Wikimedia Foundation elections/2017/Post mortem/Report

Info The election ended 11 June 2017. No more votes will be accepted.
The results were announced on 19 June 2017. Please consider submitting any feedback regarding the 2017 election on the election's post mortem page.

The 2017 Wikimedia Foundation elections included two selection processes – one to fill the three community seats on the Board of Trustees (Board), and another to select five new members for the Funds Dissemination Committee (FDC). In this document, "election cycle" refers to the period containing both elections.

This document summarises feedback delivered to the Support and Safety team at the Wikimedia Foundation – which helps to facilitate these elections – by the elections committee, the candidates from this cycle, and other community stakeholders.

Cycle timeline

The official process for these elections began in April 2017 with the Board's call for candidates and ended with the announcement of the FDC election results on June 19, four days later than anticipated. The results can be reviewed in the Appendix.

Board of Trustees election

  • 6 April: Call for Candidates
  • 20 April: Deadline for candidate submissions and Identification verification
  • 21 April – 1 May: Questions and discussion between candidates and community
  • 1–14 May: Voting
  • 15–19 May: Vote-checking
  • 20 May: Goal for announcement of results

Funds Dissemination Committee election

  • 28 May: Deadline for candidate submissions and Identification verification
  • 29 May – 2 June: Questions and discussion between candidates and community
  • 3–11 June: Voting
  • 12–14 June: Vote-checking
  • 15 June: Goal for announcement of results

Post mortem process

Feedback for this report was sought in various venues, including a dedicated page onwiki, emails to movement mailing list Wikimedia-L, and through emails to those who put forward candidacies for either election.

Feedback

Job description and presentation template

On the post mortem page, community member Michael Maggs points out that the call for candidates set out five key characteristics and experience points that would be ideal for new members to possess. However, the candidate presentations did not reflect these points and instead focused on online versus offline work, which he argued was not enough to judge their calibre as the prospective members of an oversight board for a Foundation as large as this.

It was recommended that the candidates' onwiki work be de-emphasised in favour of the promotion of the skills and qualities that the Board actually needs and is looking for, as specified in the job description itself. However, it should be noted that the Board's criteria for a good trustee and what the community is looking for are two distinctly separate sets. Political theory states that candidates will naturally tend to opt to address voter concerns over those recommended by the Board itself.

Word limits for these, some argued, placed non-native speakers of English at a disadvantage as they needed to replace words in their statements to make them fit into the limits, which is naturally harder to do.

Timing

This election cycle included both the Board elections and the FDC elections in the same year, which meant that either the elections had to be held simultaneously or consecutively. Either way, this would require a longer timeframe for the cycle to effectively run. The elections actually began on 7 April, which was much later than the initial goal of 15 March (which would have allowed for ample time for both elections).

Time was not sufficiently budgeted for some of the asks – which are documented elsewhere in this report – and as such deadlines were missed and the public timeline was sifted on a couple of occasions. In particular, there was not enough time to check FDC votes while sticking to the target date, so the deadline was moved back four days.

The committee appointments themselves should be streamlined to allow for an earlier startdate. The committee was not happy that community discussion was not allowed to happen because of these strained timelines, which resulted in the status quo being maintained even if that had proven ineffective in the past.

Translations, language, and templates

Compared to previous elections, overall translations were done more quickly, were of a higher standard, and covered a wider range of languages. This was at least partly because of the availability of strategy contractors who were able to lend their hand to translations. This is unlikely to be so easy next year, though the availability of translations may have helped to improve voter turnout this year.

