Wikimedia Foundation elections/Board elections/2009/Post mortem

OK, it's time to start thinking about what went well, and what went poorly. Let's kick off the post mortem below! --Philippe 22:08, 9 August 2009 (UTC)[reply]


  • It's my opinion that voters who have an indefinite block in more than one project should not be entitled to suffrage. Also, we need to amend the language about who is entitled to include that they must have an ACTIVE project that meets the voting requirements. --Philippe 22:26, 9 August 2009 (UTC)[reply]
    • Just make sure that we don't exclude people who have only been banned on one wiki and have made a large number of contributions to the Wikimedia Foundation sites as a whole (ie. Meta or other language/sister projects). Cbrown1023 talk 23:39, 9 August 2009 (UTC)[reply]
    • It would give too much power into the hands of the administrators of the individual projects, when we exclude people as soon as they are blocked on a few projects. I rather recommend that the edits on those projects, where are a given user is blocked, are not counted when we determine whether this user has the required number of edits. HannibalForever 00:46, 10 August 2009 (UTC)[reply]
    • Indef blocks are sometimes used to block someone until he explains himself/changes behavior (such as blocks of out of control bots) so you should recheck eligibility every time the users visits the voting page, otherwise a trivial block given at a wrong time can bar him from voting. --Tgr 13:58, 27 August 2009 (UTC)[reply]
  • The suffrage guidelines should be reconsidered. Something like "30 contributions in the past 6 months, 300 contributions in all, first contribution over 12 months ago" where a project edit, mailing list post, code submission, or bugzilla post/comment are all considered contributions should be considered in contrast to the current policy, and reasons given for choosing more or less stringent requirements. -- sj | translate | + 03:08, 11 August 2009 (UTC)[reply]
  • There should be out-of-band ways to register to vote for good contributors who somehow don't meet the automatic criteria. Foundation staff should be eligible to apply this way, but should not get automatic byes -- having foundation staff specially privileged to vote for community seats on the foundation board seems wrong. -- sj | translate | + 03:08, 11 August 2009 (UTC)[reply]
  • Aside from any other considerations, I would change the "active contribs" requirement to be lower, maybe 20 edits in the last six months. This year, and last year, many long-time, very invested community members who have a definite interest in the future of the foundation were rendered ineligible because they haven't been editing much lately. -- phoebe 05:27, 12 August 2009 (UTC)[reply]
  • Indefinite blocks that began more than three (five?) years prior should not count. There ought to be some provision for disregarding blocks from problem-ridden projects or problematic admins. (No idea how to define those notions.) Blocks that were appealed and remain undecided should be disregarded at least under some reasonable (weasel word) circumstances. Users can usually circumvent all of this easily so stringent rules simply encourage bad actors to have a stable of sock puppets. Confirmed-identity users could have least stringent rules. DCDuring 16:20, 12 August 2009 (UTC)[reply]


  • It would be great to link directly a specific language on the translation table (ie. Board elections/2009/Translation#el for Greek) and to make the table sortable. Cbrown1023 talk 23:39, 9 August 2009 (UTC)[reply]
  • Re-use translations from previous years! Let's make sure we remember to do this, we did a pretty good job this year. :-) Cbrown1023 talk 23:39, 9 August 2009 (UTC)[reply]
  • Please, please, please make sure we are not given last-minute texts the day before the vote to translate. Cbrown1023 talk 23:39, 9 August 2009 (UTC)[reply]
  • Maybe we should split the SecurePoll extension into two parts (the admin interface and the user interface) so that the translators don't have to translate the admin interface unless they want to? (translators-l post) Cbrown1023 talk 23:39, 9 August 2009 (UTC)[reply]
  • Set-up and follow some sort update schedule for the spiwiki vote interface and jump text, we can't assume that the page won't change (new translations added, old translation tweaked) and shouldn't have to keep bothering everyone for a new dump. It would probably be best if the script just made a dump every day or two of all the applicable pages. Though, it was definitely a Good Thing® that we had Tim's script to pull all the translations and easily upload them. Cbrown1023 talk 02:41, 11 August 2009 (UTC)[reply]
  • I notice that candidate Relly Komaruzaman seems to have had a much longer statement on the Indonesian page than on the English page. Should we have reached out to someone to translate the Indonesian version? Probably not having a good English version contributed to his low showing in the election.--Pharos 19:23, 13 August 2009 (UTC)[reply]
  • Why were there way less translation links for the questions than there are languages in Google Translate (not to mention there are other, often better web translators)?
  • It would be a good idea to have the candidate introductions translated first, they are way more important than committee etc. --Tgr 14:01, 27 August 2009 (UTC)[reply]



