Wikipedia:Wikipedia Signpost/2015-11-18/Special report



It’s election season again: voting will soon open for the 13th annual election of the English Wikipedia's Arbitration Committee. In a repeat of last year’s election, there are nine vacant seats on the 15-member committee, eight of which will carry two-year terms and one a one-year term. Following two withdrawals, there are 22 candidates, up on last year’s 18; five of them have already served on the Committee for at least one term: LFaraone, GorillaWarfare, Kirill Lokshin, Thryduulf, and Casliber. Many eyes will be on the number of voters, which last year shrank precipitously by more than a third, from 923 to 593.

As in previous ArbCom elections, the electronic interface SecurePoll will be used, with support–neutral–oppose ternary choice and an unusual S/S+O formula. In the 2013 election, the use of this system made a difference to who was elected compared with the number of supports alone, and last year changed who was given the one-year term.

This year has marked one of the most fractious in the history of the Committee, in which judgment voting patterns have at times shown mild evidence of the formation of blocks of arbitrators, depending on the theme. There were several drama-infused cases, including the Gamergate case, which attracted unfavourable outside press coverage; gender-related cases appear to be a point of divergent viewpoints among the arbitrators. The current election will influence whether the Committee can regain cohesion and weather external shocks, including emotionally charged cases and critical coverage by external news outlets.

As well as reading the candidate statements and question pages, community members have written 15 voter guides, almost as numerous as the candidature. Against this, however, the Signpost is providing a different angle by presenting the results of an emailed survey to candidates on both their personal qualities and their views on ArbCom-related issues. This is the methodology we have used twice this year in our coverage of WMF Board and FDC elections. With so many candidates, it is a way to provide voters with comparative data gathered on a large scale in isolation, eliminating the "herd" effect in which candidates' responses are influenced by those of their colleagues. This is at the expense of reducing candidates' views to numbers, so we invited short statements to give respondents the opportunity to state more nuanced views—taken up by only a minority. The survey and the writing of this story was designed and supervised by the Signpost's Editorial Board; editor-in-chief Gamaliel was excluded from the process because he is standing in the election.

The survey
There are several findings of interest. The candidates overall think that cases take too long, that case procedures should be streamlined, and that it's too hard for community members to extract the important messages from ArbCom’s judgment pages. They are satisfied with the voting system in the election, and believe the WMF should take more responsibility for minor issues. There is mild consensus that ArbCom's scope to manage excessive behaviour should not be widened, and that off-wiki outing is never acceptable. Consensus is less clear on transparency issues. The candidates are significantly divided on Gamergate and gender treatment.

We received 18 responses; one candidate, Kirill Lokshin, did not respond; Kelapstick apologised that he's "in a jungle" with bad connectivity; two more refused to participate on the grounds that the questions "require candidates to reveal information that they chose to withhold in their nominations or chose not to reply to the users' questions" (Kudpung), and were not "transparent" (NE Ent). Hullaballoo asked to withdraw his responses well after the announced copy-deadline, and after the data analysis had been done, a request we declined. One of the candidates who responded, Samtar, has since withdrawn.

We used a seven-number Likert scale, exploring a six-point response space from 1 to 7:
 * 1 = Strongly disagree
 * 2 = Disagree
 * 3 = Mildly disagree
 * 4 = Don’t know or can’t decide
 * 5 = Mildly agree
 * 6 = Agree
 * 7 = Strongly agree

Each candidate was invited to put a number against each of 20 propositions, and were informed that blanks would be counted as "4" for statistical purposes (these are marked red in the table). We have abbreviated candidates' usernames for reasons of space; those who have served on the Committee are marked with an asterisk. The full wording of the propositions appears at the bottom of the story. Averages and standard deviations appear first;1 then net positives (5–7) and negatives (1–3), which disregard the strengths of the responses to focus merely on which side of neutral candidates lie as a whole.

Key to the abbreviations of the candidates' names in the first row of the table below:


 * Cal = Callanecc
 * CL = Casliber
 * D = Drmies
 * G = Gamaliel
 * GW = GorillaWarfare
 * H = Hawkeye7
 * HW = Hullaballoo Wolfowitz
 * K = Keilana
 * KG = Kevin Gorman
 * LF = LFaraone
 * M = Mahensingha
 * MB = MarkBernstein
 * OR = Opabinia regalis
 * RF = Rich Farmbrough
 * S = Samtar
 * T = Thryduulf
 * TT = Timtrent
 * W = Wildthing61476

Making sense of the data
Our motivation was mainly to survey attitudes to ArbCom-related issues by the group as a whole (this is a highly relevant cross-section of the community—those who put themselves forward for election). However, before they cast their votes, editors may be interested in scrutinising the responses of individual candidates.

Data interpretation can never be 100% objective, and the Signpost welcomes critical comments and discussion on the talkpage below. Propositions C and D we regard as likely to attract a higher level of public-relations calculation by candidates, which explains the narrow, positive range (who would admit they're unprepared for managing difficult editors or can't write clear judgments?); statistics are less relevant here and are not included. Proposition A might have been in the same class, except that the focus is on evidence (candidates' "record"), with responses ranging from 1 to 7. B is the hanging judge question, which might attract more scrutiny from voters: eight favour light over strong sanctions, four sit on the fence, and six favour strong sanctions (not surprisingly, only one of them going beyond mild agreement).

