Wikipedia talk:2024 administrative elections proposal/Archive 1

Scrutineering

Factual correction: The Arbitration Committee elections don't currently use bureaucrats to scrutineer, they use stewards.
Not sure bureaucrats are a great choice for this. How about existing checkusers instead?
Not sure having to go through a special process to become a scrutineer is efficient. How about existing checkusers instead?
SecurePoll voting and the scrutineering process, if I understand it correctly, means that each voter gets checkuser'd every time they vote, possibly bringing up WP:NOFISHING issues.

–Novem Linguae (talk) 08:52, 6 January 2024 (UTC)[reply]

Trout for me, I meant something totally different in re: ArbCom scrutineering practices! As for the fourth point, if that's not a problem for ACE, I don't think it should be a problem here. For the second and third points, I doubt that the checkuser team wants this added to their responsibility, and the 'crats are responsible for moderating RfA. theleekycauldron (talk • she/her) 08:59, 6 January 2024 (UTC)[reply]

So the idea is to have bureaucrats volunteer to be scrutineers, go through some kind of RFA or ACE-like process to get confirmed, then they have to do the process to become checkusers (sign NDA, go through a second public comment process (example: Wikipedia:Arbitration Committee/CheckUser and Oversight/2023 CUOS appointments) that checkusers go through so that ArbCom can properly screen them, then get approved by ArbCom), and then the 0 or 1 people that go to all this trouble can finally scrutineer admin elections? That workflow seems too inefficient to me. Should just use existing checkusers, imo. –Novem Linguae (talk) 09:06, 6 January 2024 (UTC)[reply]

Okay, how about the 'crats can volunteer unilaterally, and ArbCom has the power to screen and appoint who it wants in times of shortfall? ArbCom can deny scrutineering power to a 'crat in an extreme case, but it doesn't need to exercise the same scrutiny it always does. Using enwiki checkusers skips one step, but they'd still have to be screened the same amount in order to get access to votewiki. theleekycauldron (talk • she/her) 09:14, 7 January 2024 (UTC)[reply]

Technical challenges

There may be other technical obstacles to using SecurePoll. It may require WMF involvement, and there is some kind of encryption key system that I don't understand completely but that makes the process more complicated. You could ping JSutherland (WMF) or AntiComposite for more details, they seem to understand the extension.
I'm not sure how fast the SecurePoll creation process is. Can a SecurePoll be spun up quickly or is it some laborious checklist that takes a special person a month to get to it?
Is SecurePoll scalable to high volumes? Can this workflow handle dozens of RFAs a year? From previous discussions, I get the impression that it cannot and that everything about this extension's workflows are laborious, but I could be wrong.

–Novem Linguae (talk) 08:56, 6 January 2024 (UTC)[reply]

I'm currently in correspondence with JSutherland about this – because of the holiday and a weekend, we haven't discussed much yet, but I'm optimistic :) theleekycauldron (talk • she/her) 08:57, 6 January 2024 (UTC)[reply]

Perhaps invite JSutherland to this talk page at some point so these details can be shared publicly. I am curious to learn about this extension, and I don't feel like many folks know how the extension works in practice. –Novem Linguae (talk) 09:08, 6 January 2024 (UTC)[reply]

Diverging from RFA norms

10 days for discussion and vetting instead of 7. These elections will be conducted via SecurePoll for the last 7 days of the election. Candidates who clear 65% of the vote are granted adminship; candidates who achieve 55% or more are granted adminship temporarily. I'd recommend sticking as close to current RFA norms as possible. The more changes you try to put in one RFC, in my opinion the more likely it is to fail. So I'd recommend 7 days of discussion and voting running simultaneously, no temporary adminship, and a cutoff for full adminship somewhere in the discretionary range (65%/70%/75%, whatever is most palatable to people). –Novem Linguae (talk) 15:12, 6 January 2024 (UTC)[reply]

