Tuesday, January 13, 2009

RFA Success

I was reading a blog post by Majorly about how broken the RFA process at Wikipedia has become. It seems to be a common topic of blogosphere ramblings, so I'll toss my own thoughts into the mix as well.

The success of an RFA system rests squarely with the bureaucrats. This needs to be a small group of people who are absolutely trusted by the community and who absolutely have the good of the project at heart. Anybody else elected to this position who does not have these traits should be removed immediately. Bureaucrats should be helpful, dedicated, inspiring. They should love the project so much that they would be willing to leave it voluntarily if they thought it would improve the project. Being a bureaucrat isn't some kind of trophy to put in the trophy case, or another badge of honor to place on the lapel: it's a promise to put the well-being of the project above any other concerns, above any personal relationships, and above any personal ambition.

That said, bureaucrats need to be both trusted and empowered to make processes like RFA Just Work. The question in an RFA is exactly this: "Is this nominee trusted enough to use the new tools to help the project?" During the process people from the community vote, optionally leaving rationales to support their votes, and then the job of bureaucrat begins. Bureaucrats need to look at each individual vote, and they need to look at the ebb and flow of overall community opinion as a whole to come up with a resolution. Here are some guidelines:
  1. Reasons to promote the candidate are listed in the nomination, and the user has a track record that should be self-explanatory. Any oppose vote needs to have an accompanying rationale. Why do you not agree with the nomination? What part of the nomination specifically do you disagree with? What other information do you have about the candidate that should be known and considered? The more information you give, the more powerful and persuasive your vote becomes. Without any information, an oppose vote really isn't anything.
  2. Votes should deal with the matter at hand: Is the nominee trusted to use the tools for the benefit of the project or not? Any voter whose voting rationale doesn't address this question directly isn't relevant and isn't counted. Saying "Oppose because user has only 423 edits and my algorithm requires 500" or "...because the user is a woman" or "...because the user is only 17", or "...because the user misuses commas sometimes" are all non-votes and really don't matter.
  3. Any votes that are unreasonable or irrational don't get counted. This counts votes from known Friends and Enemies of the nominee who obviously vote they way they do because they publicly like/hate the person in question. If your judgement on the topic is clouded by your personal feelings, you shouldn't vote. The question isn't "do I like this person?" it's "Do I trust this person to use new tools for the good of the project?" There have been several occasions where I promoted RFA candidates who I did not like or agree with personally, but who I knew to be good Wikibookians. Notice that "Good Wikibookian" is not the same as "Agrees with my opinions about Wikibooks".
Bureaucrats need to read through the list of votes, separate the thoughtful ones from the cruft and the craziness, and render a verdict based on the needs of the community and the project, not the whims of individual editors. On Wikibooks, bureaucrats and admins are empowered (although generally discouraged from) making decisions on behalf of the project even when such a decision is against the majority of voting community members. This isn't a frequent occurance, but when the chips are down the person making the decision needs to take the available evidence (in the form of votes and their rationales) into consideration. There have been a number of VFD discussions that come to mind where the majority of people voted to "keep" a book but an admin deleted it anyway because the "keep" votes didn't adequately address any of the policy violations that the book represented. Are these tough decisions to make? Of course they are. Do admins take some heat for this when it happens? Yes, they always do.

At the end of any discussion, the question has to be "which of these two alternatives will be best for the project as a whole?". Any decision-maker who doesn't take this question into account, or who willingly answers it incorrectly should be removed from their position immediately. Because it really doesn't matter what I want, or what you want, it's what the project needs that's important. Hopefully, all your bureaucrats know that.


  1. This does sum up the idealized role of the Bureaucrat nicely - thanks for writing this. I'd simply offer that many decisions are not "Which of these two alternatives will be best for the project as a whole?" but rather "What course of action will be best for the project?" There is often room for compromise; false dichotomy is one reason we eschew voting.

  2. You're right Mike, false dichotomy is a very large problem and I shouldn't have made it sound like every decision has a true or false solution to it in the end. However, this is really what many practical administrative decisions boil down to: To delete or not to delete? To promote or not to promote? To block or not to block? To protect or not to protect?

    This is, of course, a level of abstraction. Choosing not to delete a page often requires deciding how to keep it (rename, cleanup, transwiki, etc). The administrator's role in that is to push the "Delete" button or not to push it: The community as a whole will work out any of the extra details.

    These true/false decisions are left to our administrators and bureaucrats. Joe Schmoe community member cannot force an administrator to click the button, or prevent him from doing so. The use of an admin's tools are left to the discretion of that admin. What's necessary is that we trust those people to use those tools wisely for the good of the project.

  3. For what it's worth, that's pretty much how I try to approach my duties as a bureaucrat on the English Wikipedia, and I consider myself to be a pretty damn good one at times. :)

  4. Thanks for the comment, EVula! People outside Wikipedia often only hear the bad stuff about what goes on there, so it's good to hear that some of the bureaucrats there aren't the mindless perpetuators of RFA-based mob lynchings that they are made out to be.

    Evula, if I may ask: Do you do any weighting of votes on RFA (giving more consideration to votes that are more meaningful), or do you ever ignore votes completely if they're worthless?

  5. Depends on how close it is. An RfA at 60%, unless I see some serious discussion about the merits of the opposing arguments, gets closed by the numbers; similarly, an RfA in the high 80s doesn't require much more than a glance to make sure that there are a hundred sock puppets stuffing the ballot. :)

    When it comes to close RfAs, I do have a tendency to weigh the !votes differently. Two examples:

    * [[Wikipedia:Requests for adminship/Tadakuni]] - I closed this one as successful, though it was at the low end of the spectrum. The deciding factor? The fact that even the opposers were very positive about the candidate. If RfA is a trust-based system (a belief I share) and even opponents are praising the candidate, that goes a long way. Similarly, supporters who repeatedly comment that they have grave concerns about the candidate can sway the RfA. (those are basically "weak oppose" and "weak support", though I tend to ignore the bold phrase)

    * [[Wikipedia:Requests for adminship/Enigmaman 2]] - I closed this one as unsuccessful, which by-the-numbers may appear obvious, but there was some serious drama surrounding it. As a result, both the supporting and opposing arguments had to be taken into consideration (the lengthy closing statement goes into detail about how I weighed them both).

    Now, all that said, I prefer it when bureaucrats have a light touch; they shouldn't be interjecting their own opinions into the process, but instead responding and reacting to the opinions of others. We didn't get elected to promote sysops, we got elected to gauge whether the community wants someone to be a sysop. (if that makes sense)

  6. If only it were a perfect world...

    There are plenty of editors who are administrators that shouldn't be. There are plenty of people that have been rejected multiple times that should be administrators. Over the last couple years I have come to realize that the more qualified you are to be an administrator, the harder it is for you to actually become one.

    The administrator crown, such as it is, is no gained by proof of competence; it is gained by proof of political malleability. All of the things people bicker over in RfA is irrelevant... self-noms, edit counts, too many nominators, too few nominators, too much huggle, too little article building, too much bureaucracy, too little projectspace editing... in the end it all is secondary. It all takes a back seat to the most important qualifier for adminship - learning how to not piss people off.