The Signpost

Editorial

We need a culture of verification


The celebration of Wikipedia's 15th birthday threatens to be overshadowed by debates concerning governance of the various Wikimedia projects and how much of a voice the community will have in the future direction of the Wikimedia movement. These debates also threaten to overshadow another debate we should be having about the future of the community, regarding what lies at the heart of the movement and its community: the encyclopedia itself.

So far our focus has been primarily on growth. This was natural and appropriate for a movement that built the world's most widely used reference work out of nothing. In October, the English Wikipedia reached five million articles, and we loudly celebrated that milestone and every milestone beforehand. Expansion has been the watchword: expanding to five million articles, expanding stubs to complete articles, expanding articles to Featured Article status, expanding the encyclopedia to cover content gaps. We celebrated the people who wrote those articles, showered them with barnstars, and marked their articles with indicators of their quality, as we should have.

Now the community needs to have a conversation about maintaining what it has built. One hesitates to reach for the cliche comparing the growth of an organization to that of a living person, but sometimes the comparison is apt. Wikipedia is reaching adulthood. The growth spurts are subsiding. Instead of focusing on constant expansion, now is the time to turn our attention to maintenance and upkeep.

When we turn our attention to an article or topic, the results are generally positive. Some of Wikipedia's worst mistakes and embarrassments, from the Seigenthaler controversy to Jar'Edo Wens have come from the lack of attention from editors. Hoaxes and defamation can lurk in the encyclopedia because editors did not see a particular article or reacted inadequately, by ignoring it or slapping a tag on it and leaving it for others to deal with.

Assuming good faith is one of our core values, and we justifiably have a lot of faith in the results of crowdsourcing and the abilities of other contributors. That may cause us not to be sufficiently critical when evaluating encyclopedia edits. Countless times editors have challenged obvious vandalism or implausible edits and stopped there, while less obvious vandalism from the same editor goes unchallenged. For example, in the case of a fabricated Thoreau quotation, an obvious hoax was immediately challenged. So the hoaxer quickly provided a fake citation to a real book, without a page number, and this citation went unchallenged and unverified, persisting for six years, even after the hoaxer tried to undo their own hoax.

The logo of The Wikipedia Library.

As the saying goes, we need to trust but verify. We can trust the work of our fellow contributors while verifying their facts and citations and not letting their edits go sufficiently challenged. We need to encourage a culture of verification so when editors see something like the Thoreau hoax or Jar'Edo Wens, they act sufficiently by deleting it or verifying it instead of challenging it once by adding a tag and then forgetting about it. In addition to celebrating the content creators, we need to celebrate the content verifiers, like Mr. Granger, who looked at that Thoreau quote and its citation six years after it was added and would have none of it, and ShelfSkewed, who uncovered a hoax article about an imaginary war by systematically examining citations with invalid ISBNs.

Creating projects and procedures to systematically verify articles and citations is one approach we can take to create a culture of verification. The Stanford Encyclopedia of Philosophy has its authors update their topics every four years. Perhaps we can systematically revisit articles after a set period of time instead of relying on tagging and forgetting, waiting for hoaxes and other frauds to be accidentally discovered.

One such new program for verification is scheduled to run from January 15 to 23, 2016, in conjunction with Wikipedia's anniversary celebration. The Wikipedia Library has created 1Lib1Ref to encourage librarians to bring their expertise to bear on these issues by having as many of them as possible add a single reference to Wikipedia. Imagine if we could get librarians to do this every year, or Wikipedians to do this every week. It would go a long way to working on our backlog of tagged articles and to encouraging editors to think about these issues as a fundamental part of their work here.


+ Add a comment

Discuss this story

These comments are automatically transcluded from this article's talk page. To follow comments, add the page to your watchlist. If your comment has not appeared here, you can try purging the cache.
  • To do this systematically (as we should) requires a little bit of planning. We need a way of recording that references are checked, and by who. Secondly it would be valuable to use quotations, to speed up the verification process. Thirdly we need an automated system to evaluate the trust that should be placed in the verifiers.
  • It's important to understand the problems that these three facilities would solve:
    1. It is no good verifying references at random, and not recording the verification - that's inefficient, it duplicates effort and misses items.
    2. Given a quotation it is easy to see if the quotation supports the statement it references. And it is trivial (if an electronic source is used) to verify that the quotation is accurate.
    3. Fake verification is as bad, or worse than, fake references. However fake verifications are likely to follow specific patterns, notably a small clique of new users or single purpose accounts supporting relatively few editors, probably part of the same clique.
All the best: Rich Farmbrough, 17:04, 16 January 2016 (UTC).[reply]

