User:Ken g6/Reference Levels

For a long time I've thought Wikipedia's reference requirements could be better. While newspaper references are great, for instance, there are now fewer and fewer newspapers. At the same time, journalism by media such as blogging is growing rapidly. Like blogs, Wikipedia is a creature of the internet, and I believe it works better with online sources than with print sources. This should be taken into account. Wikipedia is also full of unreferenced statements that, while often valid, are officially forbidden.

In order to take advantage of the new citizen journalism, while keeping it in perspective, I propose differentiating references, by quality, into the following levels:

• Level 1: Theoretically verifiable statements of fact that are otherwise encyclopedic, but have no references. I believe these have a place in Wikipedia, if only for documenting the rapidly changing memes of our culture. If there is a dispute over the veracity of such a statement, even if the disputer also lacks references, the statement should be removed.
• Level 2: A Level-1-type statement with a reference to any web site that does not appear to have been written by the Wikipedian posting the statement. In cases where there is doubt about the authorship of the website, the usual procedures for dispute resolution apply, but a principal of "innocent until proven guilty" (or in this case convicted by consensus) should apply.
• Level 3: A Level-1-type statement with a reference to a blog or site by a person who has a Wikipedia article about them. This person is usually not the Wikipedian who posted the statement, though it may be. If it is clear that the author of the blog or site has no knowledge of the topic discussed, this should have no more weight than a Level 2 statement, but again a principal of "innocent until convicted by consensus" should apply.
• Level 4: A Level-1-type statement with a reference to a blog or site affiliated with an unbiased, factual publication or news organization which has a Wikipedia article about it. (The Onion doesn't count, for instance.) Note that this does not require the referenced text to be purely factual; op-eds fit here. If it is clear that the author of the blog or site has no knowledge of the topic discussed, this should have no more weight than a Level 2 statement, but again a principal of "innocent until convicted by consensus" should apply. Op-eds should probably also go here.
• Level 5: The normal fact-checked news article level.

Each reference level trumps every level below it. Level 5 trumps all other references.

Paying for Wikipedia references

You may notice that I haven't mentioned old-fashioned paper book/journal articles much here. I realize that in the past (including before Wikipedia) such references have been the gold standard, but there is a major problem with them for Wikipedia. If you reference a website (assuming it's not behind a paywall), I can go look at that site and verify that it says what you say it says. Any contributing Wikipedian can, because all contributing Wikipedians have Internet access. But a journal article (or a web page behind a paywall) is not so accessible. Books are similar, and although Google Book Search may make things easier in some cases, I believe they should be treated in the same way.

So for references to material behind such a paywall, I suggest the following procedure:

1. Any such reference given by anyone starts with a Level 2 reference ranking.
2. A "trusted user", other than the original poster, should then look at a copy of the reference and verify that it says what the original poster claims it says. I would like two users to look at the reference in all cases because (a) even a person with the best of intentions can misinterpret an article, and (b) there's a good chance that no one else is going to verify this reference.
3. If there is a sighting dispute and it appears that one party blatantly misrepresented a reference, a moderator should look at a copy of the reference. Blatant misrepresentation of a reference should be a vacationable offense.
4. In the event that someone finds a contradictory reference to an un-sighted pay-to-view reference, the un-sighted reference should remain, perhaps commented-out, in the text unless another reference with ranking greater than its maximum ranking is found that disputes it. (E.g. if person a cites an op-ed in a paper newspaper, that's given an initial Level 2 ranking. If someone finds a Level 3 reference online that disputes it, that should be included, but the op-ed should remain because if verified it could become a Level 4 reference. If someone finds a normal, fact-checked news article in a paper newspaper that contradicts the op-ed, that also starts out as a Level 2 reference. In that case, both references should remain, either as alternatives or commented out, until the reference with the higher maximum rank is sighted.

The sighting procedure could also be used for freely accessible online material, but it's somewhat cumbersome to deal with unsighted references as above. If it were to be used, I would suggest that for online material:

1. Any new reference to online material should begin with its maximum reference ranking.
2. A "trusted user" should be able to sight their own references.
3. Misrepresentation of a web reference should be a less severe offense than misrepresentation of a pay-to-view reference, more akin to regular vandalism.

Automatic categorization of references

For the most part, the reference categories I've suggested above can be automatically assigned by the wiki software.

• Level 1: This kind of statement has no reference, so there's no problem. We already have this.
• Level 2: Any reference with <ref> tags is at least a Level 2 reference. I suggest that a Level 2 reference link should be red.
• Level 3: If a reference has an author wikilink, but no publication wikilink, it's a Level 3 reference. References with any wikilink, but no template, could possibly be elevated to Level 3. I suggest that a Level 3 reference link should be orange.
• Level 4/5: Any reference with a publication wikilink should be at least a Level 4. I don't think there's an automatic way to distinguish between Levels 4 and 5. Level 4 refers to a statement by a trusted but unauthoritative person, while Level 5 refers to fact-checked news. An "oped=yes" option in reference templates might be useful, e.g. making the reference link green. Otherwise, I think we should just remember that news articles are more authoritative than op-eds and play it by ear.

Note that references without a template can't exceed Level 3 in this scenario, and bare URLs - even to e.g. nytimes.com - are Level 2. I think that would be good incentive to improve references overall.

References without a URL should start out unsighted. I suggest a yellow background for unsighted URLs, with their text color indicating their maximum reference level.