[Wikipedia-l] Major new thread: Looking for consensus on content metadata

[Jimmy Wales]

(At least, I hope it is a major new thread!)

Well, we've argued for a few days, and a lot of ideas have been thrown
around, and the tension between various competing principles and/or
ideals has been explored fairly effectively.

So I wonder if we could work towards some consensus that we can all
(or nearly all) agree on.  What steps might we take that all parties
to the discussion might agree on?

Can everyone chip into this thread with accomodative ideas that you
think everyone might agree with?

By way of example, in the "fair use" debate, the one thing that all
parties could agree on is the importance of prioritizing the _tagging_
of images with their status.  Whatever might end up being done, even
if -nothing much- is the answer, that tagging is something that no one
has objected to.

So what might we do in the content metadata arena that no one (or
nearly no one) would object to?

I think that there is broad support for a categorization system which
is broad and flexible, and not necessarily aimed at content
advisories, but which might, in part, be used to address that issue as
well.

Here's what I propose -- and I'm talking at the level of policy, not
at the level of technical implementation, because there's a lot more
to be thought about and said in that regard -- is that we move towards
the implementation of a content metadata or categorization system with
the following features:

1.  Categories should be non-normative in nature.  That is, "mature
content" is an invalid category, because it suggests a value judgment
that we want to leave to the end user.  "Explicit sexual content",
while still perhaps vague in some respects, is at least
non-normative... it might be a good thing or a bad thing, depending on
your purpose.

2.  Categories should be infinitely wiki-editable, especially in the
beginning, because "a priori" categorization is impossible and likely
to lead to a lot of problems.  Ideas that only sysops can create new
categories should be avoided until we actually determine empirically
that it's necessary.  (One thing we know from our wiki experience is
that something as obviously insane as letting anyone in the world edit
works amazingly well!)

3.  We should quite possibly accept _as an editing principle_, a sort
of Ockham's razor for categories -- not multiplying them needlessly,
while at the same time, not hesitating to let people experiment with
categories as they see fit.  But, we'll try not to fight about it too
much, especially at first.

4.  The categorization system should be simple -- i.e., articles can
be tagged with as many categories as we like, and that's that.  They
are not required, and if people want to work on them, they can, and if
they don't want to work on them, they don't have to do so.

5.  Especially initially, website impacts of categories should be very
minimal... i.e. we don't try anything radical regarding filtered
searches or automatic index pages or anything too exciting like that.

--------------

If we did that, we'd be introducing nothing harmful, and doing something
positive for *multiple* purposes, not just the "content advisory" purpose.
(See below...)

And we'd get to find out, in an experimental environment, whether
categorization prompts massive flamewars, etc.

And then in a while, we can revisit the issue, and see what we think
could be done with these categories.

--Jimbo

p.s. Examples of useful categories that are non-normative:

biography
math
statistics
graphic sexual content
advanced mathematical content
gay and lesbian studies
European history
American history
U.S. Presidents
sports
tennis

All of these might be useful for content re-users who might like to
extract a subset of our data.  For example, an "Encyclopedia of Sexual
Practices" might want to extract just those articles flagged with
'graphic sexual content' or 'sexual content'. A "edupedia" project
could automatically extract articles that avoid certain topics, or
focus on certain other topics.




----- End forwarded message -----