Misplaced Pages

Revision as of 22:20, 30 April 2010

SBHB's outside view

Regarding Misplaced Pages:Requests for comment/Lar#Outside view by Short Brigade Harvester Boris

Lar can be banned from acting as an uninvoled admin altogether based on Disruption

• The diffs show that instead of commenting on cases as an uninvolved admin Lar is being rude and dissmissive to a large group of very dedicated editors.
• Lar is regularly taking the oportunity to push his view rather than acting as a neutral arbiter and evaluating the cases based on merit.

His presence as an enforcement admin is highly disruptive. Polargeo (talk) 08:29, 30 April 2010 (UTC)

I do not think those diffs show what you think they show. I think while you may well have legitimate criticism of me to offer, some of which I have acknowledged, you are overreaching when you characterize my presence as an enforcement admin as "highly disruptive". ++Lar: t/c 14:11, 30 April 2010 (UTC)

I wouldn't think it was highly disruptive if I didn't think you were letting what I see as your prejudices and misconceptions manifest themselves as extreme bias, which I think is a very worrying characteristic for an "uninvolved" admin. Polargeo (talk) 16:48, 30 April 2010 (UTC)

On whether a "whatchamacallit" exists

Here, Enric Naval states "Lar really seems to believe that a cadre exists, and that WMC is leading it" and characterises this belief as a "delusion".

So let's examine this more closely. (note, to avoid unwieldy wording I will have to use a short term here, and I will use "AGW cadre". I have struggled with what a neutral short form would be for some time now, and welcome better suggestions)

First, is it even possible that such a group exists? If we had never seen any such group before, it would be a good argument that existence is improbable.

But we have seen such groups before. In the findings of fact of the EEML case, ArbCom found that there was a group of editors who were

commenting in process along "party lines", supporting each other in disputes even when otherwise uninvolved in them

and that they were using tactics including

baiting, harassment and vexatious complaints against specific users in order to have them sanctioned or driven away from participating.

So clearly we have examples of the existence of groups. Is the "AGW cadre" such a group?

The EEML group was a highly coordinated group. They used sophisticated techniques to carry out activities that subverted consensus and poisoned the editing atmosphere. While others have alleged that the "AGW cadre" have coordinated activities using emails or using Facebook. I have made no such claims, and I repudiate any such allegations until and unless evidence is introduced to substantiate them.

Making such claims is not likely to be helpful either.

But it is not necessary to use a mailing list or IRC or FB or whatever. Merely watching the activities of folk is sufficient to determine what's going on, and what "hot spots" might exist. I watch my wife's contribs and she watches mine, for this very reason. Perfectly legitimate and we all do it all the time, as there's nothing inherently wrong with it and often it can be quite helpful. (the contributions page is a tool, and like any tool, has no intent of its own, the intent is with the user)

If editors have overlapping areas of interest, watching contributions can be quite effective. Consider this wikistalk result focusing on the overlap in mainspace contributions of WMC and 5 other editors that many folk, including folk other than myself, consider among those that are candidate members of the "AGW cadre". Note the areas of overlap. Almost every mainspace article that at least three of these editors have edited is an AGW article. (the few that are not are in other areas that have had some challenges in the past... fringe science such as Cold Fusion, and creationism/ID)

That demonstrates only that these editors all have an interest in AGW, and that it overlaps with each other. Running similar analysis on other groups of editors (for example those who many could consider members of the "denier cadre") would probably show similar results. Nevertheless, it DOES show an interest in AGW (which I think none of them would deny, but it's necessary to show it if we wish to be rigorous)

Let's look at who edits the enforcement request pages: This tool: when we put "Misplaced Pages:General sanctions/Climate change probation/Requests for enforcement" in as the page to review, shows that these editors do frequent the enforcement page.

So too do others. My contributions to that page are pretty high too. So this shows only that this group of editors, with a clear interest in AGW articles, also has an interest in the enforcement page.

However, that by itself is sufficient to show that there is an "AGW cadre". But not that there is a problem that we need to act on, that is, that the editors who are statistically correlated to this cadre have malign intent, or that their activities are overall harmful. Remember, that we encourage the existence of "groups of related editors"... we call them "WikiProjects".