Debates and Questions

  • It was very disappointing that a substantial number of candidates participated in an audio Q&A for Wikivoices #45, on Skype, but then the volunteer(s) who organized it failed to edit or even post the raw audio file of the session. What a complete waste of time for the candidates. -- Thekohser 03:18, 10 August 2009 (UTC)[reply]
  • It was disappointing that there were no official debates or conversations on Meta, and only two unofficial sets of questions, both posed and answered in English, on the English Wikipedia! -- sj | translate | + 03:08, 11 August 2009 (UTC)[reply]
  • The questions on Meta were not prominently linked from anywhere -- it was quite difficult to run across them for a casual voter who was moved to read candidate bios. The answers to one question did not link to those to other questions, making it difficult to see a single candidates answers to more than one question at a time. And the summary page listing the full set of questions posed had to be updated by hand, sometimes with mistakes. All of this could be automated somewhat, so that the act of posting a question updated the TOC as well. -- sj | translate | +
    • Just a comment about advertising the questions page: the questions page was linked in the navigation, so it showed up on every single elections page. Cbrown1023 talk 03:28, 11 August 2009 (UTC)[reply]
    • I was going to make the same comment Cbrown made... I agree that the inability to see a single candidate's responses together was an annoying limitation. It would also be nice if you could select two or three candidates and see their answers side by side: Forcing the electorate to wade through responses from people they had already excluded from consideration is disrespectful. Q1 is around 44kwords of text by itself: The size of a novel. Given around 3.1k voters... if they all read at 300wpm and read the whole thing the reading of Q1 would have expended something like 3.8 man years of effort in aggregate. We should be mindful of this non-trivial resource cost and try to make it as easy as possible for voters to find the information most relevant to their decision, I'd hope that such efforts would also result in more informed voters that can make better decisions as well as greater turnout. --Gmaxwell 03:41, 11 August 2009 (UTC)[reply]

Candidate statements and voting pages

  • Accepting candidates until the day before the vote started wasn't good at all, though it probably couldn't have been avoided it sucked. (no stable pages, quite a few of the translations were out-of date, had no time to finish the translations before voting started). Cbrown1023 talk 23:39, 9 August 2009 (UTC)[reply]
    • There would have been seven days gap if it wasn't for the Centralnotice thing not working due to technical issues which the Committee had no control over whatsoever. Daniel (talk) 00:49, 10 August 2009 (UTC)[reply]
    • Even so, mandating at least a minimum period (2 days?) for final translation, checking, and cleaning of relevant pages is appropriate. If there is some crisis that pushes new candidacies up to the previous deadline for starting the vote, the vote start date should be pushed back rather than putting non-English-speaking voters at a confused disadvantage. Also, see below: the declaration period for candidacies should be separated from the statement period. -- sj | translate | + 03:08, 11 August 2009 (UTC)[reply]
  • The committee should go through and make sure that all the candidates meet the requirements, we had a candidate or two who were translated into quite a few languages and then removed... wasted work. Cbrown1023 talk 23:39, 9 August 2009 (UTC)[reply]
  • I rather like the idea of a "proposed candidate" page and a "confirmed candidate" page. No translations until they're confirmed... --Philippe 16:34, 10 August 2009 (UTC)[reply]
  • I agree - we should require that candidates list themselves as potentials, without filling out anything but their bare statistical details (no statements) a week before statements can be posted. Then they can be vetted, moved into 'confirmed', and can have their statistical bio (and formatting such as the shuffler) added to all translations while the statements come out. -- sj | translate | + 03:08, 11 August 2009 (UTC)[reply]
  • If the shuffler is going to be added to the candidates page, they need to be added before the candidates are added and the page is being translated. It was added late and not all the translations had them and some of them who did were missing a few div tags, which caused only a few candidates to be shown at once ( = really sucky). Cbrown1023 talk 23:39, 9 August 2009 (UTC)[reply]
  • The voting pages were quite light on actual information about the candidates. There was no way to get from the voting page itself to candidate details. Then again the statements themselves contained no links to the answers by that candidate to questions posed. -- sj | translate | + 03:08, 11 August 2009 (UTC)[reply]
    • By "voting page" do you mean the vote interface where people actually vote? If yes, then there was actually a link to the Candidate presentations at the bottom. (See Board elections/2009/Vote interface/en). Though, I agree that we could probably do something better here... if people just showed up on wiki that day and missed the other few weeks of notices, they would be sent directly to the vote interface with no idea who the candidates were. Cbrown1023 talk 03:28, 11 August 2009 (UTC)[reply]
      • To be clear, I meant you couldn't go to a single candidate's information, say with a link next to each name. You had to scroll to the single link that took you away from the page. Mouseover text would have been reasonable and handy, given how short the texts were. -- sj | translate | + 07:57, 14 August 2009 (UTC)[reply]
  • The voting page was confusing. This method is actually fairly simple : the principle is to define a a spectrum of preferences, where the actual values assigned to individual candidates don't matter. This could be well visualized with a list that can be drag-sorted, with a color gradient added to help indicate direction of preference (e.g., lighter colors towards the top of a list to indicate the highest preference). -- sj | translate | +
    • Yes, Kalan actually suggested something else and I think I told someone about this, but it probably wouldn't have mattered since the interface was already designed. However, for next year something like what Kalan came up with for the ruwp ArbCom elections might be better. ([1]) Cbrown1023 talk 03:28, 11 August 2009 (UTC)[reply]
      • That's a lovely example, yes. I would add the ability to rank individuals below the default "no preference" group, so that if you only feel strongly about 5 candidates, you can rank them above or below the larger group about which you have no pref. I would also add a feature that supports selecting a # of names and dragging them all at once, either together in an "equal rank" group or in a series as "123" group. -- sj | translate | +
    • Drag sort doesn't easily accommodate equal ranking, at least not in any implementation I've seen. Equal ranking is something the overwhelming majority (2647 of 3019) of the 2008 participants made use of... --Gmaxwell 03:41, 11 August 2009 (UTC)[reply]
  • Oh. Thats spiffy. I'd missed the link... Enough room there that you could move the panel to the left and display the candidate statement for the hovered candidate on the right. Shouldn't be hard to integrate that into securevote. --Gmaxwell 16:18, 11 August 2009 (UTC)[reply]
That drag n' drop implementation is *awesome*. I wish we'd been able to use it this year. -- phoebe 05:19, 12 August 2009 (UTC)[reply]
  • It is very confusing when you arrive logged out to the voting wiki. At the very least the links to IP etc. should be hidden. --Tgr 14:05, 27 August 2009 (UTC)[reply]
  • The statements are too short to be useful; to get a clear picture, one needs to read questions and candidate talk pages (and being a foundation-l resident doesn't hurt either). This further disenfranchises non-English voters. Maybe some sort of recommendation system should be considered, where voters could see the opinion of more informed members of their main community? This has its obvious dangers, but it might be still better than people voting essentially at random. --Tgr 14:11, 27 August 2009 (UTC)[reply]


