Last modified: 2011-03-13 17:46:07 UTC
A solution is to be a bit like Slashdot and do things by
vote. My schema (which I'm not particularly attached to and
which can serve as a jumping-off point) involves ranking
articles depending on how well "moderated" it is, and how
good the people ranking the article have been in the past.
- Article WikiBlah Bug is entered by Person A.
- It is made available immendiately with a big
red "UNVERIFIED" label across the top.
- Immediately, it is also put in the list for moderation
- Anyone who logs in and has indicated that they will
moderate, will be informed that the article is available for
- A formula should be used to select whom an article should
be shown to. For instance, the next "experts" to log in
after the article is posted should be asked to score the
article and this stopped once the target of (say) six experts
is reached. Similarly, it should also be offered to six (or
however many is found to work best) non-experts for scoring.
- If a person's indicated area of expertise matches the
classification the article was placed under, their vote
carries a slight weighting. These are "experts" and their
expertise is rated by how many articles they are in
concordance with their peers in.
- a person also has the option of selecting a "don't know"
button - their "expertise" in the particular subject is then
- If this person has a history of consensus with their peers
in this subject, (i.e. they seem to be getting it right more
than they're getting it wrong,) their vote carries an
additional weighting and they are considered an expert, or a
negative weighting in the case of repeated non-consensus.
The more positive this rating the more expert a reviewer is.
- Both the "specialist rating" and "general rating" scores of
the article are indicated and the "UNVERIFIED" label removed
once the required number of scores have been received and
averaged. Until those review scores are reached, readers
should be able to see both the label and the scores.
- Specialist scores (someone who knows that particular
subject well as indicated by their positive weighting in that
subject) are the major component of the article score.
- Should a reviewer feel the article has been miscategorised,
they may suggest a different category. The article can then
be split into two articles, one in each category, and the
fittest of the two, after a certain threshold is reached,
will survive while the other is removed.
This way, you will collect statistics about your pool of
experts, contributors, and articles.
I think this would improve wikipedia tremendously, as you
have human peer review. After a while you will have a huge
list of subjects that a reviewer is "expert" in, and if they
submit an article on a subject of their expertise, it could
get a provisional rating immediately.
Social animal homo erectus being what it is, you will soon
have a pool of experts on most topics, and therefore a
possibility of expanding into an "ask the experts" forum as
well. Apply the same rating system to forum responses and
you have another fine resource of wikipedia articles,
the "vox populi articles" perhaps...
Articles could also be distributed for review and scoring by
email to the required number of reviewers, or else made
available much as the Distributed Proofreading Project is.
I for one would find this to be a great confidence booster,
I'm sure others would too. Also, it ensures that the
practical grassroots experts are found in each subject. A
database of email addresses of experts in many fields will be
the result. Important to note: You don't NEED to know if
the person answering all the questions on radium is Mme
Curie - what's important is that the person is right. Their
ranking in the wikipedia expert system is their correct title
as long as they produce the right information...
Awww ain't that all starry-eyed and idealistic... But
honestly, you could become the largest knowledge exchange in
the world by doing this.
Changing summary to be more catch-phrase intensive and hence "search-friendly".
Ted, did you check out the article validation system of MediaWiki 1.4 that's
currently live on [[test:]]? That implements at least a tiny part of what you want.
BTW the phrase "database of email" makes me shudder. Any spammers around?
This kind of stuff has been thouroughly discussed on the different mailinglists
(wikipeda-l, wikitech-l) for a looong time. You can try using gmane.org and
search for karma or perhaps "trust networks". In conclusion: its very un-wiki
and encourages gaming (note that I have not read this proposal), I'm afraid even
the current CVS (1.4) feature has far too much gaming potential.
[[test:]] doesn't seem to work. Use this link: http://test.wikipedia.org/.
Could someone please post a working link to a summary discussion of this type of
thing? I'm sure you guys are sick of it, but I bet a lot of people would like
to see the FAQ-style version of what's there for review, and why or why not.
Changing all WONTFIX high priority bugs to lowest priority (no mail should be generated since I turned it off for this.)