Jump to content

Wikimedia Foundation elections 2013/Post mortem

From Meta, a Wikimedia project coordination wiki
Info The election ended 22 June 2013. No more votes will be accepted.

The results were announced on 24 June 2013.

Help translate the election.

This is the post mortem to the 2013 Wikimedia Foundation elections for the Board, FDC and FDC Ombudperson. Input is very much appreciated, and can be posted on the talk page or directly to this page. Sign with four tildes as usual.

Decisions on voting processes and candidacy rules[edit]

  • A few housekeeping notes/suggestions:

    (1) Possibly specify for candidate statements that the limit (of 1200 characters) includes spaces and is counted from display mode, not edit mode.

    (2) I suggest beefing up the deadline instructions for end of noms and end of voting, recommending that candidates and voters not leave it until the last minute, given the possibility of computer crashes and internet outages/delays—and stating that the deadline is hard and fast.

    (3) Work out ways of succinctly expressing the voting system, with brief examples, translation-friendly but not for that reason watered-down.

    (4) Organise earlier and more thorough translations of election text.

Tony (talk) 06:27, 18 May 2013 (UTC)[reply]

Based on this year's experience:

  • Incumbency exemptions for all positions up for election (i.e., no edit count requirements)
  • Candidate instructions must be absolutely clear about the "identify to WMF" process
  • Make sure there is a decision before the close of candidacies about what to do with candidates who file after the clearly posted deadline. While there seems to be a recurrent torrent of self-nominations in the closing hours of the candidacy period, it is still a bit of a stretch that someone only decided to put forward a candidacy in those last few hours. Risker (talk) 03:42, 19 May 2013 (UTC)[reply]

Limits on character count (spaces and wiki markup)[edit]

Time line, time limits and deadlines[edit]

  • It was suggested that the time line be followed rigidly to be fair to all candidates. Meaning all deadlines for statements. ID'ing to the foundation, etc. must be met for every candidate to qualify. This would infer that issues such as those listed on this page would need to be decided in advance so they don't hold up the candidates from meeting the deadlines. 03:46, 20 May 2013 (UTC)[reply]
  • Another suggestion was to encourage candidates to file early to give enough time for typical processing (i.e. web site outages, lost internet connections, etc.). This would also allow time for the translators to properly translate staements, replies to questions, etc. in to other languages before the election. 03:46, 20 May 2013 (UTC)[reply]


Election format[edit]

  • Speaking as someone organizing a three-role election, this is considerably more complex than a single-role election. I would personally recommend that the FDC/FDC ombud elections be split off from the Board of Trustee elections. It must have been a difficult decision for some candidates to decide what role to seek, but the notion of allowing an individual to post candidacies for multiple roles is a recipe for voter confusion and overload. Risker (talk) 02:46, 19 May 2013 (UTC)[reply]
  • The next election committee should be appointed at least a year before the next Board elections, so that they can carry out a good review of voting systems (3-4 months), determine which is most satisfactory to voters (e.g., something that allows ranking but also opposition of candidates)...and then have some dedicated WMF resources to write and thoroughly test software (2-3 months) that will allow the selected voting method to be in place for the next Board election. Risker (talk) 02:46, 19 May 2013 (UTC)[reply]
  • Prepare early Per comments from the election committee, preparations should be made early for the following;
    • Verify that the list of eligible voters is complete and that they meet the required criteria.
    • Notify the technical staff mw:User:Tim Starling and maybe m:User:Eloquence? of the upcoming election and schedule time to set up the special:securepoll so that it is functioning properly and also tested before the scheduled voting period.
    • Notify the Translation committee of the election and get help translating the voter interfaces.
These will help avoid possible delays in voting. 02:59, 1 June 2013 (UTC)[reply]
Just to note, the bugzilla asking for the voter list to be run was filed on May 6 after the criteria were provided to Engineering, and the bugzilla for creating the test poll was filed on May 10. Risker (talk) 02:37, 5 June 2013 (UTC)[reply]

Unsorted items[edit]

  • Use the post mortem from previous years.
  • Remember to update default SecurePoll messages on the voting wiki so that they're not misleading.
  • Include links (or at least one link in the sidebar) from the voting wiki to Meta.
  • Use Special:MyLanguage for all links.
  • Ensure candidate presentations are linked from the voting instructions and not only among a bulk of unrelated links.
  • Don't ignore electoral systems experts and mathematical theorems.