All of Rich Farmbrough's very sensible suggestions are based on an unstated assumption, namely that references can be seen to apply to a specific claim (such as a sentence, or perhaps a whole paragraph). However, the encyclopedia still contains many thousands of older articles with a list of unattached citations, listed at the end of the article with no connection to any paragraph. Therefore, a vital precursor to verification is to convert unattached citations to specific (attached) references, using inline ref tags to create traceable "little blue numbers". Chiswick Chap (talk) 19:46, 16 January 2016 (UTC)[reply]

Demur. By far the bigger problem is that some editors see "ambiguity" in a source - and parse it to reach their own preferred result. Short "quotations" do not necessarily properly represent the view of any author at all, and the result is that the "automated system" using such quotations will simply rubber-stamp horrid misinterpretations of what the work actually states as fact. And the acceptance of "facts in headlines" as proof that the claim is in the body of any source.

Rather, I suggest that:

Edits by "flagged authors" including, but not limited to, IPs, authors with under 300 article edits not marked as "minor", and any editors flagged as being found to have committed plagiarism or major multiple copyright violations, be placed in "pending" status" until a reviewer then examines them for accuracy and copyright status. And all edits giving "foreign language" sources, video sources without accurate transcripts, and "not findable" sources should also be placed in that limbo.

Which, I suggest, is more readily do-able than the initial suggestion above. Collect (talk) 14:06, 17 January 2016 (UTC)[reply]

We find a problem- so the first reaction is to use a few more terabytes of memory to provide a global resource that unnamed virtual editors are going to do complex sql matches. No. No, lets see what practical- in my everyday editing I verify dozens of references using online sources as I clean up sections or zap cn's. I use {{sfn}} in the main and place the simple tag |p=34|or |pp=34-35| as needs be I could add |four tildes or |v=exact| or |v=dubious| |v=POV |v= fail, at the same time. We need a simple traffic light code, to inform the reader the quality of the reference. So with a little help from our coding bunnies we have started to address the problem. But is a tool we need before we can start. Clem Rutter (talk) 23:32, 17 January 2016 (UTC)[reply]


While we also need improved tools, I think it very important to point out that a verification foremostly requires detailed work by editors having some domain knowledge on the subject they are checking or who at least are willing to do the leg work to pick it up on the way. So those editors cannot check just the validity of the references, they actually need to look it up, read its content and compare it to the Wikipedia content using it as a source. This is very tedious and time consuming process, which also requires a different attitude from the "tag and forget"-approach. With regard to the latter, we probably also need a more sensible (or at least prioritized) approach to tagging. Right now I still see a lot of purely "formalistic" tagging, meaning editors often without any domain knowledge of the material they are surveying simply tag any paragraph or statement that doesn't come with footnote. They do not seem to spent time on the article to figure out, whether the paragraph just contains obvious domain knowledge, where sourcing isn't necessarily needed or whether a neighbouring footnote might cover that material as well. This can result in unnecessarily long maintenance lists, which contain a lot of articles that upon closer inspection may have no real sourcing or correctness issues. So we need a sometimes more sensible approach and probably also a way to prioritize sourcing/verification issues better.

There is one tool, which has not been all that popular in the English wikipedia, which however maybe important for verification/consolidation attempts and that are the flagged revisions. So far they are mostly used as protection against vandalism. However the original idea also considered an "verified flag" indicating that a particular version of an article was fully fact checked/verified/sourced at that point. So it might be a good idea to revisit the idea and options of flagged revisions. Maybe even a more fine grained tool is necessary where you can break it down to individual section or footnotes. However tools which are very complex with have a lot of options can in practice also become more of an obstacle than a help for the majority of editors.

Another important aspect for a verification drive is to get a large number of editors to help with it. For that the (online) access to the sources is critical. So we probably need to extend (and possibly simplify) the use of Wikipedia library and the resource exchange request. Ideally all regular Wikipedia editors should have access to all (online) resources.

Finally there isn't necessarily a contradiction or conflict of goals between expansion and consolidation/maintenance but they go hand in hand as we do verify articles while we expand them.--Kmhkmh (talk) 02:50, 20 January 2016 (UTC)[reply]



       

The Signpost · written by many · served by Sinepost V0.9 · 🄯 CC-BY-SA 4.0