Wikipedia:Pending changes/Request for Comment February 2011/Archive 1: Difference between revisions
Content deleted Content added
MalnadachBot (talk | contribs) m Replaced obsolete font tags and reduced Lint errors. (Task 12) |
WOSlinkerBot (talk | contribs) m Fix misnested tag lint errors |
||
Line 557:
=== PC was a trial, so where are the results? ===
It is completely pointless discussing where we should go next without some data. I haven't read all of this massive page but it seems to be repeated over and over that there is no analysis being produced. The trial was pointless if it cannot produce any data. What I am looking for is some metrics on performance, especially comparing articles protected by PC against articles protected by some other means or unprotected. [[User:Spinningspark|'''<
:Results is a bit like fishing for ghosts. You might be wanting to know - how many defamatory comments especially in relation to living people were stopped from being published via wikipedia and mirrored around the www on articles that would not have been protected at all, and considering the more open edit situation that pending offered in comparison to semi protection, how many unconfirmed IP users were able to make an accepted addition when they may have not bothered when the had to request it on the talkpage, questions regarding such figures are unanswerable. We can say, the more low watched BLP articles we add pending protection to the more defamation we will stop publishing all over the web for sometimes lengthy periods of time, and we know that unconfirmed IP users have benefited from the ability to post direct to the article when semi protection would have stopped them. [[User:Off2riorob|Off2riorob]] ([[User talk:Off2riorob|talk]]) 00:49, 23 February 2011 (UTC)
::It is only like fishing for ghosts because we never got a proper trial. Without a proper trial all we have are people like you who in good faith can make claims about how the unknowables point in your favor and people like me who can (again, in good faith) claim the opposite. If we had an actual trial of the tool we could make real empirical claims. Until that point we ae really just arguing from priors, which will never result in one side convincing the other of anything. [[User:Protonk|Protonk]] ([[User talk:Protonk|talk]]) 01:19, 23 February 2011 (UTC)
Line 563:
::::There are a few points to be made here. First, if the downsides of PC represent intangibles or unknowables then we should be more cautious about implementing it rather than less. We should be respectful of the unknown, not derisive. Second, statistical inference is not alchemy. It is entirely possible to create a sample of articles subject to long term semiprotection which are similar along a number of dimensions (e.g. page views, reversions, edits per editor, edits as a whole, etc.) and split that sample into a control and treatment. The treatment group is assigned PC for the duration of the trial. The control remain semi-protected. You can then compare outcomes for the two batches of articles. Obviously there will be a [[Hawthorne effect]], but this can be minimized by not publicizing the articles in the treatment (even though people will find out) and never revealing the articles in the control. compare this to the mess we had with the actual PC "trial" where the tool was turned on and demoed but no inference about its efficacy or impact is possible. [[User:Protonk|Protonk]] ([[User talk:Protonk|talk]]) 03:07, 23 February 2011 (UTC)
::::I should also point out that the total sample doesn't have to be very large. Something like 200-300 articles total would be more than sufficient, even if the effects we are measuring are small. [[User:Protonk|Protonk]] ([[User talk:Protonk|talk]]) 03:22, 23 February 2011 (UTC)
::::: Off2rioro complains that we are not specifying what kind of statistics we are looking for. Let me be clear, my question here is are there ''any'' statistics whatsoever giving ''any'' kind of metric. This discussion is woefully short of facts. Apparently, the only fact that can be stated with any certainty is that there is no ''evidence'' that PC has any benefit. [[User:Spinningspark|'''<
::::::I can give you some statistics. I've used PC on some 40 pages. Each page (100%) contained (search-indexed) vandalism before it was PCprotected. None of the pages (0%) showed any vandalism while PC was in effect - in Google or to casual browsers. 100% of editors were able to edit each article while the protection was in place. Those benefits seems fairly convincing to me. -- [[user:zzuuzz|zzuuzz]] <sup>[[user_talk:zzuuzz|(talk)]]</sup> 08:45, 23 February 2011 (UTC)
:::::::This is a fallacy. Spammers take the easiest path. In this case, they just go spam something else instead. Do you want to take on a trial of putting PC on all of Wikipedia? We could never keep up with the reviewing load and the distraction would be equivalent to having everyone fight at the front door while the vandals go blow the entire back wall off the place. How many subsequent good IP edits were lost because editors didn't see that their efforts had any effect? These problems make a statistical analysis of PC impossible. Please don't waste your or our time. —[[User:UncleDouggie|UncleDouggie]] ([[User talk:UncleDouggie|talk]]) 09:12, 23 February 2011 (UTC)
Line 571:
:::::::Data is not the plural of anecdote. [[User:Protonk|Protonk]] ([[User talk:Protonk|talk]]) 18:04, 23 February 2011 (UTC)
::::::::Second that. [[User:Spinningspark|'''<