SecurePoll software[edit]

This one might be trickier than it looks. The Election Committee shouldn't have the role of deciding which of a user's eligible accounts does the voting (except by eliminating bots), and there is no consistent way across hundreds of wikis to identify alternate accounts. Risker (talk) 03:33, 16 June 2013 (UTC)[reply]
Agreed. It's a very difficult task. For reference, wikitech:SecurePoll#Voter eligibility describes how to generate the voter list. 04:45, 16 June 2013 (UTC)[reply]
Yes, except I understand that's quite a bit out of date. It certainly didn't take 3 days to run the lists this time, several steps that were taken aren't included, and other steps noted there are no longer done. The SecurePoll extension needs a major rewrite and redocumentation; it's not keeping pace with all of the excellent improvements that have been made in MediaWiki since it was initially developed. I've already proposed that it be added to the 2013-14 roadmap for the Engineering department. Risker (talk) 04:57, 16 June 2013 (UTC)[reply]
Agree. Making seperate comment. 19:59, 16 June 2013 (UTC)[reply]
  • The SecurePoll documentation at mw:Extension:SecurePoll and the "user manual" at wikitech:SecurePoll both need to be updated. Prefereably before the next election. 20:07, 16 June 2013 (UTC)[reply]
  • SecurePoll training We should try to get multiple people trained on using the SecurePoll software so that we are not reliant on a single person running the software during elections. 20:21, 16 June 2013 (UTC)[reply]
  • Something to consider is whether or not we as a community feel that we *must* have our own software to run votes, or whether we should consider making use of an external organization to do a lot of the election work. Aside from English Wikipedia Arbitration Committee elections, the only other times this software has been used is (a) for Board/WMF elections and (b) for cross-project referenda, specifically the license change and the personal image filter referendum. It may be difficult to justify the ongoing cost of developing and maintaining this software and then ensuring that (a) Board and Board-related elections/uses are considered a #1 priority that supercedes other work and (b) there are sufficient staff educated and competent in its use. If additional voting methods are considered, there will also be the need for writing the programs that enable those voting methods. This is also a security extension so needs to remain in-house. Risker (talk) 20:25, 25 June 2013 (UTC)[reply]

Voter participation[edit]

1809 contributors determined the results in this poll.

How does this compare with historic participation levels? does the participation trend match contributor activity trends? - Amgine/meta wikt wnews blog wmf-blog goog news 21:35, 24 June 2013 (UTC)[reply]

There were 3368 valid votes in the 2011 elections. This represents a 46% reduction in participation since the last election. It is difficult to identify reasons why people choose not to do certain things. We do know, however, that 534 people got as far as authenticating to vote and received a ballot but then did not click "submit". We do not know if this is new or if this is something that has been happening for years, because the last two elections were hosted on SPI servers instead of WMF servers and we do not and did not have access to that data for previous elections. We had several comments about the "difficulty" of reaching the voting page, and I do not know if there is a great solution for it (sending people directly to the ballot without introducing them to the candidates is just as suboptimal as forcing them to stop on Meta); on the other hand, this situation where so many people got as far as an authenticated ballot but did not vote suggests there is something else going on here. That 534 non-votes is 30% of the number of actual votes, and that is a statement by itself. Risker (talk) 22:33, 24 June 2013 (UTC)[reply]
Yeah, I think it's relatively clear that the difficulty in voting made less people vote but it shouldn't be a big contributor to less then previous elections because previous elections had the same issue (going back to your home wiki/active wiki) and actually often did not have as clear directions on that. The 534 non-voters (who almost certainly got a ballot, but did not submit it) is very concerning to me though. If we continue to use SecurePoll then the upcoming SUL unification will make the process much easier (we will be able to just give them a link to vote on, they will not have to go back to their main account) however the non-voters are tougher. My gut says it was because of the long list of questions and candidates (testing has shown pretty clearly in many other places that the more options you give the more likely people are to freeze and not finish) but I don't know that for sure and I'm not totally sure a great way to fix it. We though about two specific 'elections' you had to vote in but trying to send someone BACK to their home wiki to start the voting process AGAIN seems like it is even more likely to destroy voting. Jalexander (talk) 23:16, 24 June 2013 (UTC)[reply]
Well, we actually developed it that way and then had a lot of problems with the two SecurePolls; we were later told that it's not possible to run two system-wide SecurePolls at the same time. (English Wikipedia successfully ran two separate SecurePolls at the same time in May 2010, but on a single project.) This again goes back to the software rewrite that SecurePoll needs if it is to continue to be our go-to election software. Risker (talk)
Ah, that's true, the whole cookie issue. Jalexander (talk) 00:38, 28 June 2013 (UTC)[reply]

Speed with which results were released[edit]

I have had some users make personal inquiries of me wondering how come we were able to release the results so much sooner than in previous years. There are several factors that probably interacted in this:

  • This year the committee monitored the votes throughout the election to identify any that required additional verification (flags for duplicate cookies, more than one user on an IP address, very similar usernames, etc). As issues were identified, they were addressed. Thus, at the close of the polls, there were only a few votes that needed to be reviewed. We also had the assistance of some software that helped to "pull out" the votes that needed additional verification; it had first been designed for the personal image filter referendum, where there were 25,000 votes that needed to be reviewed.
  • There were significantly fewer votes.
  • Two experienced checkusers were reviewing the votes for problems.
  • The voting method was simpler and the results required less review by the Committee (support/neutral/oppose versus Schulze method)
  • The election was hosted on a WMF server instead of a third-party server, which meant faster response times in all vote-related processes including the dump and tally.

There are probably a few other less relevant factors (including the Committee being determined to finish the process), but these are likely the key ones. Risker (talk) 23:03, 24 June 2013 (UTC)[reply]

Were all the voters checkusered? Seems overkill to me, and a violation of the privacy policy perhaps, if so. PiRSquared17 (talk) 23:27, 24 June 2013 (UTC)[reply]
None of the voters were checkusered. The election administrator interface provides IP and useragent information. Checkusers, however, are experienced at quickly identifying overlaps and issues in that data, and conversely eliminating non-concerning information. Risker (talk) 23:31, 24 June 2013 (UTC)[reply]

Neutral votes[edit]

Neutral votes didn't 'count' in the selection process. Is this a weakness? A candiate with a profile of 20 support, 200 neutral, 2 oppose would do better than one with 200 support, 20 neutral, 21 oppose. Point is that a very low-profile candidate ignored by most voters only needs to 'recruit' a few supporters to get elected. Any thoughts on this The Yowser (talk) 16:17, 25 June 2013 (UTC)[reply]

There was an additional criterion, which is discussed on one of the talk pages, but may not be in the description of the election. Candidates for Board of Trustees seats were eliminated if they did not receive the support of 10% of the voters (or, in this case, a minimum of 181 votes). Candidates for the FDC seats and the FDC ombud seat were eliminated if they did not receive the support of 30 voters. The criteria for the FDC/FDC ombud roles was deliberately much lower, because this is the first election for these positions, and we could not gauge whether or not the community would actively participate in their election or if they would essentially ignore that segment of the election. Now that we have had a successful first election, the level of support for FDC/FDC ombud candidates can safely be raised. Risker (talk) 18:03, 25 June 2013 (UTC)[reply]
"discussed on one of the talk pages, but may not be in the description" = it's not a rule. --Nemo 21:31, 27 June 2013 (UTC)[reply]
No, it was a rule[1]. It just was not well advertised or incorporated into the introductory pages. It was also a moot point, as every candidate met this minimal threshold. Risker (talk) 14:49, 1 July 2013 (UTC)[reply]

Count of neutral votes[edit]

  • Only one candidate was able to get more pro- than neutral-votes

Looks for me as if too many voters were trying unsuccessfully to find their favourite candidates ... and it's uncertain, how many canceled their voting during the process to "find a favourite candidate". (at least one I've to confess ;-)

May be trying to get supporting people in relating (or perhaps in most favoured) national wikipedias would be helpful for future candidates to get a wider support of the community - and perhaps also support in translating issues.

Jaybear (talk) 14:22, 26 June 2013 (UTC)[reply]

You raise an interesting point, but looking back on the 2011 results the most strongly supported candidate (i.e., the one who received the most rankings above other candidates) still only received support of about 40% of the electorate; of course, the two systems aren't statistically comparable. Even in elections where the candidates are much better known by the electorate as a whole, there is always a high level of neutral voting; this is seen each year for English Wikipedia Arbitration Committee elections. Indeed, we anticipated the possibility of an even higher level of neutral 'voting' for FDC positions, as this is a new group that has had little interaction with the community as a whole, and whose work has almost no perceived impact on the projects themselves. I for one would have liked to see more individual Wikimedia projects discussing the election and encouraging participation; however, I'd be concerned that bloc voting for certain "nationally favoured" candidates is promoting the agenda of one small group over the benefit of the community as a whole. Risker (talk) 15:04, 26 June 2013 (UTC)[reply]

Finding SecurePoll[edit]

Maybe it was just me (so more opinions appreciated), but I had a hard time finding the actual voting software from the banner. I clicked the banner, and then was lead to a page on meta. The page wasn't straight forward to me how to get to the vote. I would have expected a big button on top CLICK HERE TO VOTE, but I believe the only thing I could find was an instruction to go back to my home wiki, and look up Special:SecurePoll there. I am quite well versed in how Wiki works, so I was able to figure it out eventually, but I guess this might not be the case for everyone. Or did I misunderstand the instructions? Were there more people with this or a similar problem? Effeietsanders (talk) 21:16, 25 June 2013 (UTC)[reply]