  • Not our finest hour this year, I think we as a committee can agree. We didn't do a great job learning from last year. We need to remember to re-use things like mail out scripts, rather than reinventing the wheel. A full committee meeting or two might help as well. --Philippe 16:36, 10 August 2009 (UTC)[reply]
  • I know the ElecCom is probably sick of it by now, but if you all can sit down & quickly jot down your technical notes on what to do, what was problematic, etc. in a short report form, I image the next committee would thank you :) It's something to do when it's all still fresh in your mind, though. -- phoebe 05:22, 12 August 2009 (UTC)[reply]
  • Some kind of "test-run" might be worth-while before the actual election happens. It won't catch all the errors, but it should help catch some. Also, please help review people's work. Even people who are amazingly magical can make mistakes when they're doing a lot of work with a close deadline (this includes creating pages, translating pages, importing/tweaking pages and translations, the election code, etc.). Anyone can check really, it's just good if you make sure that someone does it. Cbrown1023 talk 15:54, 12 August 2009 (UTC)[reply]
  • How advertising this time? I remember a few posts from people saying "I had no idea this was going on". A CentralNotice is really great and is the best option, but then there are people who hide all notices and get annoyed when they don't know... e-mails are also good, but do we need to try to reach out to people in different ways? Village Pump posts? Posts to major mailing lists? Cbrown1023 talk 19:05, 13 August 2009 (UTC)[reply]

Candidacy requirements

  • Instead of making candidacy requirements a simple equivalent to suffrage requirements, I would suggest that each candidate must get the support of a nominating petition of of some fixed number of editors with suffrage (perhaps 10 or a few more) before they can be confirmed as candidates. This would winnow down the candidate pool to only the realistic options, and if we choose this structure it could also allow the participation of non-editor candidates, so long as they have sufficient support among the Wikimedian editor community.--Pharos 17:54, 12 August 2009 (UTC)[reply]
We tried something like that previously and it created a huge confirmation orgy. Everyone would get "10 or a few more", look at the 2008 results: The person who performed worst relative to the winner was still preferred over the winner by 13% of the voters. I've racked my brain trying to come up with some way to cut down the options to make the material more readable— but I've got nothing. :( Even if we could manage some reasonable criteria it would probably only manage to eliminate one candidate or so, which probably wouldn't be enough of an improvement to justify the trouble. --Gmaxwell 18:37, 12 August 2009 (UTC)[reply]
  • It also turns it into a popularity contest, heavily favoring those known on en:wp with its large pool of editors. On the other hand, I agree with Pharos that some sort of enhanced confirmation process could be valuable. I wonder if better explanations of what the board is for (i.e. don't run to make a point about Wikipedia content) would help. -- phoebe 18:20, 13 August 2009 (UTC)[reply]