I'm out of time for right now. Others have put forward information around some of those points. I may follow up later, but for right now, do we have general acceptance that such a cadre does exist (regardless of anything else about it such as who might be its leader (if such exists... I've not directly made such a claim) or what have you)? Or is there a need to demonstrate this existence in more detail? ++Lar: t/c 14:11, 30 April 2010 (UTC)

I'm not sure what your wikistalk result purports to show, but I suggest that if you ran the same whatever on Cla68, Heyitspeter, GoRight, MarkNutley and Atren, you'd find a far, far stronger whatever the tested variable is. Hipocrite (talk) 15:16, 30 April 2010 (UTC)

I believe this is the link he meant to do. And here is your suggested comparison. Of course, if you included Short Brigade Harvertor Boris' previous account then the correlation would be even stronger. TheGoodLocust (talk) 16:47, 30 April 2010 (UTC)

I don't even know what those things purport to mean. Could someone describe what the column "No." and "Similarity" are, and what the rank order of the various pages are? Hipocrite (talk) 16:50, 30 April 2010 (UTC)

Well, I don't use that tool often but "No" appears to be the arbitrarily defined number used to identify the area of overlap . It appears to be buggy as well since it doesn't increment sequentially correctly (nor does it seem to put them all in reverse alphabetical order as apparently intended). You'll notice in the first example that after the number 16 it jumps up a huge number and then continues incrementing++ until later on when it jumps up again unexpectedly.

The "similarity," as far as I can tell, shows two things; first, it shows how many of the group have participated in that article (e.g. 2 of 6, 6 of 6), and the 2nd column identifies those individuals by their previously assigned number. The limitations of this are obviously that it doesn't show the quantity of activity overlap, simply the areas of overlap, and, as previously mentioned, it doesn't show the overlap from previous accounts (a concern in this case). TheGoodLocust (talk) 17:04, 30 April 2010 (UTC)

The odd jumps in sequence appear to occur when it starts displaying the next person's edits. This may be intentional, but it certainly is ugly from a UI perspective. TheGoodLocust (talk) 17:20, 30 April 2010 (UTC)

If that's accurate, there are fourteen-hundred plus items that only person 1 edited, 11 hundred plus that only person 5 edited, 150 that only person 4 edited, and a massive four thousand that only person 3 edited. This seems to be evidence that if you have people with ten of thousands of edits, they overlap. I wonder how many std. devs away from random chance these results are. Hipocrite (talk) 17:33, 30 April 2010 (UTC)

No, you are misreading it, I said the numbers don't increment sequentially correctly and should mostly be ignored. Every single article mentioned shows overlap between at least two people - every one. Also, you are incorrectly assuming that these count edits that overlap (which would be difficult to do) - these counts areas of overlap. You can't compare apples to oranges (edits to areas).

From my count the article overlap, and keep in mind this is just articles, not talk and other areas, includes approx. 520 articles (not counting SBHH's previous account). The overlap is considerable if you consider the fact that they make thousands of edits (usually reverting others?) in these articles. For comparison, the group you suggested we compare edited in 33 of the same articles. TheGoodLocust (talk) 18:01, 30 April 2010 (UTC)

Thanks for the analysis... ++Lar: t/c 18:33, 30 April 2010 (UTC)

Dividing that by median articles edited might show something. Which group would you say has greater coorelation? Hipocrite (talk) 18:27, 30 April 2010 (UTC)

Actually the best thing to do would be to look at the # of edits they make to each of those overlapping articles. I recall several of these editors making hundreds or even thousands of edits in several of those overlapping articles. I have neither the time nor the tools to make a proper analysis (and my statistics is quite rusty). Cheers though. TheGoodLocust (talk) 18:40, 30 April 2010 (UTC)

Just to be clear: So far, due to time reasons, all I think I've established is that there is a (at least one) group of editors that have an interest in this topic area, as evidenced by the number of articles in the topic area (and, oddly, in related topic areas such as ID, creationism, fringe science, ect) that at least two of these editors have edited. Although I haven't done a statistical analysis, I'd say that the overlap between these 6 editors (there are others I feel that are in this group, I just picked those six as a starting point) is several standard deviations away from the overlap we'd see among 6 editors of comparable edit counts picked randomly. Hipocrite, did you seriously want to debate that point? I really didn't think that it was debatable, actually. I was just trying to be rigorous.