Alongside the good things that happened with translations this year, several issues were raised:

  • Because the FDC candidates all submitted candidacies last-minute, there were very few translations of their statements and answers by the time the voting began. This could be mitigated by allowing more time as a buffer to gather translations.
  • While translations are improving, there is still the issue of sending out notifications and messages in English as a fallback. This again needs more translation work and additional buffer time. At least one user requested that the Wikimedia Foundation fund translators to help do the work of publishing notices across non-English projects.
  • Relatedly, the use of translations by default in SecurePoll should have some fallback for those who wish to read the raw, untranslated English statements. This will prevent a "weird" translation from being the only one a voter will see.
  • The requirement for candidates to speak English was criticised. This requirement objectively could bar potential candidates who may be otherwise perfect for the Board, but who could not easily take part in discussions in real-time. The majority of the world does not speak English whatsoever, so requiring that Board and FDC members speak the language is arguably shutting off the expertise from the majority of the world's population.
    • This is not something that the elections committee could decree; this would need to be an active decision from the Board of Trustees itself.

Additionally, templates are becoming more convoluted in the Meta-Wiki ecosystem surrounding these elections, which is something that should be addressed in the downtime before next cycle in 2019.

Candidate pool and conduct

The Board of Trustees election process attracted high-quality candidates, many of whom would probably have slotted in well on the Board. There were fewer candidates this year than in previous years, with nine ultimately standing; that candidate pool is the smallest since 2005 (6) and less than half the size of the pool in 2015 (21).

The order of the elections was praised by some, who were pleased that the Board elections were held prior to the FDC elections. This meant that Board candidates who were not selected by the community were able to apply for a seat on the FDC, something that would not have been possible if this order were reversed or condensed to by simultaneous. This did also mean that candidates were examined twice, and could have led to "double jeopardy" in that candidates were penalised by voters for seeking power a second time.

It appears that overall the candidates in this year's election were more popular than those in 2015 – this year, the lowest support percentage was 58.25%, while eight of 2015's candidates polled lower than this.

Some feedback received in the post mortem period pointed to candidates acting out of process on the talk page of the elections. In particular, it was arguably unfair to other candidates (who may not have seen the Meta-Wiki talk pages) to answer direct questions and undermine fellow candidates. In future cycles, it should be explicitly disallowed, better moderated by the committee, or (if allowed) better publicised to the candidates.

YouTube hustings

This cycle marked the first time that "hustings" – public, group debates between the candidates – were held, bringing all candidates together for a facilitated discussion. The hustings were hosted on Google Hangouts and recorded for later viewing by potential voters.

While there were some issues, the concept was sound, and it allowed for a deeper exploration of the candidates. However, there were some criticisms levelled at this process:

  • Timing was a challenge, and some candidates spoke at greater length than they were theoretically allowed. This might be fixed in later years with more aggressive moderation, assuming it is something the committee is willing to run in future cycles.
  • Other criticisms of the process included that it focused on improvised responses to questions, which had several issues. Firstly, this disadvantaged non-English-speaking candidates who had to spend more time thinking about what they were trying to say and therefore had less speaking time. It also did not accurately reflect the reality of what being on the Board would entail.
  • The questions were not geared towards Board candidates; rather, they felt like questions one might ask a prospective executive director.
  • The hustings were organised at the last minute, which some candidates felt was disrespectful to their time. Additionally, the viewership for the debate was low, which meant its overall usefulness to the voters must be called into question.

Voter turnout

Despite the Board elections lasting one day fewer, voter numbers were roughly the same (47 fewer than in 2015). This may have been down to banners or to the availability of translations. The turnout was almost three times that of 2013, and almost a thousand votes more than the previous second-highest (2007, which attracted 4,170 valid votes).

On some wikis, turnout dramatically improved. For example, on the Polish Wikipedia, the number of voters more than doubled –this is likely to be due to Dariusz's candidacy. Wikimedia staff – able to vote this year – were the likely cause for Meta-Wiki's 131% increase. Full statistics on voter turnout, broken down by wiki and compared to the previous election in 2015, can be reviewed on the stats page.

It may be worth investigating how we can further break down these voter demographics, perhaps by affiliate membership or by account experience level. This might provide the FDC with some insight as to the wants and needs of the people they will ultimately be serving, and provide opportunities to further encourage engagement. This kind of breakdown, however, may not be technically possible or even advisable, and would likely need to remain internal to the Wikimedia Foundation or to those on the election committee who have access to private information.