* - As I said, and as Zzuuzz appears to have understood very well, the benefits to the project are loud and proud. The issue is do you care? Are you bothered that defamatory content sits in our articles about semi notable people unnoticed for months? Do you think publicity about such things are bad or good for the project? If you gave the living subjects a vote for this protection to be kept or switched off what do you think they would vote? Do you feel any responsibility to the subjects of our articles? How many of our living subjects do you think would like to ''opt out'' of the project if we refuse to protect their article from defamatory additions? Ask yourselves questions like this. Its a difficult position to be in imo when we have an almost totally unidentified not notable user base rejecting additional protection for the identified notable subjects of our articles. It is absolutely undeniable that pending protection keeps defamatory content from being inserted and remaining in the BLP articles of semi notable low watched articles and through insertion and publication in our article also mirrored all over the world wide web. If you reject this then what do you support to stop such additions? Liberal indefinite semi protection? Allowing living subjects to opt out? Account only editing? Have a look at this, the latest of a constant stream from the BLPN, [http://en.wikipedia.org/w/index.php?title=Mahmoud_Hassani_Sorkhi&action=history defamed as a criminal on the 24 November], removed three months later.[[User:Off2riorob|Off2riorob]] ([[User talk:Off2riorob|talk]]) 11:41, 23 February 2011 (UTC)
*:So let's define our requirements. PC is a rather inefficient, error-prone tool. If you're so worried about BLPs, and with good reason, let's find a real way to protect them. How about starting with the low hanging fruit like having the edit filter reject any BLP edit without an edit summary? The next step is a big red warning/captcha/flashing neon thing if an unconfirmed user changes anything more than punctuation in a BLP without including a reference. —[[User:UncleDouggie|UncleDouggie]] ([[User talk:UncleDouggie|talk]]) 13:28, 23 February 2011 (UTC)
Line 599:
::::::Well, I don't like it, we have enough vandal fighting tools is just not cutting the ice. I think I have expressed my position pretty clearly, so, removed from my watchlist, let me know what you decide. [[User:Off2riorob|Off2riorob]] ([[User talk:Off2riorob|talk]]) 21:50, 23 February 2011 (UTC)
:::::::How often does this actually happen? How much does it matter when it does? How often is an IP responsible? How many IP vandals wouldn't set up accounts to do it anyway? Seriously, if you want to protect frequently vandalized articles, ''surgical'' use of PC as an alternative to semi protection might make sense, but when you're talking about finding miscellaneous slanders out of the whole database, maybe searches for "pedophile" and various vulgar terms might work better. Speaking of which, we could use some kind of special terms for the Search box that limit a search to BLPs, rarely edited articles, newly created articles, uncategorized articles, etc. so that we can search "Special:BLPs Special:unedited-7-days fucked" and get back an answer. [[User:Wnt|Wnt]] ([[User talk:Wnt|talk]]) 05:07, 24 February 2011 (UTC)
::::::::Caution needs to be exercised providing tools that find weak articles. For instance identifying unwatched articles has always been considered a bad idea as it is more useful to trolls than those trying to protect articles. [[User:Spinningspark|'''<
:::::::::I think Wikipedia is too deep into the Knowledge Is Good business to be having second thoughts now. It's not just that the whole database is public and someone else could offer the tool, but more that we otherwise assume that good editors can outwit vandals, with positive results. In any case you should at least have a tool that looks for unwatched BLPs with juicy terms in them and sends those out to a group of editors to go over. [[User:Wnt|Wnt]] ([[User talk:Wnt|talk]]) 16:32, 24 February 2011 (UTC)
::::::::::We have tools that do this already. [[WP:STiki]] flags high risk edits for review using it's own algorithms. It also now has a mode that routes the "overflow" from [[User:ClueBot NG]] for review that consists of edits below the automatic revert threshold and reverts that would otherwise violate the 1RR rule for bots. —[[User:UncleDouggie|UncleDouggie]] ([[User talk:UncleDouggie|talk]]) 21:30, 24 February 2011 (UTC)
Line 606:
==Alternatives to PC==
Are any alternatives to PC being considered? As Wnt stated above, the major downside of PC is that it prevents a good faith IP editor from immediately seeing the result. Wiki is meant to be a form that does just that - immediate. Any protection scheme that builds walls around articles is going to be bad from that point of view. Far better that the guilty are directly attacked rather than the innocent editors harmed through colleteral damage. It seems to me that the biggest problem is shared IP addresses and IP hopping connections. Problem editor on a fixed IP or with an account are rapidly stamped on. Technical means of distinguishing different editors on the same IP or the same editor on different IPs would largely obviate the need to protect articles at all. I am thinking here, for instance, of AOLs [[X-Forwarded-For]]. I really don't have much clue over what else may be available but ir may be worth opening this discussion up a bit beyond the narrow polarised for/against PC this has became. [[User:Spinningspark|'''<
:Makes sense to me. See also [[#Should we define our requirements first?]] [[User:Yaris678|Yaris678]] ([[User talk:Yaris678|talk]]) 12:40, 24 February 2011 (UTC)
::PC is less restrictive than semi-protection, which is the main alternative. It enables unregistered and new users to make edits directly to the article, whereas they would otherwise be locked out completely. We have already [[WP:XFF]], active checkusers, some very skilled and dedicated admins, swift indefblocks, IRC feeds, various automated screening tools, and massive rangeblocks in place. None of these can prevent vandalism, sockpuppets, dynamic IPs, or open proxies slipping through. Blocks and patrollers are not all they're cracked up to be. Protection is an ''essential'' tool, and PC is just another form of protection. -- [[user:zzuuzz|zzuuzz]] <sup>[[user_talk:zzuuzz|(talk)]]</sup> 13:52, 24 February 2011 (UTC)
| |||