PS: once I got to the voting software, it was all quite easy - much easier compared with past years! I was already fearing the ranking of the candidates, then it turned out I didn't have to. Effeietsanders (talk) 21:17, 25 June 2013 (UTC)[reply]


Please, use Schulze method or something similar next time. For me as one of the voters, it's much more simple and friendly. Sorting candidates according to my preference is a straightforward and easy task, it's actually a real joy compared to the strategic voting required by the Support/Oppose system, which is time consuming toil with a chancy result. Honestly, from the beginning of the election, I am kind of haunted by the only "important" question: Where can I vote against the people who decided to change the voting system to the unfriendly Support/Oppose system? Of course, I understand that selecting the right voting system was just one of many duties and there may be some voices of unhappiness with the Schulze method, but I am really disappointed by the only arguments I've found. --Tchoř (talk) 23:54, 25 June 2013 (UTC)[reply]

The Schulze system used in prior elections was the one that is intended to give only one winner in an election, but was then used to produce multiple winners contrary to what the system was designed to do. It was also quite easy to strategically vote using Schulze, no more difficult than using support/neutral/oppose (although perhaps with less effect). There are few preferential voting systems that are specifically designed to produce multiple winners, most of which are designed to work with slates of candidates (expressly forbidden for the Foundation seats). Some form of cumulative voting might be worth examining, although I understand it was dismissed in the past as being too prone to hijacking by factions. The Schulze method has fallen out of favour by many of the organizations that have used it in the past (the English Wikipedia article on the topic requires a big cleanup and is severely out of date) and many have moved on to other systems. It is unfortunate that you felt the need to manipulate your votes; supporting candidates you thought would do a good job and opposing only those you thought would not, remaining neutral on those for whom you formulated neither opinion, is the optimal method for voting using the support/neutral/oppose system.

In any case, I understand the desire to rank candidates or otherwise weight one's votes rather than just either support or oppose candidates (or say nothing about them). We certainly didn't have time this year to give alternate voting systems a thorough examination, given that we had to write the election rules for two new positions, and that some members were able to give only very limited time to committee tasks. I'd suggest that consideration for a different voting system start some time in late 2013, because it will take lots of work, a very diverse group to consider it, and committed resources from Engineering to develop the software and thoroughly test it before the next election. Risker (talk) 04:55, 26 June 2013 (UTC)[reply]

To balance the discussion a bit: I would like to express my support for the system chosen this year. Sure, there might be downsides - but it is much more userfriendly in my view of the world. Everyone understands support/neutral/oppose immediately as a principle - even if you don't understand all the strategic details. Also, I found it much less time consuming to be for/against someone than to rank them in a detailed way. Especially if they keep mixing up the order of the candidates :) Anyway, just saying that the opinions are probably scattered. Effeietsanders (talk) 07:10, 26 June 2013 (UTC)[reply]
Opinion I just want to say that I want anybody from this group and nobody from this group can be easily expressed in Schulze, it translates to giving the same low number to the people of the first group and the same high number to the people of the second group (Everybody from the first group is better than the members of the second group). You are not forced to have and give detailed ranking and the value of your vote is not significantly diminished by your decision to not use the detailed ranking.
On the other side, S/O does not give me the freedom to express my detailed opinion and I am often quite disappointed when I realize my vote had no real influence on the outcome (and all the time spent on studying the profiles of the candidates was wasted), because the real competition happened between the candidates which I had been forced to put into the same Support / Neutral / Oppose group.--Tchoř (talk) 08:49, 26 June 2013 (UTC)[reply]
Thanks for your further thoughts, Tchoř. I find it interesting that you felt your vote had more influence on the outcome using the Schulze system than the S/N/O system, as that's quite different from the theory for each of these systems; supposedly it is easier to "game" S/N/O than Schulze. On the other hand, I do actually agree with you that a really skilled voter could potentially have more impact by strategically voting using Schulze. Risker (talk) 15:08, 26 June 2013 (UTC)[reply]
I am not sure that 'gaming' is the right word, but S/N/O and similar „simple“ systems are indeed much more vulnerable to the tactical voting. And people often do vote tactically in such systems not because they want to somehow artificially increase the influence of their vote, they just do not want to completely waste it (even pretty staunch supporter of Constitution Party probably decides that it is better to use his vote to support some of the major candidates when it comes to the important election – would you call such tactical voting 'gaming'?). I do not want a system where I can leverage my influence thanks to skillful tactical voting or actually effectively waste my vote when not resorting to tactical voting (= S/N/O and similar, in many situations). I just want a system where the best tactical vote almost always coincides with the honest vote, so there is nearly no space and no motivation for tactical voting neither for me nor for others. (= Schulze and similar).--Tchoř (talk) 12:53, 27 June 2013 (UTC)[reply]