Note that we could well find other groups of 6 editors that have similar overlap in this area. I make no claims that we can't. Just that I found this one group. That's all, so far. ++Lar: t/c 18:33, 30 April 2010 (UTC)

Lar, as someone who does work in a similar-esque kind of area, yes, I'd make the argument that if you picked six editor who all edited one article extensively that the "pattern" you alledge is hardly interesting. IE, pick six editors who edited Al-Qaeda extensively but also have massive edit counts, and you'll find an identical overlap. What is interesting is that there are editors who overlap with these editors, and eachother only in this topic area. IE - there are gadflies just here to fuck things up, and they are on both sides. Why are Atren and WMC showing up at The Hockey Stick Illusion to throw a wrench into things? Hipocrite (talk) 18:53, 30 April 2010 (UTC)

My concern isn't whether you personally find the pattern interesting or not, just that you (and everyone else for that matter) acknowledge that the pattern exists. You may want to ask the 6 editors I used whether they want to deny that they have an interest in this area and edit in it extensively. So far that's all I'm trying to show. Gotta fly. ++Lar: t/c 19:06, 30 April 2010 (UTC)

It seems that what you're saying is so trivial as to be irrelevent. Probably next time you want to enter a trivial fact into evidence, you should just state it clearly, like you did here. I'll cop to having an interest in the area, and I edit it, for various definitions, "extensively." Here's a list of other people who have to cop to that also - Cla68, marknutley, Atren, heyitspeter, and so on and so forth. Big "sowhat" here, to be honest. Hipocrite (talk) 19:13, 30 April 2010 (UTC)

Oh, and I compared 6 editors who stood out in the Al Qaeda history and here are the results. They overlap 58 times and the only article that they've all edited was Al Qaeda itself (far less overlap when you look at the details). TheGoodLocust (talk) 19:22, 30 April 2010 (UTC)

(undent) you didn't correct for edit count. Hipocrite (talk) 19:22, 30 April 2010 (UTC)

Nor am I going to go through every single Al Qaeda editor and look at their edit counts. Each of those editors have a good # of edits and I doubt you'll find many (any?) editors to that article with as many edits as the AGW group (it is much easier to push up the edit count when reverting). TheGoodLocust (talk) 19:26, 30 April 2010 (UTC)

Note:

Group 1 (ATren et al.) - 5.7% of the articles on their list are edited by all members of the group (average 513 unique pages per group member edited in Misplaced Pages)

Group 2 (WMC et al.) - 0.8% of the articles are edited by all members of the group (average 5950 unique pages per group member edited in Misplaced Pages)

Assuming a simple linear model between editing and coincidence , it seems to me that you'd expect the proportion of coincidence in the latter group to be an order of magnitude higher. Now add to the fact that the latter group has 37% of its edits in mainspace (unweighted average per person), while the former group has 23% of its edits in mainspace (unweighted average), you'll see that the likelihood of the people in Group 2 editing the same articles is much higher than the likelihood of the people in Group 1 and the appearance of conspiracy should be much higher for the first group. Unless, of course, you approach the issue with a strong confirmation bias. Then you'll see what you expect to see. Guettarda (talk) 19:59, 30 April 2010 (UTC)

Did you correct for the number of users? You might need to add in another editor to group 1. Hipocrite (talk) 20:02, 30 April 2010 (UTC)

No, Group 1 has 4, Group 2 has 6. I didn't pick the people. But the point is that Group 1 has an order of magnitude higher coincidence with an order of magnitude fewer pages edited. Which suggests that anyone looking at these pages and saying "Group 2 is a whatchamacallit", but not saying that about Group 1 has misinterpreted the numbers. Guettarda (talk) 20:06, 30 April 2010 (UTC)

