Page MenuHomePhabricator

Automate attribution claims on Wikipedia
Closed, DeclinedPublic

Description

Author: wikimedia.org

Description:
People often make an anonymous change to a Wikipedia page, and later want to
claim it as their own. The current method of fulfilling such requests is so
highly manual, and the demand for this service so great, that the page on which
users are supposed to make such requests has been changed to include a big red
box saying "Service suspended - Edits have not been reattributed for some
months. There are no indications when requests made here will be executed, if
ever."

I don't see why this is not simply an automatic service. There's a great demand
for it, and, although I am not familiar with the wiki code, it seems that it
should be almost trivial to implement. Specifically, something along these lines:

Just put a "Claim attribution" link in each page's "Toolbox" . If you click it,
and if your IP address matches any of the IP addresses of anonymous comments on
the page, then they're instantly and automatically assigned to your username.


Version: unspecified
Severity: enhancement

Details

Reference
bz3538

Event Timeline

bzimport raised the priority of this task from to Lowest.Nov 21 2014, 8:52 PM
bzimport set Reference to bz3538.
bzimport added a subscriber: Unknown Object (MLST).

Many (most?) IP addresses are dynamically assigned from a pool or represent a proxy
which may serve several, dozens, hundreds, or thousands of people.

Claiming all edits by an arbitrary IP without a human review can't be done.

wikimedia.org wrote:

Why not?

Exactly what "review" can be done that allows a person to determine that
Username X, currently at IP address a.b.c.d, currently claiming that he or she
wrote something that was written by someone at IP address a.b.c.d, is or is not
lying?

wikimedia.org wrote:

Assuming that "And that's why" is referring to what I wrote and by implication
saying that no such review can be done, rather than referring to the fact that
you just set it to "wontfix" and wanted to tersely imply that this conversation
is beneath you, then:

My point is that what I described is, as far as I can tell, the way that it
always has been done - a human checks that the IP address matches, and that's
the extent of the "review". That is something that can easily be done by a
computer.

So if the reason for the disabling of the manual version of this feature was
that it was too highly manual for the sheer volume of requests, then the fact
that there is no good "review" that can be done automatically is not relevant -
there's no good "review" that can be, or in fact was, done manually anyway.

Now if the reason for the disabling of the manual version was that there's no
good review that can be done, that's (potentially) another story. But just read
the "discussion" page for the "Changing attribution" page - maybe I missed
something, but there seems to be NO indication that this was disabled due to the
theoretical possibility of abuse. NONE.

On the other hand, there is a WHOLE LOT of discussion about how tedious this is,
how long it takes, that developers have (what they perceive as) better things to
do than manually fix attributions, and so forth.

And, once again, there is a great demand for this feature. Lots of people
accidentally post something only to realize that they're not logged in, or want
to take attribution for things that they wrote before they made an account.
Frankly, the lack of the ability to do this just plain sucks, and theoretical
concerns about the possibility of abuse seem to be almost irrelevant in comparison.

But if you're REALLY concerned about abuse, just add a flag indicating whether
or not a change was previously anonymous, but claimed by its current author.
People can take the attribution with a grain of salt, if they perceive any need
(which, frankly, seems unlikely at best).

I don't know what you're talking about. What's "the disabling of the manual version of
this feature" refer to?

wikimedia.org wrote:

It was a longstanding policy that, when a person wanted to claim attribution for
an anonymous comment, they post a request on the page "Wikipedia:Changing
attribution for an edit"
(http://en.wikipedia.org/wiki/Wikipedia:Changing_attribution_for_an_edit).

They would, while logged in, say what page/revision they were claiming, then log
out, and, using only their IP address, would say something like "Yeah, this is
really me".

Theoretically, someone with appropriate permissions would then update the database.

I say "theoretically" because this would often take a LONG time - weeks, even
months. Because nobody really wanted to spend their time doing this. But at
the same time, there was a HUGE demand for it. So the backlog got greater and
greater and greater. And the exact way in which requests were made was
constantly morphing in an attempt to make the process more efficient - for
example, "Type your request into a table, laid out like such-and-such". But it
never helped - demand always outstripped supply, by a huge margin.