I'd be willing to go up to 70/65, but I will point out that the median RfA candidate last year scored 28 points higher than the median ACE candidate. The cutoff has to go down, or else I worry that we're right back where we started and don't actually pass more candidates. A full seven days is needed for voting, and since all candidates need to be voted on simultaneously, I'm not sure we should be allowing people to make up their minds and walk away right off the bat. There should be at least a moment of reasoned discussion to look at before people cast what will likely be a vote they'll never revise. theleekycauldron (talk • she/her) 09:08, 7 January 2024 (UTC)[reply]

Support

I have no questions or comments to share yet, but just wanted to express my support for this proposal. Though I'm relatively new here, I've certainly seen the problems at RFA that the proposal is attempting to reduce. I appreciate the attempt to fix these problems, and look forward to a future RfC. NotAGenious (talk) 14:44, 7 January 2024 (UTC)[reply]

Draft comments

Proposal seems in good shape though I have a few pieces of feedback from the perspective of someone unlikely to support but willing to stand aside.

The arguments in the "current system" section are not equally compelling. If these are based on the consensus reached in various discussions, they should be linked to. If they are interpretations of multiple discussions, they need better explicated. I think the problem that is being papered over in this section is that there's no real consensus on what the problem is; everyone has their own belief as to the central problem, many contradictory, but no particular view has commanded consensus. Not adequately addressing this leads to confusing drafting and points for opponents to (disingenuously) pick at. Some examples:
- Point 1 states "Participants frequently write toxic or incivil votes and replies without consequence" but point 2 undermines it saying that "Participants are also frequently badgered or criticized for their rationales." The problem isn't that toxic or incivil comments go without consequence. The problems include (1) "a systematic failure of administrators and bureaucrats to enforce norms" and (2) "criticism adds to the cacophony and corrosive atmosphere". These are, in fact, already in the proposal, but because of how the issues are framed, these points wind up undermining each other rather than being synthesized into a coherent problem statement.
- Point 3 states "its format frequently fails to raise coherent, consistent cases for opposition or its rebuttal" but this claim seems poorly motivated. If this were true, why do we have cases where a single well-reasoned oppose shifts the tide of an RfA? Why do we see cases of supporters changing their opinion and moving to oppose? Why do we have a coherent (if undocumented) set of norms and expectations for candidates (i.e., 2 years, 10k edits, 1 piece of good content, need for tools)? Why do we see cases of editors who repeatedly ask silly questions or give inane oppose rationales stopping after community feedback and intervention? I'm not saying the claim is wrong per se, but it ignores the successes of the system, its reformers, and its participants to make an alternative/replacement seem more attractive. While likely unintentional, it is a recipe for attracting strong opposition from community organizers (i.e., the reformists who have historically been quite successful).
- Point 3 states "Participants are expected to read mountains of text...in order to reach a decision on their vote" and frames it as a problem. I suspect that there are a few people who view this as a feature and not a bug, and it's worth considering how the opposition will frame a rebuttal. For example, if we agree that this is a problem, the proposal is not obviously solvent: participants may still need to read through mountains of text during the discussion phase with the only substantial gain being the lack of one-word "supports" that anyone with basic reading comprehension skills should figure out they can skim past. If we don't agree that the volume of text is a problem, the proposal trades social cohesion for governance risks. We want an informed electorate, and what is painted as a flaw is in fact a design feature (a meatball:PricklyHedge) that creates a cost (but not prohibition) for uninformed participation leading to a higher floor on discussion and decision-making. There are issues with this (e.g., there are systemic biases in which genders, economic classes, and cultural backgrounds feel more comfortable barging into areas where they have no competence), but those issues are what needs addressed if the argument is to be discounted and consensus reached.
