Unprotection experiments

edit

Page protections are a last resort to stop disruption, and in rare cases pages are protected indefinitely rather than for a set time period. While this stops disruption, that very fact makes it hard to determine when it is safe to remove protection. Since the only way to know is to remove protection and watch, pages can wind up protected for much longer than intended or even needed. Many of these protections were likely forgotten about: placed on pop-culture pages that have faded out of the public eye or to combat widespread problems that have long-since been resolved.

In May 2021 Wugapodes (and others) began reviewing indefinite move protections in place for over a decade, and some of these pages were also indefintiely protected against anonymous editing for a similar period. As part of the move protection review, Wugapodes began removing long-term edit protection as well in order to see whether protection was necessary (see WP:TRYUNPROT). The effects of page protection are immediate and obvious, so our documentation and policies have adequately considered when and why to apply page protection. What happens when a page is unprotected after a long period, however, has not been adequately documented or understood. This section (and eventual essay) documents pages unprotected as part of the experiment, qualitative analysis of editing activity, and eventual outcomes of the unprotection experiment.

  • Dragon (protected due to vandalism on 17 January 2009; unprotected 4 May 2021) an unconstructive edit was immediately reverted by ClueBot. An editor added an unreliable reference which was reverted by a presumed page watcher.
  • Dance (protected due to vandalism on 15 January 2009; unprotected 3 May 2021) 6 edits since unprotection (last six edits date to February). One new user made productive contributions and was welcomed. An IP added personal opinion which was removed about three hours later. No bad faith edits after 8 days.
  • Flower (protected due to vandalism on 13 January 2009; unprotected 3 May 2021) within 24 hours of unprotection, an IP was warned for making a test edit where they intentionally misspelled a section heading; reverted 15 hours later. No other edits afterwards.
  • Cattle (protected due to vandalism on 30 May 2008; unprotected 2 May 2021) 25 edits in 7 days after unprotection (prior 25 edits date to January). Two IPs made good faith but net negative changes which were reverted. Another was warned for test edits which stood for about one minute. Four IPs vandalized: one was immediately reverted by ClueBot, two were reverted within a minute by recent changes patrollers, and a third was reverted by a helpful IP after about 30 minutes. One IP and three registered users have made constructive, non-revert edits after nearly a week. The page was temporarily semi-protected on May 11 following a number of disruptive edits that were not promptly reverted.
  • David Beckham (protected due to vandalism on 7 January 2009; Beckham retires from football May 2013; unprotected 2 May 2021). Two editors (one IP) were reverted and warned for making test edits, both were reverted after about 12 hours. An IP vandalized the article and was reverted by a recent changes patroller in about a minute. An IP made a good faith edit which led to a removal of poorly sourced information. Another editor removed a poorly sourced statement due to (presumably) increased meatball:PageChurn. An anonymous editor was introducing unverified material on May 10 and 11 which was not promptly reverted. To prevent BLP violations, the page was pending changes protected for two weeks.
  • Frog (protected due to vandalism on 5 January 2009; unprotected 2 May 2021; protected indefinitely 3 May 2021) page was unprotected roughly 40 hours and experienced 11 edits. Of those, 7 edits were spamming sock puppets and the remaining 4 were reverts. Because of the volume and persistence of the problem, the protection was reinstated.

The above data were updated 23:50, 11 May 2021 (UTC)

The below analysis was written 03:14, 9 May 2021 (UTC)

At this point, data are limited but point towards generally positive outcomes. Of the six test cases, only one had immediate justification for reprotection, and anonymous editors have generally interacted positively with the remaining five articles. The most questionable is Cattle which has experienced disruption roughly once a day, but while anonymous editors have made both positive and negative contributions, existing soft-security processes such as recent changes patrolling and page watching have removed the negative contributions rather quickly. While many positive contributions were ultimately rejected, others were reworked or led to further improvements by more experienced editors all while introducing readers to the editing interface.

This increase in "reader engagement" with the editing interface likely comes even from the "negative" contributions such as test edits. While an apparent majority of anonymous edits were tests that ultimately required volunteer labor to fix, it also provided an opportunity to meatball:WelcomeNewcomers who seem interested if not yet constructive. Similarly, the test edits were generally "conservative" manifesting as changes to single words or digits. While nonetheless disruptive, they are easily reverted and help identify readers who might be productive if they were given direction. For this reason, it may be better to use the {{subst:Welcome}} series of template (e.g., {{subst:welcome-anon-test}}) which provide links to introductory material instead of {{subst:uw-test1}} which links them to the sandbox and provides little direction for their curiosity. Perhaps uw-test1 could be improved by adding a "call to action" similar to {{welcome}}.

Preliminary results are consistent with the theory of meatball:WikiBreathing and suggest that the way readers interact with out content has changed since 2009. The process of WikiBreathing was first documented on WikiWiki where the ability to contribute was periodically restricted and then loosened to manage wiki growth. As a model of the meatball:WikiLifeCycle, it theorizes that wikis go through a period of growth until a point where the community cannot cope with the influx of new members. At this point the wiki begins to "breathe out" by restricting contribution and allowing a net outflow of contributors. This is consistent with the history of Wikipedia prior to 2008 and the period of protections being reviewed. 2009 marks the peak of Wikipedia's entrance into the mainstream, with community size beginning to fall from that point (see Wikipedia#Launch and early growth). As the encyclopedia and wikis as a concept entered the mainstream, high-visibility pages became targets for test edits and general vandalism as the public began to experiment with an encyclopedia that anyone can edit. The existing community was not large enough to incorporate this influx of contributors and so the soft-security processes broke. To protect the encyclopedia, Wikipedia began to breathe out and make pages more difficult to edit. This trend continued into 2011 when the community added inactivity criteria for administrators (yielding our single highest year-over-year drop in administrators on record) and asking the Wikimedia Foundation to restrict article creation to auto-confirmed accounts (which the Foundation refused until 2017). This sequence mirrors other wikis (in direction, if not magnitude) such as MeatBall and WikiWiki which revoked editing access following major vandalism attacks between 2010 and 2015. The pages present in the data above are moderate- to high-visibility pages on major topics of public interest protected due to vandalism during a period of WikiExhaling, but have circumstances changed enough that continued WikiExhaling is actually harmful to the project?

While the protection accomplished its goal, the data present the possibility that the current effects are a net negative. The obvious harm is in preventing good faith anonymous edits. While many anonymous edits were not retained in full, readers were able to point out areas of the article that needed more attention through action rather than requiring them to write an essay on that talk page. Less apparent though is its affect on meatball:PageChurn and whether the increased presence on Special:RecentChanges increases edits from advanced editors. Preliminary results at David Beckham support the idea that the increased PageChurn positively affects articles. That article saw substantive copyedits by two editors new to the page (no edits to it within two years), and both occurred soon after an IP edit. While the page had been edited and expanded, the IP edits seemed to have instigated copy-editing distinct from what seems to have largely been writing and expansion over the last few edits. This suggests that the PageChurn brings controversial or sub-optimal sections of articles to the attention of experienced editors who are then able to apply the appropriate policies and improve the article. Long-term protection seems to have harmed these pages by making them harder to find and reducing how often regular editors are drawn to them, ultimately reducing the level of policy-informed copy-editing they receive. More data from more pages and for longer periods of unprotection is needed.