Every so often, some kind soul with appropriate permission would take it upon
themselves to try to make a dent. But that was rare, and the demands always
reflooded soon after anyway.

Eventually, the service was "suspended", as noted in a big red box at the top of
the page:

"Service suspended

Edits have not been reattributed for some months.
There are no indications when requests made here will be executed, if ever."

It is, frankly, almost unbelievable that the reason for this was theoretical
concerns of abuse, given that:

(1) Such abuse always could have been done;

(2) It was always obvious that such abuse could have been done;

(3) There was no (or virtually no) discussion on the importance of such
theoretical abuse;

(4) There was LOTS of discussion about how tedious it was, and how demand was
too great for the people with appropriate permissions to keep up with.

I don't know where you got this idea of the "reason"; the reason that manual
reattribution was suspended was that it required direct database manipulation available
only to a small number of cluster administrators.

That's why we now have Special:Renameuser, allowing relatively quick reattributions to
be run by existing trusted users on the wikis.

avarab wrote:

(In reply to comment #7)

That's why we now have Special:Renameuser, allowing relatively quick

reattributions to

be run by existing trusted users on the wikis.

Special:Renameuser doesn't do attributions, only existing user => nonexisting user

wikimedia.org wrote:

"I don't know where you got this idea of the "reason"; the reason that manual
reattribution was suspended was that it required direct database manipulation
available only to a small number of cluster administrators."

Which is not necessary.

Unless you can explain to me what relevant task a cluster administrator could do
that a "Claim attribution" link on each page cannot do, this is simply not germane.

I'm not claiming that there is no such task, but you haven't pointed one out
yet, and earlier posts of yours seemed to imply that there is none.

Because Special:Renameuser hadn't been written yet.

wikimedia.org wrote:

"Because Special:Renameuser hadn't been written yet."

Could you please explain how that is relevant?

First, Ævar Arnfjörð Bjarmason just indicated that Special:Renameuser is not
applicable to this discussion.

Second, even if it is applicable, then if it was written to allow "relatively
quick reattributions" (as you said), then why are reattributions not being done?
Or if they are being done, why does the page where users request reattributions
indicate that, hey, sorry guys, you're out of luck? Either it is

(A) Not applicable;

(B) A poor solution to a problem with an easy solution; or

(C) Not being used.

Or else the page where people request reattributions should be changed to tell
people how to submit reattributions to this wonderful new thing, instead of
telling people in a big red box that it's just not gonna happen.

Special:Renameuser doesn't do everything yet; currently it hits the primary easy case
of changing existing user accounts to non-overlapping new names.

Resolving as WONTFIX again as we will under no circumstances add the requested feature
(allowing anyone editing from a certain IP to claim all edits from that IP to their
account without review).

If you want to request further improvements to Special:Renameuser to handle the other
use cases, file a bug for that.

wikimedia.org wrote:

"Resolving as WONTFIX again as we will under no circumstances add the requested
feature"

Fine, but:

"(allowing anyone editing from a certain IP to claim all edits from that IP to
their account without review)"

You said this was the problem before, and I asked you a question in response.
Twice, in fact. You did not answer it, but here you are claiming the problem
again. So, I will ask you again:

What relevant review can a human do that a computer cannot?

A human can read the edits and make an educated guess about whether it looks like they
are likely from multiple different people, and can refuse to make the change if it
seems to be an illegitimate request.

wikimedia.org wrote:

Okay, it's now Bug 3539, framed in terms of Special:Renameuser.

wikimedia.org wrote:

"A human can read the edits and make an educated guess about whether it looks
like they are likely from multiple different people, and can refuse to make the
change if it seems to be an illegitimate request."

I would enjoy seeing an example, in any of the many, many reattributions that
were done in the past, of this being anything but theoretical.