Wikipedia:Flagged revisions/reliable revisions

The reliable revisions proposal is a proposed procedure for the usage of the flagged revisions MediaWiki extension (see mw:Extension:FlaggedRevs for the technical description of the extension).

Please discuss this proposal on the discussion page.

Summary
Flagged revisions allows certain editors to flag certain revisions of articles in different ways, such as rating a certain revision or setting a certain revision as a stable revision.

Reliable revisions is a proposal to implement flagged revisions, flexibly, in such a way that readers of the English Wikipedia have the option to acquire:


 * stable, possibly and hopefully reliable, but potentially out-of-date content, or
 * unstable, possibly and hopefully not unreliable, but potentially up-to-date content

in accordance with their own preferences.

Reliable revisions constitutes a partial compromise between sighted revisions and quality revisions, two other proposals for the implementation of flagged revisions written long before this proposal, as well as a new approach. Sighted revisions treats flagged revisions liberally and primarily for fighting damage; quality revisions treats flagged revisions conservatively and primarily for recognising quality; and reliable revisions recognises that both of these purposes can be fulfilled in a single implementation.

This proposal recognises that flagged revisions is a feature which has the potential to provide readers of the encyclopedia with more reliable content and a method of seeing how reliable this content is.

This proposal recognises that the burden of decision between stable and unstable content should lie with readers of the encyclopedia.

This proposed implementation of flagged revisions would, if implemented, give readers the ability to make their own decisions about the content they want to read.

Fact sheet (what you must know)
This section lists the minimum you must know about this proposal. For explanations of some of the terminology used, you should read the full details of the proposal.

Groups
The patrollers group


 * can rate aspects (accuracy, depth, readability) of article revisions up to accurate–moderate–good,
 * cannot downgrade aspects of article revision ratings if they are higher than these values, and
 * is useful for vandal- and spam-fighters and other editors dedicated to maintaining quality.

The reviewers group


 * can rate aspects of article revisions up to featured–featured–featured,
 * can downgrade aspects of article revision ratings to any level,
 * can set a stable revision for an article, and
 * is useful for editors dedicated to improving quality.

Members of the sysops group (administrators) can


 * grant and revoke access to and from patrollers and reviewers.

Interface
The interface is


 * clear, and
 * simple.

Details of proposal
By default, there are three aspects of revisions that may be rated – accuracy, depth, and readability – and five levels of rating, ranging from "Unapproved" to "Featured", as demonstrated in the following table:

For the sake of simplicity, an article revision's rating will be simply referred to in three point accuracy–depth–readability form for the purposes of this proposal.

Patrollers and minimal rating of revisions
Proposal: This proposal suggests that a certain user access group of editors, patrollers, be permitted to rate revisions up to accurate–moderate–good, for use in fighting damage. This group would not be permitted to downgrade aspects of ratings that were higher than this level.


 * Explanation and reasoning: It seems useful for editors (such as recent changes patrollers) frequently engaged in fighting vandalism, spam, and other damaging modifications to have the ability to give article revisions a basic rating – at least sighted–basic–acceptable. It seems worthwhile to give such editors the ability to carry out minor "verification" of content, hence the maximum rating level of accurate–moderate–good.  Such editors should probably not be permitted to lower an aspect of a revision's rating when it was above one of these levels.  Access to this user group would be granted fairly liberally to practically all editors who demonstrated a desire for the feature and had minimum experience in fighting vandalism, spam, or other damage.

Reviewers and full rating of revisions
Proposal: This proposal suggests that another user access group of editors, reviewers, be permitted to rate revisions up to featured–featured–featured, for use in giving readers of articles some idea of the quality of certain revisions. This group would be permitted to downgrade aspects of ratings to any level.


 * Explanation and reasoning: It seems useful for editors frequently involved in the production, development, and quality of content to have the ability to elevate advanced revision ratings to any level (such as the ultimate featured–featured–featured), and have the ability to downgrade ratings to any level. Access to this user group would be granted on a more conservative basis than would access to patrollers, since editors in the group would have a high responsibility for ensuring quality and would thus need to be fairly experienced in editing.  A requirement of having contributed significantly to some form of recently-featured content could be sensible.

Stable revisions and reviewers
Proposal: By default, the extension also allows a stable revision of an article to be maintained. This proposal suggests that editors in the user access group reviewers, but not those in patrollers or any other groups, be permitted to set a revision of an article as stable, i.e., that it contained information thought to be accurate, have a moderate depth, and have good readability. The current revision would be displayed by default to all readers of an article, logged-in or logged-out, and a suitably obvious but discreet link between the current and stable revisions would be placed at the top of these revisions.


 * Explanation and reasoning: A stable revision of an article would only be permitted to be set if it was thought to be accurate, have a moderate depth, and have good readability. All editors in the reviewers user access group would be permitted to set this revision.  The stable revision would be easily accessible to readers via an unintrusive, small but noticeable link just under the article title.

Role of administrators
Proposal: Administrators would be permitted to grant or revoke access to both the patrollers and reviewers user access groups.


 * Explanation and reasoning: Administrators would be permitted to grant or revoke access to patrollers and reviewers because they are (supposed to be) trusted, experienced members of the community, who, in most cases, are qualified to make a decision in regards to a user's fitness for access to quality maintenance functions.

Interface
A header would appear on top of all articles which had at least one flagged revision, and would look similar to:

Stable revision | Current revision | Find out more

Accuracy:   Depth:    Readability:


 * Clicking the "Stable revision" link would navigate the user to the stable revision of the article.
 * Clicking the "Current revision" link would navigate the user to the most recent revision of the article (the one they would see by default).
 * Clicking "Find out more" would navigate the user to Flagged revisions or a non-technical description of the extension.
 * The purpose of the three sliders, "Accuracy", "Depth", and "Readability", is obvious: to indicate the quality of the article revision in each of these aspects.

If we implemented a cookie-based default revision preferences system, then the header could become:

Stable revision (make default) | Current revision (make default) | Find out more

Accuracy:   Depth:    Readability:


 * The additional "make default" links would make either the stable or the current revision of articles display on default article view.

Cookie-based default revision preferences?
Users should be permitted to define whether or not they want to see stable or current content by default. The only feasible way of implementing such a system, in my opinion, is to maintain a cookie-based preferences system for users to define their preferred revision to display upon default page view.

Major differences from the German Wikipedia implementation

 * The current revision of articles would be displayed by default to all readers, as opposed to readers seeing the last "sighted" revision. I believe that this addresses the main issue that many German Wikipedia editors have with the system.
 * Access to the patrollers group would be granted liberally on a case-by-case basis by administrators. Access to the reviewers group would be granted more conservatively, but still on a case-by-case administrator basis.
 * The link to the stable revision would be prominent, clear, and simple, meaning that readers could easily find out what flagged revisions was, how it worked, and whether or not they wanted to use the stable revision.