Obviously the group with fewer people will have a greater % of complete interesections if based on random chance (and also since their sample size (4 articles) is smaller). Also, I looked at SBHB's account and compared it with WMC's and together they have 173 articles in common (I didn't count repeats). I'm just pointing that out since his older account adds quite a bit more overlap as well. TheGoodLocust (talk) 20:10, 30 April 2010 (UTC)

The simplest comparison would be to compare groups of two - get rid of some of that noise. TheGoodLocust (talk) 20:13, 30 April 2010 (UTC)

Actually, we should stop using the tool, as it's been declared "busted" by it's creator. So right, all of this was a waste of brain. Hipocrite (talk) 20:16, 30 April 2010 (UTC)

I believe he was referring to the broken enumeration, which I'd already pointed out was broken, but which doesn't really matter in any substanitive way. TheGoodLocust (talk) 20:47, 30 April 2010 (UTC)

But deleting Dave and me destroys all of Lar's conspiracy theories, since we're now talking about 4 people with advanced degrees in (pretty much) the same field. Not to mention that with an order or magnitude more edits, you'd expect an order of magnitude more intersections. So again, these are entirely innocuous numbers. Guettarda (talk) 20:49, 30 April 2010 (UTC)

Deleting you and Dave means the groups can be compared more accurately. Also, I compared the % of complete intersections between the two groups, something which you brought up and which, if edits were random, would decrease as edit counts rose - instead over 10% (likely far higher w/ Boris' old account included) of their shared articles (35+ articles - which they edit a LOT) were edited by every single one of them. TheGoodLocust (talk) 21:08, 30 April 2010 (UTC)

Oh, and since you decided to look up %'s for group of 4 vs a group of 6, I turned the group of 6 into a group of 4 (SBHB, WMC, KDP, and SS) and found that 10.8% of their overlaps were complete (they all edited in those articles), which is a much stronger figure since that was based on 35+ samples of complete overlap while your 5% figure was based upon 2 articles of complete overap. Also, I didn't take into account SBHB's previous account, which, as I showed, adds a significant amount of overlap. TheGoodLocust (talk) 20:23, 30 April 2010 (UTC)

Yes, more edits = more overlap. That's my basic thesis. And yet group 1, with a fraction the number of edits, has similar or greater proportion of overlap. So it undermines Lar's conclusion that there's some sort of evidence here against Group 2. Against Group 1 - maybe. After all, nearly all of their similarities are in climate change articles, while the rest of us overlap in all sorts of random places (Buju Banton, Great Pyramids...) Guettarda (talk) 20:52, 30 April 2010 (UTC)

More overlap, but less complete overlap (something you brought up). Also, the other group, which has far fewer edits, isn't nearly as statistically robust (e.g. when you said 5%, but based that off a mere 2 articles of complete overlap). As I previously stated, if we could compare the number of edits in the overlapping areas it would look far worse for WMC et all. TheGoodLocust (talk) 21:08, 30 April 2010 (UTC)

Oh, and for fun I included SBHB's old account and discovered the articles of complete overlap, where they all participate, goes up to 55. I'd be curious if there is any other group that has that much complete articlespace overlap. TheGoodLocust (talk) 21:47, 30 April 2010 (UTC)

Complete overlap for all spaces on wikipedia goes up to 160 pages for that group. TheGoodLocust (talk) 21:52, 30 April 2010 (UTC)

rfc/u?

Should this not be posted here  ? mark nutley (talk) 21:50, 30 April 2010 (UTC)

Yep. Thanks Mark Polargeo (talk) 22:03, 30 April 2010 (UTC)

Should this be in the admins section? Not sure because it is not a complaint against use of admin tools but is a complaint regarding admin privileges. Any comments welcome. Polargeo (talk) 22:07, 30 April 2010 (UTC)

I'd assumed you hadn't posted it to the appropriate area because this was some sort of elaborate joke and that you couldn't possibly be serious in your request. TheGoodLocust (talk) 22:20, 30 April 2010 (UTC)