Wikipedia:Timed article change stabilisation mechanism

This article describes a technical and policy proposal for a very lightweight peer review scheme to make all the articles in Wikipedia vandal-resistant and improve quality and more importantly perception of Wikipedia.

Summary

The proposal is a change to Wikipedia's software so that less well-established editors' changes would be delayed for a day or so before being shown to the users as current articles.

The purpose of doing this is to permit a lightweight form of peer review on edits. It does this by giving a chance for more experienced editors to spot these changes in their watchlists before they go live, permitting corrections and vandalism removal to be performed, improving the overall reputation, wikipedia standards compliance, and factual accuracy of Wikipedia. Vandalism is expected to be greatly discouraged by delay of appearance of changes and by the review.

• The idea scales with Wikipedia, and lessens the workload of both admins and editors.
• The implementation issues don't seem overly difficult, and no changes to any existing databases would be required.
• The idea, if implemented, should reduce the number of vandalisms, since vandals' changes won't appear immediately if at all; this would deter this mode of interation with the wiki.
• It makes it harder/more expensive to use the wikipedia as a link farm, since it costs time/money to get an established account

This, if implemented, should raise the reputation of the wikipedia in the international community, and minimise the perception of the wikipedia being an encyclopedia that anyone can vandalise or a way of raising the google index of a site.

Problem & Philosophy

One problem we seem to have is vandalism by a small minority that are mostly low-edit count and anonymous editors. This small minority of vandalising editors tend to 'test edit'/vandalise Wikipedia. When they do that, these changes appear immediately to the public.

On the other hand, very many low-edit editors also do great work, so we don't want to require all low-edit editors to jump through big hoops just to put their changes into the wiki.

So, a mechanism is needed to filter out vandalism that is lightweight to use and as much as possible is in keeping with the current esprit d'Wikipedia.

We really do want anyone to be able to edit anything, regardless of how long they've been using Wikipedia, provided they are contributing useful information. Ideas like voting or formal peer review add a lot of complexity and overhead to the users of Wikipedia, and may very well be unworkable for every little change. We also want to promote 'liveness' of the wikipedia- articles should not hang around in purgatory for indefinite periods; otherwise people may decide not to contribute if their changes aren't ever seen.

We'd like to be able to dissuade vandals from bothering in the first place. If they don't think their vandalism will appear, they may well not try.

Proposed Solution

Unlike the semiprotection policy, the following proposed policy would automatically apply to ALL articles and project pages by default:

First some definitions:

Immature editors

An 'immature editor' would be defined as an editor who fulfills one of the following criteria:

1. Making edits anonymously;
2. Making edits from an account that is less than one month old; or
3. Making edits from an account that has made less than 50 edits.

The criteria for selecting the number of edits to become mature (50) should be that vandals should not be able to easily make the number of edits required to become mature in order to then commit instantly-appearing vandalism.

Edits by an immature editor should not appear "live" to the public for 24 hours from the time of the edit. (The time period of 24 hours was selected because many editors check their Wikipedia watchlists only once per day. If a shorter time period was selected, malicious edits by an immature editor might thus not be caught prior to going "live.") An edit within 24 hours by another immature editor would 'reset the counter', and the current article would then only change after 24 hours had passed since the other immature editor's edit.

These numbers are rough approximations for the purposes of this proposal, and could conceivably altered over time, or even on a per-article or per-editor basis. Smaller stabilisation periods like 10 minutes might work, but might possibly let too much vandalism through.

Current version

The current version would be whichever is newer of:

1. the newest article version that has been edited by a mature editor.
2. the newest article version by an immature editor that has not been subsequently edited for at least 24 hours

Note: that this rule could theoretically mean that a new edit would never become visible, if edits by immature editors repeated continuously. In practice, this would not happen frequently. (The reason we need to take this risk is to handle repeated vandalism or corrections; if edits became visible after one day regardless of any corrections, then vandalisms would eventually be visible anyway, even if the vandal or another low edit editor had fixed it.) We don't really want all the intermediate edits anyway, the fact that it is rapidly being edited is a sign that something is potentially wrong with the article.

Mature editors

An edit to the article by a mature editor would make the edited article immediately current.

This would be as it currently is, but it would also mean that any vandalism could be tidied up immediately if they were somehow missed in the 24-hour period. It would also represent a mechanism by which an immature editor's edits could immediately be made "live," if the changes were so wished (for example, on an article describing a current event), by a mature editor making a dummy edit to the article.

Mature editors committing vandalism

Any logged-in editor caught making inappropriate edits would have their account permanently locked out.

They would have to start from scratch as an immature editor. This would mean that most visible edits would not be bogus, even if a mature editor turned rogue. Well-established editors are much less likely to vandalise anyway, as they have invested something in Wikipedia.

Advantages and implications

Introducing these changes would deter vandals in the first place, since their changes will probably be removed before appearing. Not only would visible vandalism be decreased, but actual vandalism attempts should decline as well.

Implementation

These rules may seem computationally expensive to implement, or difficult to introduce. They really aren't.

An "embargo list" of Wikipedia articles could be kept. This list would contain the comparatively small number of articles that would have been edited in the last 24 hours by immature editors, and for each such article would specify which article revision was currently being shown to the public. Any article in Wikipedia that had not been edited in the last 24 hours by immature editors would work normally.

The existing database entries would not need to be changed. A new set of database records would go alongside the preexisting ones, tracking embargoed articles (these records would track article ID, the date the article was last edited, and the article's current visible version) and tracking editors' maturity (these records would track User ID and edit count).

The relatively expensive calculation only needs to be done during an edit, and once for each editor. Luckily, the embargo list can be made self-cleaning, since embargo list entries would no longer be required 24 hours after their last edit and could then be automatically removed. Information on editors' maturity could be stored in a list separate from the current editor list, since it is difficult to add information to a preexisting list.

Things that would *not* change

The only thing that would change is the 'current' article; everything else would not change. Most especially, the following wouldn't change:

• The version presented for editing to editors would always reflect the most recent edits, regardless of whether that edit was made by an immature or a mature editor. (We do not want to branch Wikipedia).
• The article's history would display all edits, including those made by immature editors.
• If editors or the public select a specific article version from the article's history, it would be displayed, even if said version reflects an edit by an immature editor.
• both logged in editors and the general public see the same current article (some non editing users have accounts to record settings such as skins, if the wikipedia starts treating people with accounts differently to non logged in users, then the wikipedia becomes less useful and friendly to such readers.)

Other changes that aren't absolutely essential

A few other changes to the wikipedia software would be a good idea to go with this proposal:

• when an editor edits an article where the current version isn't the same as the article the editor is editing a warning should be shown (in this scheme editors see the same current version as the general public, so it's not obvious, and we want to minimise confusion.) Alternatively, it was suggested on wikien-l that the edit button be replaced with 'view latest' when the current article isn't the latest.
• when an immature editor is editing to make a change that isn't going to go immediately current, it would be a good idea to warn the user about that and hint that vandalisms probably won't appear at all
• in change lists, it is highly desirable to highlight articles that are not current
• at least for mature editors: when an editor sees a version that is not the latest version, a link to the latest version should be included at the top of the article.

Bugzilla Reference

See Bug ID 4397.

WolfKeeper 22:46, 20 December 2005 (UTC)

See also

• Article validation proposals meta mention of this
• Meta:2006 proposed approval for anonymous edits