[Foundation-l] [WikiEN-l] Talk pages Considered Harmful (for references)

[Gwern Branwen]

An update: I managed to fix the double-counting problem I mentioned
was skewing the numbers upwards, and fixed a few other issues. (In
retrospect, the solution was almost trivial: just discard any URL that
appears *twice* in the diff, since none of the edits would repeat an
added link.)

The updated numbers are:

- My anime references: <8%
- My non-anime references: <3%
- Krebmarkt's references: <4%
- Total references used: <4.15% of 1206

As one would expect from fixes removing false positives, all the new
figures are smaller. I invite people to go through and double-check -
everything you need is provided.

On Thu, Dec 22, 2011 at 1:10 PM, Ken Arromdee <arromdee at rahul.net> wrote:
> The rest of that, about deletionism, may be at least as interesting.

Or it's a rant, depends on your own inclinations, I think. (I do well
on things like belief calibration and avoided political bias on tests,
but who knows whether my beliefs on Wikipedia are correct.) Sue
Gardner liked it, at least.

> I wonder how the ban on canvassing is affecting deletion.  Our system is set
> up so that informing the very people who would be affected most by deleting
> an article is not permitted.  (And of course, we have WP:OWN, which prevents
> even *recognizing* that some people may have a particular interest in an
> article not being deleted.)

It helps deletion, unsurprisingly; see the study quoted & linked in
http://www.gwern.net/In%20Defense%20Of%20Inclusionism#fn22

On Thu, Dec 22, 2011 at 1:48 PM, Rob <gamaliel8 at gmail.com> wrote:
> This makes a lot of sense.  Many times I've removed these from the
> article for valid reasons - text/link dumps, mal- or unformed
> sections, etc. - and placed them on talk so editors could use them for
> future edits.

They don't use them, as I've shown.

On Thu, Dec 22, 2011 at 9:46 PM,  <kgorman at berkeley.edu> wrote:
> This rate, without additional context, is meaningless.  As Rob pointed
> out, there are many different reasons for moving
> references/links/citations from an article to a talk page, and unless you
> have more information about why people are moving these to talk pages, the
> rate at which they move back doesn't really mean anything. By labeling
> this rate a 'failure rate' you are strongly implying that success would be
> keeping the link in the article.  I don't believe this is right - I
> believe that 'success' is doing what's best for the article.
>
> Even if 99% of things that were moved to talk pages were not subsequently
> returned, I would not find this at all disturbing without evidence that a
> large portion of the removed things should not have been removed.
> Frankly, I would be surprised if 10% of things that I personally moved to
> talk pages were moved back in to the article space.

You and Rob have apparently completely missed the point of the
exercise, the reason why I invested so much manual effort into this.

I didn't look at a bunch of anonymous edits, precisely because I
*knew* someone would say 'oh they're from dirty anonymouses and so
they are probably crappy links - why be bothered by a 10% or a 1%
rate?' This is wrong, but it has a surface plausibility and there's no
point in compiling data that can be so glibly dismissed.

So I looked *only* at known good links, links I and Krebmarkt had
hand-selected as useful. Again, feel free to go through the links and
look at them! My first 2 anime links were RSs for a director's next
movie, and box office receipts; Krebmarkt's first 2 links were RS
critics' reviews for manga that both have (note the present tense) 0
reviews in their articles. And so on.

There is a known rate at which these links ought to be included. It's
>90%. (I am being charitable in not saying 99% or 100%.) The actual
inclusion rate is <10%. The difference should bother us.

-- 
gwern
http://www.gwern.net/In%20Defense%20Of%20Inclusionism#the-editing-community-is-dead-who-killed-it