Explaining the voting system

  • As someone who got very confused by the voting system last year, and was even more confused this year (because it was for three seats, not one), is it possible to have a really well-written explanation of the voting system? For example, this year, I voted for 5 candidates, and didn't express preferences for the other 13 candidates. However, I wanted to express preferences for 5 candidates, abstain on another 5 or so, and cast a 'negative' vote for around three candidates (the phrase "vote off the island" came to mind). However, I couldn't work out how to do this within this voting system. I also fear that those who only voted for one person, or two people, and left the others blank, ended up giving the other 17 candidates their "second choice" vote, or the other 16 candidates their "third choice" vote. In an election for three seats, I think this does actually skew results, unless I'm misunderstanding how it works (quite likely). So there are two questions there: (1) Are abstentions and rejections possible in this system? (2) Is this system suitable for n seats if people cast less than n votes? Carcharoth 00:40, 13 August 2009 (UTC) Please move this comment somewhere else if it fits better in a different section.[reply]
  • Votes are relative, so you have to rank everyone in order to give someone a "negative" vote -- all you can say is you prefer all these other people to the people you ranked last. Unranked votes are always counted last, so in practice, no, there is no good way to abstain on voting for a candidate. Agreed we need a clearer system; the drag & drop widget linked above might help illustrate it. -- phoebe 18:16, 13 August 2009 (UTC)[reply]

The correct ballot for "Prefer ABCDE, Abstain FGHIJ, Oppose KLMNO" is ABCDE=1, FGHIJ=2, KLMNO=3. The whole notion of "giving the rest their second choice...skewing results" is the wrong way to think about how the process works.

Here is how you should think about it: If you had six candidates (ABCDEF) running for some seats you could conduct 30 (6*5) two person elections, one for each pair, i.e. AvsB AvsC AvsD AvsE AvsF BvsC BvsD... etc. In the AvsB election voters could say they prefer A, B, or have no preference. Using this process you could then select the first winner to be the person who won all 5 elections they were a participant in (this is called the condorcet winner), then for the next seat the person who won all elections except the one against the first winner, and so on.

However, conducting 30 two-way votes is a lot of work for the voters and with 18 candidates there would be 306 instead of 30! If we assume that the voter's preference has no loops, i.e. A is better than B who is better than C who is better than A, there is a much easier way than running 30 elections: We can instead run ONE election, and ask the users to simply place the candidates in order based on their preference. A>(C,F)>B>(D,E) for example. That ordering can then be converted to the 30 pair-wise elections, and judged in the same way the pair-wise elections were judged.

So there is no harm in voting (AB) > (CDEF...). What you're saying with that ballot is that you prefer A and B to all the rest and have no preference between A and B, and no preference between any of the rest. Likewise, you can easily create a "Prefer, Don't care, and oppose" ballot.

The "pairwise matrix" given in the results are the scores in these 306 virtual pairwise elections. For example, 992 people preferred Ting to Kat, and 948 people preferred Kat to Ting which leaves 1000 voters who expressed no preference between these two candidates. In this election Ting won these pair-wise votes against all 17 other candidates, so he was the unique condorcet winner. If you remove Ting, Kat would be the condorcet winner.

There are many different condorcet voting methods. These methods differ in how they behave when there *isn't* a single unique condorcet winner. The particular method that we use, the Schulze method, meets a number of desirable voting-theoretic criteria but I don't believe I can provide a simple and intuitive explanation of how it works like I hope I provided for the concept of ranked preferential voting and condorcet selection. Fortunately, an understanding the basics of ranked preferential voting is all you really need to have in order to create a ballot which reflects your honest intent.

There are a couple of improvements I'd recommend for the JS ranking tool. For example, I'd recommend that the candidates start in the middle with the possibility of ranking people below the starting point— so that creating "Prefer, Abstain, Oppose" ballots doesn't take a lot of work. Cheers, --Gmaxwell 20:52, 13 August 2009 (UTC)[reply]

  • Re your last point, I agree, it would make sense to have a drag below/drag above option in a visual ballot like that, thus obviously illustrating the answer to the question that Carcharoth asked. I think it's a pretty common outcome for people to say "well, I like this person, and I don't like this person, and I kinda don't care about the rest", and having an easy way to vote that ticket would be good & would probably encourage participation. -- phoebe 23:43, 13 August 2009 (UTC)[reply]
    • Ah! I was just discussing this with Mako (and above); I agree. -- sj | translate | +