[Foundation-l] Oral Citations project: People are Knowledge

[David Gerard]

The great thing about an oral history citations project is that it is
a first and active method to remedy one of the big problems with
English Wikipedia: the epistemology - how we decide we know what we
know - really is completely and utterly broken at the edges.

(I realise this is foundation-l, but en:wp is a third of Wikimedia by
most measures, and this discussion shows its ways of doing things
getting into everywhere else.)

The trouble is that all through history, turning information into
knowledge has required human judgement and nuance. People do four-year
humanities degrees to *start* getting *any good at all* at this stuff.
But Wikipedia being Wikipedia, the whole thing has to be (a) reduced
to a three paragraph guideline (b) which calcifies into policy (c)
misinterpreted by socially-inept teenagers (d) with the
misinterpretations being perpetuated well past the point of actual
failure.

Thus we end up with blithering insanity like the phrase "reliable
sources" being used unironically, as if being listed on WP:RS
*actually makes a source humanly reliable*. This is particularly
hilarious when applied to newspapers - no-one who has *ever* been
quoted by the media would think this way.

(For those of you aware of the hip Bayesian way to calculate
uncertainty, this is what happens when your network has allowed
probabilities of 1 or 0.)

Now, the sourcing method we have almost works. Its successes are
important and useful. But there's a lot of denial that it breaks
really badly when misapplied, and that the misapplications are even a
problem. WJohnson's earnestly put forward this viewpoint in this
thread; his argument appears to be that we don't have a perfect
solution so therefore this must not be a problem and doing something
that doesn't work *harder* must be the right answer.

Somehow we have to get the nuance back. All this stuff is produced by
humans, and working assumptions that it isn't are *broken*.

The oral citations project appears to be a first step to even
acknowledging that the present methods actually break at the edges.
This alone makes it a good and useful thing. And, y'know, we might
actually learn something.


- d.