Questions

This year, questions were collated into a set of ten questions common to each candidate. Some of those providing feedback felt that this process was, in theory, a good one:

  • It cut down massively on the time candidates had to set aside to answer questions.
  • The process cut down on the number of questions that candidates had to answer, in particular it eliminated questions that were pointed or leading.
  • It, in theory, allows for better questions which can more accurately judge the potential quality of new trustees.

Along with this praise, some issues with it were raised during this feedback process:

  • The community was confused by what this process entailed and how it was operated.
  • The process for this was not sufficiently explained to the committee themselves, which led to confusion and strained deadlines. More time should be allocated to this in future, potentially two or three weeks.
  • The timeline for the publication of these questions was tight, and ultimately not stuck to, leading to candidates having to answer questions in a few weekdays. This, in turn, impacted those with day jobs who were potentially unable to get to the questions in time. It was recommended also that these questions be published before, not after, a weekend.
    • There was little time allotted to questions and answers from the candidates to the FDC. Only five days were allowed for this process in this cycle. This was done deliberately, as historically the "default" questions are the only ones asked.
  • The fact that questions were not included in the final set of curated questions led to some feeling disenfranchised, and some argued that this setup did not allow for meaningful examination of the candidates. There is the valid argument that pointed questions are more "provocative" and can help to reveal how candidates react to this kind of questioning.
  • While not expressly prohibited, the committee may want to make a call on the issue of questions directed to individual candidates. Some were asked and answered, but – especially when questions are collated – individual questions and answers should be expressly allowed or disallowed.

Communication

There were several issues with both wider communication and communication with the candidates themselves:

  • There was no public discussion of the timescales or any consultation on the election process; as a comparison, the affiliate-selected Board seat selection process is usually publicised months beforehand, and a similar process would be welcomed for these seats. Additionally, a greater volume of announcements was requested by some, or at least in greater venues.
  • Communication with candidates was lacking for both elections. There was no outreach from the elections committee during the election cycle, even to confirm details (outside of the YouTube hustings). Some basic outreach would be appreciated, potentially including an email before the results a few hours before they are public.
    • (It should be noted that the results were published, in both cases, immediately after they were signed off on; we were not holding onto final results for long.)
  • There remains some confusion about the use of the term "election" to describe this process. Since the "winners" are actually just put forward to the Board for their approval, it is recommended they are called something different to distinguish this. Consistency in communication here is also advisable going forward.
  • A positive from this cycle was the use of new banners, which seemed to be well-received on the whole and may have been a contributing factor to the high turnout.

Voting system and software

A support-neutral-oppose system was used for this election, which has been common in these elections since 2011. However, it may be worth investigating making use of a new system for the next cycle, since there are several flaws with the current system:

  • It encourages "tactical voting" since an oppose vote is far more powerful than a support vote. This means that, in theory, a candidate with 999 supports and one oppose vote (99.9%) would still lose to a candidate with 200 support votes and no oppose votes (100%).
  • This weighting of votes in opposition also means that reelection can be especially difficult, which in turn leads to unstable, fluid boards. (This is not necessarily a bad thing and did not impact the result this year.)
  • The current system renders "neutral" votes essentially useless, which is either a good thing or a bad thing depending on what the neutral votes are intended to signify.

If the community discussion (or committee discussion, whichever is used to determine the new system) results in consensus for a new, uncoded system, this should be developed and tested well before the next cycle. This would mean the discussion would need to begin at some point in this calendar year if it is to be implemented ahead of the next cycle.

There was praise for the software used to run these elections. At least one person argues that SecurePoll should be more widely used for elections across the Wikimedia movement, both for online and offline positions.

Post mortem

The very nature of the post mortem is unstructured and free-flowing, and it is not clear how or where feedback should be given. In the future, it may be worth the committee asking some direct questions of voters to better organise this process.