The theory of change is left implicit; the causal connection between the principles and the proposed process is not stated. This is a negative for comprehension, but more importantly it is a negative for evaluation and future improvement. What parts of this proposal are crucial for the change you wish to achieve and which are arbitrary parameters that we can adjust later? For example, the vote threshold is probably not crucial to achieving a more collegial atmosphere or attractive process---it's just a requirement of a voting system and we can tune the level based on practicality later. By contrast, having a developer remove the names of withdrawn candidates is directly in service of improving on the issue of unreasonable scrutiny identified in the principles. Identifying these causal links in your theory of change helps identify what is a core issue and what is a side issue when determining consensus of the discussion, and it helps us figure out what to change in the future. How will we know whether this process achieved its goals? If it fails to achieve a particular goal, how will we know what assumptions were incorrect and need to be improved? If, two years from now, discussion is still not an improvement or the quality of promoted candidates goes down, then we can identify what parts of the theory need revised without throwing out the whole thing.
Your ideals of design should be revised to be less reactionary. They should proceed from shared norms and values, and ones not clearly identifiable as having broad consensus need to be more thoroughly justified. Some examples:
- Ideal 3 states that participants should "be incentivized to give useful information or insightful analysis on a candidate's fitness, rather than the minimum text needed to justify participation" which only makes sense in reaction to current system point 3. In the proposed system, there is no minimum text needed to justify participation, so it's not so much an ideal as it is a repeated criticism of the alternative system. The rest of the proposal does not identify what incentives exist in the system that would achieve the ideal, further confusing what purpose this ideal serves. These paint the other principles negatively: are the principles truly ideals that any system should attempt to achieve or are they just re-framed arguments against the current system?
- Ideal 4 strikes me as internally inconsistent. It states that "Participants should be able to vote their conscience...free from social pressure to vote one way or another" which, if I'm reading uncharitably, is silly. The portion I elided gives an example of why this ideal is inconsistent: "discussion and questioning" that participants are encouraged to read is a kind of social pressure but because the supporters of the outcome view it as a "good" or "acceptable" form of social pressure, it gets a rhetorical pass. What is actually being advocated here is a freedom from consequences of a choice by removing the ability to verify that choice. That is not necessarily a bad thing; in systems where threats of violence and monetary extortion were used to influence voting the need to shield victims from those criminal consequences necessitated removing the ability for criminals to verify choices. The issue is that the principle assumes that secret ballots are superior and comes up with a post-hoc justification for this, but this is not a universal belief even in political governance, let alone Wikipedia, with the Iowa caucuses being a famous counter example. Is the claim of this proposal that the people of Iowa are unable to vote their conscience free of social pressure because of their open ballot, and, conversely, that the people of Australia experience no social pressures to vote in a particular way because of their closed ballot? The ideal suggests that our implementation of a secret ballot will be utopian with perfectly informed and independent voters experiencing no social pressures or influence at all. I find that hard to believe, and a naive design goal. I think the proposal would be more compelling if it were transparent in what kind of social pressures it wishes to privilege and what kinds it wishes to disincentivize because that, historically, has been the kind of debate which actually convinced societies to move between open ballot systems and closed ballot systems. The choice between a secret and open ballot isn't
The community sanctions/contentious topics designation might be a point of contention that may or may not be worth it. You may wish to run it as a separate question. The benefit there is that it would generalize to the open ballot process and achieve the goals in the new and old systems rather than just the new one.

