Thursday, August 30, 2007

On Ignoring All Rules

One of Wikipedia's most important, and most embattled, policies is the infamous "Ignore All Rules". It's been there for a very long time, and in that time has changed quite a good deal from the original "If rules make you nervous and depressed, and not desirous of participating in the Wiki, then ignore them and go about your business" (which is the same form it had when I joined up in 2004, although more bits of it were wikilinked then than in the original). Today's "If the rules prevent you from working with others to improve or maintain Wikipedia, ignore them" is different from this in a very subtle and pernicious way: you are only permitted to ignore the rules if you are supported by a clique. Sadly, IAR has been subjected to constant attacks of this sort, almost entirely by people who either fail to understand IAR, or understand it but object to the principle behind it.

I admit that I don't pay a lot of attention to what IAR (or any other policy on Wikipedia) actually says. My experience is that Wikipedia policy documents are edited mainly by "policy wonks": people who seem to have a fetish for generate huge reams of policy without necessarily contemplating whether the policies they write either reflect actual practice or are actually calculated to benefit the encyclopedia. (There are also quite a few people who are constantly manipulating policy for "nefarious" purposes, but I don't want to dwell on that.) These people mean well, but that doesn't mean that their efforts yield positive results. The road to hell, after all, is paved with good intentions.

IAR is especially problematic to policy wonks. By and large, these are people who would prefer that everything be governed by well-defined, clear rules that anyone can understand so there is no excuse for not following them. The problem with IAR is that it gives explicit license to not follow the rules: anathema! Their preference would be to get rid of IAR, but Jimbo won't let them. Since they can't get rid of it, they spend a great deal of energy trying to find a way to write it that narrows its scope as much as possible while not offending the Jimbo-God by actually deleting the policy. The two thousand edits IAR has experienced in the past two years are the collective papercuts of these editors circling about, seeking to find the chink in IAR's armor that allows them to stab it dead, and never understanding that IAR reflects a attitude of the "old core" Wikipedia community that cannot be killed without killing that core.

The real problem here is that IAR is the "Zen koan" of Wikipedia policy: if you understand it, you don't need the words, and if you don't understand it the words (no matter what they are) won't help. Unfortunately, the proportion of Wikipedia's editor base today that understand what Zen is, or more generally understand how to balance opposing tensions, is small and shrinking fast. Many of Wikipedia's policies are in opposition to one another, and it is necessary to strike a balance between two extremes in order to move forward in a reasonable way. (This is not the sum total of IAR, but it is a significant part of it.) Unfortunately, the "follow the rules at all expenses" crowd (many of whom are Aspies, or so it seems) doesn't tolerate balancing tests well: they want hard, immutable, objective, brightline rules, not multifactor tests that evaluate competing interests in what is inevitably a subjective way.

I don't have a solution to this problem. We can't ban all the Aspies, or even exclude them from editing policy. The best we can hope for is that people will make more of an effort to try to teach new editors what things like IAR really mean. But we've been trying to do that for the past two years and it hasn't helped worth a damn.

Monday, August 20, 2007

Meatpuppets, elitism, and structurelessness

Over the past month or so there have been a number of people accused of sockpuppetry, meatpuppetry, or other forms of inappropriate canvassing. That, combined with my recent "apology" regarding the sockpuppetry/meatpuppetry situation involving Poetlister et al, got me to thinking about Wikipedia's whole issue with meatpuppets and sockpuppets. On top of that, a friend reminded me of Jo Freeman's classic essay on the Tyranny of Structurelessness (which should be required reading for all Wikipedians). Rereading that essay brought me a realization: Wikipedia's policies on meatpuppetry and on canvassing, amongst other effects, entrench the current power structure.

Let me be clear about one thing: I am not questioning the good faith of those who enforce the sockpuppetry policies; I'm sure they think that what they're doing is for the good of the project. Nor do I think that they have, as an intention, the goal of perpetuating their own power cliques. But nonetheless, that's the effect.

Here's how it works: In any AfD, RfA, or other such voting-like activity on Wikipedia, if a group of editors not already generally known to be an "elite" appears and all votes together on an issue, and a member of an existing elite notices it, there is a good chance that an allegation of sockpuppetry or meatpuppetry will be made. Also, if editors should make the mistake of taking a position on such a vote in any communication where it can be seen, that editor will be almost certain to be charged with canvassing. In both cases, what is happening is Wikipedia's immune system is reacting to the presence of a foreign elite. The effect of these policies is to marginalize editors who, being not part of an existing elite, seek to form a coalition of their own to increase their influence. Wikipedia's governing philosophy holds that such influence peddling is inappropriate and officially excoriates those who do so, but in practice this is only applied to people who are not already members of recognized elites and who get caught.