Let's deal first with the five propositions on which there appears to be clear consensus among the candidates:
 * K: Cases need to take less time. There are 16 positives, 12 of them agreeing or strongly agreeing; 2 negatives, both mild disagrees; 0 neutrals. Three of the four candidates with experience on the Committee agree (i.e. "6"), with one mild agreement ("5").
 * L: Case procedures need streamlining. 13 agree (only three of them mildly); 4 disagree (none strongly); and one is neutral. There was a high correlation with the responses to the previous proposition (K). One candidate with Committee experience flipped from agree to mildly disagree.
 * R: The WMF should take more responsibility for minors issues. 14 positives (8 of them strongly); 2 mildly disagree; 2 were neutral. Over the past few years this has been the subject of discourse with the Foundation, which no doubt would like to minimise its legal exposure and the need to deal with the plethora of jurisdictional and linguistic issues on its sites.
 * I: It's easy enough for the community to get the messages in judgment pages. 13 are negative, 4 of them strongly; 5 are positive, all but 1 of them mildly; no one is neutral.
 * T: I'm satisfied with the voting system for ArbCom elections. 14 agree (5 strongly); 2 disagree (neither strongly); and 2 are neutral.

On four questions there is only a modest consensus:
 * E: ArbCom's scope to manage excessive behaviour needs to be widened. 11 are negative (only one mildly); 5 are positive (none strongly); 2 are neutral.
 * P: Off-wiki outing is never acceptable. 11 positives (four strongly); 5 negatives (three strongly); 2 neutrals.
 * O: ArbCom should disregard off-wiki evidence. 10 positives, 4 negatives (three strongly); 4 neutrals.
 * H: Arb burnout is a major problem to address. 10 agree (3 strongly); 2 disagree; 6 are neutral.

Consensus is less clear in two related questions about transparency, especially the second one:
 * F: All case evidence should be on-wiki. 10 are negative (4 strongly); 7 are positive (none strongly); only one is neutral.
 * G: Arb discussions on cases should be on-wiki. 9 are positive (2 strongly); 6 are negative (1 strongly); 3 are neutral.

On five questions, the candidates show no consensus:
 * J: Accept fewer cases, leave more for AN/I etc. 8 are positive (2 strongly); 6 are negative (none strongly); 4 are neutral.
 * M: ArbCom was at its worst in handling Gamergate. 8 are positive (2 strongly); 6 are negative (none strongly); 4 are neutral.
 * N: I was satisfied with the Gamergate decision. The results were spread through the range 1–6, with 3 neutral. Interestingly, candidates were on average less positive about the decision than the handling of the case.
 * Q: ArbCom has treated men more sympathetically than women. 8 positive (4 strongly, 1 mildly); 6 negative (none strongly, 4 mildly); 4 neutral.
 * S: I'm happy with the Audit Subcommittee arrangements. There is certainly a negative majority (9, 4 of them strongly), with only 2 positive (a "5" and a "6"); but 7 were neutral.

Voting
The two-week voting period will open at midnight on Monday 23 November (although some voters may be confused as to whether this refers to midnight start or end of Monday). Voters are advised that the arithmetic of the ternary system means that opposing all candidates they are not supporting, rather than voting neutral for them, is a more powerful confirmation of their supports. An election feedback page has been established.

1 Standard deviations are a measure of spread. They can be roughly visualised as the space that contains a third of responses above the average and a third below it. If 5.0 were the average and 0.5 were the standard deviation, two-thirds of responses, roughly, fall within the range 4.4–5.5. The larger the standard deviation, the more divergent the candidates' views. Full wording of propositions:
 * (A) I have a record of minimising drama in emotionally charged onwiki situations.
 * (B) I see myself as being on the tough side of the spectrum, favouring strong over light sanctions.
 * (C) I’m prepared to manage difficult editors under the radar where privacy is imperative.
 * (D) Drafting cases in plain, simple language will be one of my strengths as an arbitrator.
 * (E) The current scope of ArbCom’s powers to manage excessive behaviour needs to be widened.
 * (F) Case evidence should always be onwiki for the sake of transparency.
 * (G) Discourse among arbitrators about a case should not be conducted behind closed doors.
 * (H) Arbitrator burnout is a major problem and the Committee should address it urgently.
 * (I) ArbCom’s judgment pages are easy enough for community members to extract the important messages behind judgements.
 * (J) ArbCom should accept fewer cases and leave more to community forums such as AN/I.
 * (K) Cases generally need to take less time.
 * (L) Case procedures need to be streamlined.
 * (M) We saw ArbCom at its worst in the way it handled the Gamergate case.
 * (N) Given the difficult circumstances, I was satisfied with the final judgments on Gamergate.
 * (O) ArbCom should disregard evidence posted off-wiki in its judgments/sanctions.
 * (P) Off-wiki outing of a Wikipedian is never acceptable, regardless of the circumstances.
 * (Q) ArbCom has tended to treat men more sympathetically than women.
 * (R) The WMF should take more responsibility for managing issues related to editors who are minors.
 * (S) I’m happy with the current arrangements for the Audit Subcommittee.
 * (T) I’m satisfied with the oppose–neutral–support voting system in the annual election.


 * Editor's note: Andreas Kolbe and Rosiestep made important contributions to help frame the questions as well as format the data on this page. Editor-in-chief, Go Phightins!, was also involved in the formulation of the questions and curation of the data. However, Tony1 deserves the bulk of the credit for this coverage. Questions or concerns can be directed to the Signpost talk page or the editor-in-chief's talk page. Thank you.