Having said all that, my goal isn't to discourage and I hope to see this put forward in the near future. I think this is probably the best proposal I've seen so far, and that's no small achievement. I'm writing all this because I think it's good. Whether you take the advice or not, hopefully it gives some perspective on the kinds of opposition it may face going forward and provide some points to reflect on in future proposals. I especially like the point that it is an alternative rather than replacement, as this is a pattern I'd like to see used more. I appreciate that it considers the fact that a single process doesn't fit all. I also like that it gives crats something to do for once and implements a system that doesn't overburden groups like CUs and stewards that already have expansive tasks and backlogs. These are nitpicks, but it's the nits holding back the formation of consensus. Hopefully by identifying those points, they can be better addressed and a true consensus formed. — Wug·a·po·des 21:22, 18 January 2024 (UTC)[reply]

@Wugapodes: this is the kind of criticism i live for :D let's get into it!

Fair point. My overarching view is that there are two systems we have to keep opposition (and rebuttal) in check: the social pressure and the formal consequences. The latter basically doesn't exist, which is what I meant by CS#1. The former creates a vicious cycle, because I would contend that it prevents good and informative opposes as much as it prevents bad ones. People who want to mouth off are protected by the discussion-as-vote model because they don't care about the social consequences, while people who could otherwise construct a good argument are too scared of the social consequences to exercise their right as a member of the community to have a say in how a candidate is elected. That makes bad opposes more prominent, which invites the most caustic supporters to rebut (which skew caustic for the same reason the opposes do), making opposing more costly, driving away more would-be good opposers, perpetuating the cycle.
Have reworked :)
Have again reworked :)
- This is something I hope I've made clear in the aims and objectives section (frmly "Ideals of design") now, but the idea is that admins and 'crats are responsible for enforcing this. They have two new tools with which to do it: requiring clear consensus to overturn their actions, and the fact that an action can't directly infringe someone's right of equal suffrage, thus lowering the stakes.
- I guess my only comment on this would be that the current system isn't accomplishing what it wants to with social pressure. It's not keeping incivility out (it does presumably scare off some people, but if you stop 4 out of 5 rocks chucked at a pond, the pond's calm is still ruined), but it does disincentivize people from voicing what might well be a legitimate grievance. People who are brave enough to comment are welcome to do so – if they violate norms, there'll be problems, but it shouldn't interfere with their right to have a concrete say in the results.
I mean, in theory, people can oppose one part or the other, and the proposal can pass with modifications. I'll think about this one, though.

With all that in mind, I've taken a stab at reworking it :) let me know what you think. Thank you so much! theleekycauldron (talk • she/her) 01:32, 19 January 2024 (UTC)[reply]

@Theleekycauldron: I like the changes a lot, and don't have much to say on the proposal as it stands. Some responses with an eye to the long term:

I think this is reasonable, but I would need to see more data. In the last year, we've had 4 unsuccessful RfAs that broke 100 votes and two successful RfAs that fell below 80% (high opposition) compared to 10 successful RfAs with over 80% support (low opposition). That ~37% of serious RfAs routinely get high levels of opposition suggests that the social cost of good opposition is not unreasonably high. The hypothesis you have is that some portion of those low opposition RfAs are actually high opposition RfAs but are artificially inflated due to social pressures. That's certainly possible, but I think there's a confound with who gets run. Most people don't run an RfA unless there's a good chance for success, so it's not surprising that there are a lot of low opposition RfAs---we rarely see high opposition ones because they never get run or get withdrawn before they really start. So I think the lack of "moderate opposition" RfAs is largely due to a sampling bias towards over-qualified candidates and less to do with the social cost suppressing opposition. This is an empirical question though, and I think the proposal will shed some light here.
"the current system isn't accomplishing what it wants to with social pressure." The purpose of a system is what it does. The instances of incivility have a role in the system as well, they're not accidents but part of what the system is meant to produce. It's not the badgering that stops me from posting opposes, it's being in the same section as people showing up and grinding axes or raising personal grievances. It leads to editors with serious concerns discussing them first in the neutral section or talk page in order to form a critical mass and come to a consensus on whether a concern is actually a blocker before anyone needs to toss their hat in with the axe grinders. The proposed system is attempting to produce that exact same outcome by separating discussion from voting, but the hope is that it won't also result in incivility. It's worth remembering that none of our systems prevent all incivility, so proceeding from the assumption that it can be removed is in my opinion a trap. The question is how to manage it and reduce the negative consequences while maintaining the purpose conflict serves in social systems, and the "two new tools" you're proposing help with that.
Similarly, "Disincentiviz[ing] people from voicing what might well be a legitimate grievance" is arguably a feature: RfA is an exhibition of the candidate's and their allies' ability to get the benefit of the doubt. If a bunch of people I trust are vouching for someone while I'm on the fence, I'll probably just sit out, not for fear or social cost, but because I trust their judgment. If the candidate turns out to be a bad admin, I don't get bonus points for having opposed them; I'm not going to go around to my colleagues gloating about how I knew their judgement sucked. So why spend my time writing feedback I am certain the candidate's mentors are already giving them in private and being a blocker to their promotion? It's far better to just sit out; not everyone needs to know what I'm thinking at all times. Even my OP here references my willingness to stand aside when this proposal goes live, an actual option in meatspace consensus decision-making models which exists here even if we don't talk about it much.
The idea that leaving a comment at RfA is the only way to "have a concrete say in the results" is simply one I don't subscribe to. That's the core disconnect in my experience, many people believe their voice only matters if it winds up in the final yea-nay tally. Polls scratch that itch, but the reality is that there are many ways to have a say in the results without participating in a poll. Using this RfC as an example: will I not have a concrete say in the results if I choose to stand aside? Does this feedback and the subsequent edits it inspires not count as having a concrete say? People seem to view RfA as a singular event in a way they don't view other processes. Even the development of an email list for candidate mentorship and development acknowledges the reality that there are ways to have concrete impact without "Voting". Recall the first bullet point w/r/t the lack of "moderate opposition" RfAs: there are multiple stages of gatekeeping prior to transclusion that lead to a specific kind of candidate not even making it to RfA. Most admin hopefuls "fail" long before they make it to RfA; the time to be a real blocker was way earlier so it's rarely worth me casting an oppose.
In my experience, for large process RfCs, it helps to keep things focused and assume you won't reach consensus on anything not asked. There's the possibility of it happening, but there are just so many participants and so many ideas getting floated that it can be hard to find a consensus stronger than the actual proposal.