The Schulze system has one awful downside: it's nearly impossible to read the results. That huge table with all comparisons doesn't let the reader know why someone won. The plus/neutral/minus system is much more simple: one learns about each candidate and classifies them in three groups. --NaBUru38 (talk) 19:27, 27 June 2013 (UTC)[reply]

"The Schulze system used in prior elections was the one that is intended to give only one winner in an election, but was then used to produce multiple winners contrary to what the system was designed to do". ROTFLOL. --Nemo 21:32, 27 June 2013 (UTC)[reply]
Oh Nemo. Do you really think that the creator of the Schulze system, who has invested a great deal in promoting his voting system, is really the most neutral person on whether or not the Schulze system is the best system? I'm assuming you understand the concept of a conflict of interest. Risker (talk) 14:38, 1 July 2013 (UTC)[reply]
As Effeietsanders above I'd like to express my support to the voting system used this year. -- MarcoAurelio (talk) 14:58, 1 July 2013 (UTC)[reply]

...or other[edit]

I'd like to suggest Graduated Majority Judgment, a voting system I explain in en:User:Homunq/WP_voting_systems and which was successfully used to help resolve the long-running controversy on abortion article naming on en.wikipedia.

  • It's similar to Schulze in that a voter can give nuanced judgments about different candidates
  • It's also similar to Schulze (and unlike the 3-option approval-with-abstentions system used, or Score voting as used on Yelp etc.) in that a voter who wants their vote to have the strongest possible impact does NOT have to remove all nuance from their vote by exaggerating all ratings to top or bottom. A less-than top rating will still count as full support if it is above the winning median, and an above-bottom rating will still count as full opposition if it is below the winning median. This makes optimal voting easy in most cases (just vote honestly), with only rare need for complex strategic analysis. (Of course, as the Gibbard-Satterthwaite theorem shows, no election process can possibly be strategy-free in all cases; but GMJ is so in most realistic ones.)
  • Unlike Schulze, it does not have the "awful downside" of hard-to-read results. Each candidate gets a single number and the highest number wins.
  • Unlike Schulze, there is never a reason to vote your favorite candidate below top or your least-favorite above bottom.
  • If you are interested, it can be extended to a proportional multiwinner form (email me from my en.wikipedia account for details).

Also, as an active contributor to the electorama wiki and electorama election methods mailing list, I strongly encourage you to reach out to those communities before you run another vote. I'm not going to pretend there's any consensus there on what the best voting method is, but there's a lot of expertise on what pitfalls to avoid, and there are even enough people with programming experience and interest that you might be able to get help with the programming side of things. I am certain that if you had talked to the people there beforehand, you would have been fully aware of the strategic implications of any system design decisions you were making. The system you chose isn't a bad one, but it could and probably would be better. You can contact me using the email function on English wikipedia if you want to discuss this further. Homunq (talk) 23:08, 4 July 2013 (UTC)[reply]

I really hated trying to vote by the Schulze method last time. It is, for me, much easire to rate than to rank the candidates. I'd read one candidate's statement, and say I liked what he had to say, then another, and decide I liked that one even better. I ended up first rating each, and then after I got all the ratings, producing a ranking. There was also a problem with the way the system was described, so I did not know how to handle equal rankings — do you number the ranks 1, 2, 2, 2, 3, or 1, 2, 2, 2, 5. Since it was never stated, and I did not want to invalidate my vote, I forced my ranks to be consecutive, and arbitrarily separated those I'd like to rank equal. -- BRG (talk) 00:38, 5 July 2013 (UTC)[reply]
Actually, Schulze is designed so that 1, 2, 2, 3 and 1, 2, 2, 4 are both the same. But yes, I also prefer rating, and that's one of the reasons I think GMJ is a good idea. Homunq (talk) 04:27, 5 July 2013 (UTC)[reply]

Report to the Board of Trustees and to the community is posted[edit]

I have written a personal report to the Board and to the community about the election, with a few recommendations. It is found at Wikimedia Foundation elections 2013/Post mortem/Report from Risker. Please feel free to comment on the talk page. Risker (talk) 03:32, 30 July 2013 (UTC)[reply]