"Elite", here, is defined as Freeman defines it: "a small group of people who have power over a larger group of which they are part, usually without direct responsibility to that larger group, and often without their knowledge or consent." Elites are simply "groups of friends who also happen to participate in the same [] activities"; in this case, editing Wikipedia. The existing elites have protected channels of communication (usually, email, IM, or private IRC channels) that allow them to engage in canvassing activity without it being seen by outsiders. Wikipedia policy explicitly encourages, in fact nearly mandates, that such power networks operate covertly.

The reason why this situation has developed and has been permitted to develop is, of course, complex, but fundamentally derives from Wikipedia's misguided love of consensus as a means of governance. Wikipedia's insistence that its votes are not votes and its use of pseudoconsensus (what Wikipedia calls "rough consensus" but which really means "a supermajority of some sort or another depending on how we feel today") as a means of governance leaves it very vulnerable to being manipulated by vote fraud systems. If Wikipedia would either use a true voting system with proper safeguards or, alternatively, actually earnestly discuss issues until a true consensus of involved parties is formed, they could develop governance systems that would be less vulnerable to sockpuppetry and other vote manipulation systems. The failure to do so has made "hunting socks" an increasingly important activity, which has led to ever-increasing use of checkuser, which itself has implications for the privacy-conscious. It's my conclusion that Wikipedia's failure to adopt methods of governance that are inherently sockpuppet-resistant has harmed it and will continue to harm it.

What made me think about this was a comment Greg Maxwell made to me about featured picture voting on Commons. Maxwell proposed to use pairwise voting to, over time, identify the "best" images on Commons. Basically, editors would be presented with two random images and asked to pick one or the other as the "better image". This method will, over time, identify the best images (assuming editors don't lie) and is virtually immune to canvassing and other vote manipulation methods because voters do not get to choose the images; they are chosen randomly. One can, of course, reload the page over and over again until one gets the image one wants to vote for, and then vote for it, but there are so many images that it would take a very long time to make this work. This is a great example of a sock-proof voting system. Maxwell only mentioned it to me because he had tried to discuss it with another Wikimedian, who simply couldn't see how it was better than the current system, which is apparently riddled with sockpuppetry accusations and canvassing efforts (which Maxwell, as one of Common's checkusers, ends up having to deal with).

This sort of approach won't work for RfA or AfD, but I thought of a way to at least reduce canvassing effects: randomly selected juries. Basically, whenever an "question" arises ("Should $EDITOR be made an administrator?", "Should $ARTICLE be deleted?", etc.) a discussion page is created as currently, and anybody who wants to can post comments on that page. However, there is no voting. Instead, after the discussion has run a reasonable time, the question goes to the jury pool. The system then randomly selects a certain number of editors, who will be asked to evaluate the discussion and vote on the question. An editor who does not vote within a reasonable time will be replaced automatically by another random editor. A sufficient supermajority of the jurors in favor of the question will lead to its being accepted; otherwise, the status quo is retained. The reason why this is much more immune to sockpuppetry and such than the current system is that there is no way for an involved editor to get the opportunity to vote on any given question. It also tends to undermine elites because the selection of jurors is completely random and cannot be influenced by an elite, and if individual juror votes are kept secret, the risk of elite retaliation against jurors is at least minimized.

Of course, this idea will not be implemented: the elites will be the big losers and will exercise their substantial will against it to ensure that it does not come to pass, and I'm sure there are endless objections hiding away in Wikipedia's philosophy. A pity.

Sunday, August 05, 2007

The difficulties of targeted advertising

The current furor in the social web world is over Facebook's advertising system. Apparently, several UK companies have withdrawn their advertising contracts from Facebook because their ads appears on the BNP's community page (requires Facebook login). For those who don't know anything about British politics, the BNP is a far-right political party with strong nationalist, racist, and anti-immigrant attitudes. They are quite marginalized in the UK and enjoy little popularity. Anyway, apparently Facebook's targeting advertising system only allows advertisers to target by geographic region, which means that any advertiser that targets "UK" is going to have some risk of their ads appearing on the BNP page. Of course, no sensible advertiser wishes to be seen as supporting the BNP, but that's how people are going to take a Vodaphone ad showing up on the BNP Facebook page, and so Vodaphone yanked their ads from Facebook until Facebook comes up with a solution.

This should be a cautionary tale for anyone who wants to propose advertising on Wikipedia. Advertising on Wikipedia faces not only this problem, but additional ones, that will make managing an advertising environment on Wikipedia even harder. Relatively few people are going to want to have their advertisement on the page for Hitler, and the ones that do are likely people that Wikipedia won't want advertising anyway. An advertisement for a neo-Nazi group on the Hitler article is not going to look good for Wikipedia. Meanwhile, The Xbox 360 article (the 18th most viewed in July) will be an obvious target both for Microsoft and for Microsoft's competitors, both of which will lead to perception-of-neutrality issues. And advertisers are likely to be unhappy to be seen in conjunction with especially puerile vandalism, which remains a problem despite the best efforts of Wikipedia's obsessive recent change patrollers.

