Wikipedia:Pushing to validation

The following seems to represent the consensus position reached in formal and informal discussions at Wikimania 2006. For background information see the original abstract, and an audio of the discussion or a written summary (incomplete). At the German Wikipedia, people seem to have reached a similar position with respect to the first two stages shown below, though they discussed the second stage (validation) much less than the first (stable versions).

Process
There would be potentially three stages in the full validation process, with the article available for view after any stage.
 * 1) A stable version of an article would be tagged as usable and unvandalized.
 * 2) A validated version.  The stable version would be fact-checked (using Wikicite?) by a team of subject experts, probably from the trusted members of the appropriate WikiProject. This might involve temporarily protecting the page from edits during the validation, or alternatively a short-lived fork would be needed.  After the work is completed this version of the article would be made visible and tagged as an uneditable validated version.  Any user could look at links from this validated version to see exactly what source was used to check a particular fact.
 * 3) Independent review. An outside expert, with a reputation in the field, would check that the validated article was balanced, up-to-date and complete, then sign off to say that they approve of the article.  See the Blessings project.

See the Wikicite project page for a visualization of this process.

Even if we could just reach stage 2 on a core of 1000-2000 articles, this would be a great step towards giving Wikipedia credibility among librarians and academics (and comedians?!) as an information source. We might want to start with the more important Featured articles, and the better quality Core topics.


 * Comment : hopefully none of the above would be "uneditable" -- save in the sense that no revision of any article can be edited in place in MediaWiki. We need to expose metadata saying "this is a validated version", and to have validators adding validation-flag content to the main article, in a way that lets readers of the most current article know which parts have been validated.  We don't need to add bottlenecks to the public editorial process.  The blessings project is a description of a way to add validation/review flags and comments -- and a way to allow for browsing of specially flagged revisions -- without interfering with the current default reading/editorial flow.  +sj + 15:49, 11 August 2006 (UTC)

Infrastructure

 * 1) The changes needed to implement stable versions should be available soon, and should begin on the English Wikipedia once trials on the German site are complete.
 * 2) Fact-checking could be done quietly on a sub-page of the article, by a team assembled at the appropriate WikiProject.
 * 3) The Wikicite software may be ready soon, and this will make the fact-checking process efficient and relatively easy.
 * 4) Wikipedians involved in fact-checking could (if they wish) make available their credentials as subject experts.  These could be checked by official representatives of Wikipedia, and then included in all validations performed by that individual.
 * 5) A system of links to the literature sources used in fact checking, from each fact in the validated version.  This might be able to work with Wikicat or similar, and it would reduce the "Footnote-creep" problem seen in some articles.  Any user could see when and where the fact was validated.
 * 6) We will need a project to collate names of subject experts who are willing to work with Wikipedia.

Criticisms and issues to be resolved

 * 1) Stable versions ("unvandalized versions") need to be available on en.  Will the default be the stable version or the editable one?  That issue will probably be resolved elsewhere.
 * 2) How will we persuade people to work on the difficult/tedious task of fact-checking?  A validated FA may become regarded as a "super-FA" with very high status – this may be an incentive.
 * 3) Who will be allowed to be on the validation team?
 * 4) Temporary protection or minor fork? Should we protect the page from edits during validation, with any updates/comments added to the talk page?  This is surely not a big problem in "mature" articles.  Would it be better to allow a temporary fork of the article as it undergoes validation – this would be a "dead-end"as the validated article would be uneditable.  If a system of fact-checking can ensure any errors (hopefully few) fixed in this version are also fixed in the main wiki version, then forking could be (in effect) eliminated, but that would require careful design of infrastructure.
 * 5) World Book allows unpublished testimony from subject experts, as long as this is recorded – could we set up a system to do the same?  Many topics such as pop music don't have a viable academic literature – but an interview with the band themselves should be able to count, surely?
 * 6) How can we persuade big-names to work with us?  In some areas, there may be no big-name PhDs, how would we handle these cases?  How would we ensure that our big-names don't try to insert POV?