Points of information for forthcoming cycles

People often submit their nominations near the deadline; that leaves not enough time to properly translate their presentations. We should plan the timelines for the elections with some time allocated for translations (a week between the submission period ends and Q&A starts, between Q&A ends and voting begins).

The template for the submission should be aligned with a letter from the Board (if there is one), because the Board identifies the needs and important points should be highlighted somehow. Other templates also need to be cleaned up and better-maintained in the downtime before the next cycle.

Elections will, as of the next election cycle – anticipated to be during 2019 – be the responsibility of the Support task force within the Support and Safety team at the Wikimedia Foundation. This means that in charge of facilitating these elections will be Joe Sutherland, as opposed to James Alexander. To that end, staff is currently in the process of producing documentation on the social and technical process of running an election, which should allow for an easier transition should the lead be changed again.

The issue of an unopposed candidacy came up in the feedback here. This was the case in this cycle's FDC Ombudsperson election, where only one person stood. The committee should consider what ought to happen in the future in the likely event this happens again; should there be a defined "minimum percentage", or should the result just be called before voting begins?

The committee should also discuss updating the requirements around voters who meet the developer criteria, but not the regular voter criteria. These were very outdated at the start of this cycle and had to be revised at short-notice.

The voting system is also something that should be looked at, as it is a matter of importance for several voters and other community stakeholders who consider the existing system outdated and open to abuse.

Overall, the key takeaway from the feedback is to be as mindful of candidates' time as possible. Systems need to be in place that adequately and critically judge a candidate's skills while appropriately respecting their time and effort. Being willing to stand as a trustee candidate, in and of itself, is a big commitment, and this should be reflected in the process they must go through to become one.

Lastly, the committee will need to select a new community advisor, as the existing advisor (User:Risker) is stepping down.

Appendix: Candidates and results

The full results can be accessed and verified on Meta-Wiki.

Wikimedia Foundation Board of Trustees election results

Candidate Support Neutral Oppose Support Percentage S/(S+O)
María Sefidari – User:Raystorm 2,557 1,997 566 81.88%
Dariusz Jemielniak – User:Pundit 2,332 2,255 533 81.40%
James Heilman – User:Doc James 2,435 2,105 580 80.76%
Chris Keating – User:The Land 2,002 2,530 588 77.30%
Peter Gallert – User:Pgallert 1,810 2,726 584 75.61%
Yuri Astrakhan – User:Yurik 1,736 2,648 736 70.23%
Milos Rancic – User:Millosh 1,646 2,598 876 65.27%
Abbad Diraneyya – User:عباد ديرانية 1,436 2,748 936 60.54%
Abel Lifaefi Mbula – User:BamLifa 1,465 2,605 1,050 58.25%

FDC members election results

Candidate Support Neutral Oppose Support Percentage S/(S+O)
Michał Buczyński – User:Aegis Maelstrom 365 468 117 75.73%
Lorenzo Losa – User:Laurentius 345 476 129 72.78%
Liam Wyatt – User:Wittylama 346 469 135 71.93%
Osmar Valdebenito – User:B1mbo 317 504 129 71.08%
Katherine Bavage – User:Leela0808 327 482 141 69.87%
Chris Keating – User:The Land 301 510 139 68.41%
Frank Schulenburg – User:Frank Schulenburg 273 501 176 60.80%
Nurunnaby Chowdhury – User:Hasive 262 512 176 59.82%
Viswanathan Prabhakaran – User:Viswaprabha 254 507 189 57.34%
Ad Huikeshoven – User:Ad Huikeshoven 201 545 204 49.63%
Abel Lifaefi Mbula – User:BamLifa 154 549 247 38.40%

FDC Ombudsperson election results

Candidate Support Neutral Oppose Support Percentage S/(S+O)
Kirill Lokshin – User:Kirill Lokshin 432 436 82 84.05%