Basically, Wikipedia will have to manage its targeting system both to protect advertisers from being exposed to juxtapositions that the advertisers don't want and to protect Wikipedia from juxtapositions that Wikipedia doesn't want. Managing this will be a lot of work, and in the latter case also cut into revenue, since the juxtapositions that will be eliminated are, in many cases, the ones that advertisers will want most. A sobering lesson to those who seek to monetize Wikipedia.

Thursday, August 02, 2007

Regrets

The recent checkuser finding that Orderinchaos had sockpuppets leading to the block of Zivko85 and Zivko's subsequent explanation of how he came to share IP address with other editors leads me to need to make the following statement regarding a similar case from some time back.

In December 2005, SlimVirgin asked me to run a checkuser on RachelBrown, Poetlister, and a specified IP address from the UK. I remember running the checkuser but don't seem to find my response -- likely it was on her talk page, which has been deleted and purged so many times that it's likely that my response is currently a deleted revision which I, being a lowly non-admin peon, am not permitted to see. (This bothers me somewhat.)

From what I recall, I found evidence of some shared IP use, but the patterns were such that IPs frequently used by one user were infrequently used by another, and vice versa. At first I suspected that this was a case of friends or coworkers occasionally editing at one another's location. At the time I didn't know enough about British IP ranges to recognize IPs that were likely to be public wireless access points, residential, or commercial (and really I still don't). However, I was convinced (pressured, really) by others to set aside my doubts regarding the reliability of the conclusion and so reported the lot of them for sockpuppetry, in what was almost certainly a miscarriage of justice.

I regret this mistake. I was much more naive then and more full of myself than I am now. I apologize to Poetlister and all the other parties unfairly besmirched in this incident. There's nothing I can do at this point to materially alter the outcome or remediate it now, but at least the truth is out.

I hope that those currently entrusted with checkuser and administrative rights will think carefully about the history here before making any final decisions regarding Orderinchaos and his friends.

"Expand" tagging on Wikipedia is a joke

This morning, doing random page browsing, I wandered across Wikipedia's "article" on Lucine Amara. At the moment, exclusive of noncontent templates and external links, it reads: "Lucine Amara is a diva". There are two external links, one to EB (which has broken syntax that invalidates the link, leading EB to throw a 500 error; removing the | fixes this, however) and one to Sony, which throws a "403 Forbidden" error.

This article was created in May 2006 with exactly the text it has now. In the 15 months that have passed since, only four subsequent edits have been made to the article. Three days after it was created, a bot added bullets before the external link to EB (without also fixing the obvious syntax error). In December, an editor added the "expand" tag, and a week later another editor added the link to Sony (which is now dead). In January, a bot replaced the general "expand" tag with a datestamped one -- with the wrong date (while the bot put in January, in reality the article has been tagged since December, not that this really matters).

The only other related change is that in April, another bot declared that this article is "within the scope" of WikiProject Biography and that it is "supported" by WikiProject Musicians. If this is what their support accomplishes, I'd hate to see what happens to articles that they don't support...

This article has existed for 15 months now without any substantive change. It has been tagged for expansion for 8 months, again without any substantive change. Any person with a web browser can go to the EB article and obtain, even with the cut excerpt that EB gives you if you don't pay, her date and place of birth, birth name, singing part (soprano), and notable place of employment (the New York Met). And yet, none of this has happened. Just what is it that WikiProject Biography and WikiProject Musicians do? Why do people bother to put "expand" tags on articles if nobody is actually going to expand them?

Wednesday, August 01, 2007

Wikipedia to Jayjg: Please leave

I previously commented on Andrew Gray's call for the parties involved in this current debacle to take a break. The main feature of the discussion on the mailing list has been Jayjg's persistent attempts to pretend that this request is not directed at him. Andrew followed up in an effort to make it undeniably clear for Jay. Jay's response is to pretend that he has no past involvement. Jay's strategy here is simple: Jay's main evidence of involvement is to be found in the oversight logs, which are secret. Very few people have access to them, and the people who do are under orders not to reveal them. Very clever, Jay.

On top of that, Jay forwarded a truly odious email from ArmedBlowfish that seeks to compare SlimVirgin to a rape victim. What happened to SlimVirgin is she was sloppy and made edits she regretted, and asked her friend and buddy to make them go away. Her own carelessness led to her being outed. This isn't a raping. At best, it's comparable to having your car stolen because you left the keys in it. For Jay to endorse ArmedBlowfish's blatant troll by forwarding it to the mailing list (ArmedBlowfish is apparently moderated on wikien-l, probably for good reason) is simply odious. It is hard for me to express how angry I feel about this, especially as a rape victim myself.

Jay, I now explicitly join Andrew's call: Depart, I say, and let us have done with you. In the name of God, go!