as style-corrections) should probably follow same "automod protocol". It is mostly an implementation question of what kinds of moderations should be included in this class; capitalizing has been discussed the most, but obvious various artist album break-ups and even obvious spelling errors might be included (Not an exhaustive list). --DonWulff

This would create an 'elite moderator' class that have proven to be valuable moderators. Moderations by Elite moderators are approved unless voted otherwise. This feature requires an overview page with 'most recent AutoMods' to make it possible to review these moderations in bulk without voting. All users can simply click on the AutoMods to put a reversal up for vote, and see all other pending moderations of this Elite Moderator. If a reversal is approved, the number of approved moderations for the 'Elite moderator' should be decreased.

  • I seem to have misread this on first time, so I should stress that the AutoMod reversal should be treated as just another, standard vote, and not an automatic reversal, or the potential for damage from single bad guy rises through the roof. This "reversal" button should probably be provided for all moderations, not just the automatically approved ones, and should provide good feedback for determining whose moderations can be trusted. I am thinking reversal-mods should never be automods, altough it does suck that people will get wrong metadata in the meanwhile. --DonWulff
    • Yep, the AutoMod reversal would be a standard vote. --Johan

Elite moderators are trusted for life with a a limited form of write-access to MusicBrainz.org. There is a remote possibility that Elite moderators turn evil. As suggested by Donwulff, the AutoMod access is automatically revoked if the Elite moderator no longer fulfills the precentage X of good AutoMods. An alternative is to put the 'revoke Elite status for moderator Y' up for vote, but I prefer the automatic method.

The number of votes is currently overwelming. The Elite moderator feature would recognize good MusicBrainzContributors and apply their changes directly with an effective mechanism to revert their changes and revoke the Elite status. --Johan

  • What about putting up a fourth voting option, such as "Trivial Yes". When three trivial yesses are reached unanimously, then the vote is passed. This would cut down on the number of votes and save time for minor editing changes. Perhaps this option would only appear for contributions that come from high quality moderators, those who already have proven that a high percentage of their moderations are passed unanimously anyway. It would just represent a level of trust that that the moderator knows what he's doing. --Eric
    • How about a sliding scale of moderator acceptedness, where as a moderator gets more positive votes, the number of votes required for his suggestions to pass reduces? Also, people with better ratings could vote more than one vote at a time, so that it only takes one supermoderator vote to pass any change, but it takes 5 normal votes to pass a normal users change. A level 2 user would only need 4 votes to have a change accepted, and would cast 2 with each vote he approves. --RjMunro
    A simple weight/karma system such as rjmunro suggests would have the effect that a new bad users would have very little power and would quickly descend into negative karma. A powerful good user that turns bad would quickly lose their power and thus could have a limited effect on the data. However, I would suggest that someone's quality as a moderator does not necessarily reflect their quality as a voter, so the two should be separate. A lot of good moderations should increase your moderation power a lot, but your voting power only a little. A lot of good votes should increase your voting power a lot but your moderation power only a little. This type of karma system isn't really suited to the Wiki model, but I think it is suited to MB, due to the fact that most of MB is data that is largely either right or wrong. --Singletoned