All that said, it's obvious which way the wind is blowing. For it's problems, I think the previous RfA system was an interesting system that came up with cool strategies for producing particular social outcomes. The problem is that a lot of those systems relied on social norms that reinforced systemic bias and prickly hedges that produced negative social outcomes elsewhere in the encyclopedia. Those issues deserve to be fixed, but the moral and ethical aspect often clouds analysis of the system. Personally, I think there's actually consensus on the kinds of outcomes we want to see, but the moral baggage of RfA is a vote versus RfA is a consensus discussion leads to ideological shouting matches rather than a discussion of how to build a system that achieves those shared objectives. A big concern I have for most "RfA is a vote" reform arguments is that they often boil down to people advocating for whatever system they are familiar with from their own government without considering those systems also have problems that need to be managed or the practical issues of implementing such a system. I like that the proposal doesn't fall into that trap and instead puts forward a whole system that considers the fact that it also has trade-offs which need to be managed in order to achieve the shared underlying goals. — Wug·a·po·des 22:12, 21 January 2024 (UTC)[reply]

Consensus

Just noting that Wikipedia:Requests for adminship/2021 review/Proposals#Closed: 8B Admin elections had 72 in favor and 39 opposed, and somehow closed as consensus against, which was overturned to no consensus.

With those numbers and the right closers, I believe a second RFC would find consensus to try admin elections.

The consensus for admin elections should be a completely separate issue from the technical details. Step 1 is to launch an RFC and figure out if the community wants this (it does). Step 2 is then to take that community mandate and explore how to do it.

Arguments that say "I'd support this but we don't have the technology to do it" should be weighed as supports, not as opposes. The step 1 RFC should be to figure out the community sentiment, not the technical details. –Novem Linguae (talk) 09:01, 6 January 2024 (UTC)[reply]

Temporary adminship

Hi. A question about the proposal: For those receiving between 65% and 70% support and thus obtaining "temporary adminship for six months," what happens at the end of the six months? Regards, Newyorkbrad (talk) 02:29, 22 January 2024 (UTC)[reply]

@Newyorkbrad: oh, good question! in my mind, it works same as any other perms request – the granting bureaucrat grants the sysop flag with a time limit, and it automatically expires. No automatic right of resysop, since the mandate has expired, but not a for-cause desysop either. A temp admin is welcome to extend their mandate by launching another RfA (either route is fine). theleekycauldron (talk • she/her) 02:33, 22 January 2024 (UTC)[reply]