User:Cenarium/Passive reviewing

Passive reviewing is a proposal to extend the reviewing mechanism introduced as part of the pending changes implementation to articles that are not subject to pending changes protection but have been marked as needing review by the edit filter or a bot. Readers would still see the latest revision to the article, even if it hasn't been reviewed yet, so this would serve as a tool for passive monitoring of articles, as opposed to a form of protection. Reviewers could check for edits that are unacceptable in the sense of the reviewing guideline, the same way that pending edits are checked. This proposal complements and is complemented by deferred changes, which allows the edit filter and bots to actively defer an edit to an article by forcing it "pending", which would require more stringent criteria than merely marking for review.

Rationale
The aim of passive reviewing is to improve the monitoring of articles, particularly biographies of living people, using the already existing reviewing system of pending changes, but leaving aside its restrictive aspect. Over time, recent changes and watchlists have become largely inefficient for this purpose due to the high volume of edits unequally spread over so many articles. A large number of articles are unwatched, or watched intermittently by few editors who are likely to miss numerous edits. Semi-automated tools can only provide a partial response to this issue, since they can only check one edit at a time, and provide little coordination abilities. By providing reference points known to have been checked by trusted users, this new system would allow site-wide coordination of the reviewing efforts and prevent work duplication. In addition, the edit filter and bots would tag suspect revisions so that they are reviewed in priority. To sum up, advantages over the current system would include that :
 * A new edit has to be checked only once for vandalism or other obviously unacceptable content.
 * Multiple edits in a row can be checked and no edit is forgotten.
 * If an article is vandalized and subsequently edited by another user, vandalism can remain hidden or only partially removed, this will be prevented.
 * Edits automatically assessed as suspicious and likely to require urgent review are prioritized, and will therefore being taken care of faster.

Reviewing
Reviewers can mark a revision as reviewed, which informs that the revision should not contain any vandalism, WP:BLP violation, or other obviously unacceptable edits. The precise criteria for reviewing are defined in the reviewing guideline. When an article has not previously been reviewed, reviewers must verify that the current revision contains no remnants of unacceptable edits. Reviewers have access to Special:ProblemChanges, which can be filtered by category and tag. Unreviewed revisions tagged by the edit filter or bots as needing review would be displayed there, enabling reviewers to quickly take care of edits with a high risk of being problematic. As with pending changes protected pages, autoconfirmed users would be autoreviewed when editing a stable version (except on pending changes protected pages at level 2 or when deferred).

Technical aspects
The FlaggedRevs extension needs tweaking to support putting a page in a flaggable state but still show the latest revision, while maintaining compatibility with the current implementation of pending changes and ensure compatibility with the parallel implementation of deferred changes. Not all tags should be displayed at Special:ProblemChanges, only those indicating a problem. It may possibly use in addition a weighing system for prioritization. In an active system like pending changes, it makes sense that the oldest edits are displayed first. But in a passive system, prioritizing the review of suspicious edits is more efficient, due to the distribution of the survival time of vandalism and page views.