User:TenOfAllTrades/CDAresponse

Flaws in this process noted by TenOfAllTrades

The nature and development of the CDA process

Caution: Imagining CDA as a 'reverse RFA' or mirror image of RfA is misleading and harmful.

The CDA proposal has been portrayed as a 'mirror image' or 'reverse' of the RfA process. This representation is potentially quite harmful, as it implies certain symmetries. While at the most simplistic level one could argue that the two processes are equal and opposite – one flips the switch on, one turns it off – in practice the circumstances surrounding each case are very different.

RfA is voluntary; CDA is compelled. An RfA starts at the time of an editor's choosing. A CDA is imposed by other editors, generally at a stressful time, and probably right after the admin was involved in a major conflict.
RfA generally begins with the presumption that a candidate is competent and ought to be promoted. CDA demands at least prima facie evidence for misconduct, misuse of tools, malice, or incompetence. CDA is inherently confrontational.
RfA begins with (at most) one or two nominators' support in the tally. CDA begins with the admin 10 votes in the hole, as all of the nominators can be expected to vote.
RfA is predominantly about an indvidual's contributions and efforts. CDAs will almost always arise from extensive interpersonal conflicts, and will often fit within a pattern of broader disputes among multiple parties.
RfA places virtually all of the evidence and discussion right on the RfA page. CDA hides the admin's statement and counter-evidence on the talk page, leaving only the nomination.
RfA is a frequent, everyday event. Standards are well-established, most candidates typically face the same sort of questions and examinations. Most participants (candidates and voters) in the RfA process know what to expect, and have extensive experience in it. CDA will be an exceptional occurrence. Neither parties nor voters will have much idea of what to expect, nor will there be an extensive history or archives to look back on.

The glib assumption that since the effect of the two processes mirror one another then the mechanics of the processes ought to mirror one another too does not stand up under scrutiny. It is a specious, even dangerous false analogy.

Proposal development was hampered by set-in-stone votes barring review of previous decisions.

While this proposal has nominally been under development since at least mid-November, in practice the process being voted on now is essentially identical to the proposal drafted by Uncle G on 5 October 2009: [1]. Despite the many issues raised below, any suggestion of substantial changes to the proposal have been rejected as the issue was 'already decided'. A series of rigidly-structured majority votes have been used to introduce minor tweaks (number of days for nomination, adjusted eligibility rules, suggested percentages for 'consensus' to reduce Bureaucrat discretion, etc.), but there has never been any question of – or opportunity to – modify the original process in light of subsequent discussion.

Kim Bruning recently offered perhaps the most insightful analogy; I will allow those words to speak for themselves: [2].

Policy proponents have not identified any specific situations where CDA would be useful, nor any metrics against which it should be evaluated.

The proposal's proponents have repeatedly rejected calls to describe any specific situations – past, present, or future – where they believe that CDA could be used to benefit the Wikipedia community. In their FAQ, they explain this reluctance on the basis that such discussions might lead to "abusive sidetracks". At some point, the community will find out how this proposal will be employed. I think it would be far better to be open about this before it becomes policy rather than after — and it isn't fair to ask Wikipedians to vote on this pig in a poke.

As recently as 2 February, Hammersoft was still trying to get the proposal's proponents to express clearly and straightforwardly what the purpose and goals of the process should be. He was asking how we should define success or failure of the process — what problems is it expected to solve? His request was briefly rebuffed, and ignored.

Issues of procedural fairness

By far my greatest concerns about this process turn on its gross unfairness to its participants — especially the administrator being examined. The proponents of the process have been very concerned about creating a process which is very rapid, which has a low barrier to entry, which sysops should find genuinely threatening, and which has a 'democratic' appearance. Unfortunately, the result is a process that does not contemplate an administrator who wants to defend his actions, situations where there is misconduct by multiple parties (particularly by the nominators), or any interest on the part of participants in examining the evidence or discussing the situation.

Accusers enjoy a privileged, protected position from which to attack.

The sole issue open for consideration in the CDA process is whether or not the subject admin is to be desysopped: yes or no. While the CDA proposal offers the vague acknowledgement that the nominees' conduct will also be subject to scrutiny, the plain fact is that this process only generates one actionable response to one specific yes/no question. Unlike all other dispute resolution procedures on Wikipedia (RfC, RfArb, AN/I, etc.), the only party who can face sanction as a direct outcome of CDA proceedings is the sysop.

An unjustly accused admin – or even a justly accused admin who is not solely to blame – has no directly parallel venue in which to address the conduct of his accusers. Taking nominators to AN/I during the course of the CDA is bound to provoke an inconclusive but destructive firestorm. Any other option (RfC/U, RfArb) which might allow the community to deal with nominators' misconduct will run much more slowly than CDA. An admin who survives CDA might then have relitigate the same issues at RfArb. An admin desysopped by CDA will be accused of seeking revenge if he attempts to address the nominators' misconduct. A CDA will muddy the waters and exhaust the community's interest in dealing with any non-admin portion of the dispute.

Canvassing is essential to ensure the participation of involved parties, but will unbalance the decision-making portion of this process.

This process operates on short timelines. The nomination window is a maximum of seven days long, and the voting period is a further seven days. In order for a nomination to be successful, editors who share concerns about a particular admin will need to be able to organize themselves to marshal their arguments and evidence. This will require coordination and requests for participation. Similarly, the admin under scrutiny should be able to invite the participation of other editors who may have insight into or experience with the dispute in order to present a thorough, balanced defence. Whether such coordination occurs on-wiki, by email, over IRC, or at Wikipedia Review, it will happen. Moreover, in the preparation of a complaint (and in the presentation of a response), broad solicitation of involved and interested perspectives and testimony will often be highly desirable. To allow the community to fairly judge an administrator's actions, we want to have access to as much relevant detail as possible.

Unfortunately, once all of these editors have become involved in the preparation of a nomination (or a rebuttal), there is no way to ask them all to return to blissful ignorance for the voting stage of the CDA. The first, and loudest, voters on CDA are almost certainly also going to be the most deeply-involved parties.

CDA voters are self-selected, not community-selected. Many voters will be partial and biased.

As a corollary to the above point, it should be noted that participation in the voting is not in any way limited beyond the bare 'usual' blanket editing restrictions barring socks and banned/blocked users. While this has occasionally been presented as making the process 'fair' or 'representative' or allowing for a 'jury of one's peers', in practice we must acknowledge that there will be three rough groups voting in any given CDA.

Nominators and friends. These individuals drive the nomination and/or have had run-ins with the admins in question. However good-faith the nomination was, this group will end up attracting every POV-pusher who ever received a 3RR block.
Admin and friends. This bloc contains all of the people who think the admin is overall doing a good job, who disapprove of railroading them through a CDA, plus all of the people who think that any of the editors in the first camp are disruptive, wikilawyering wastes of time.
Crusaders. This last category contains all of the self-appointed Guardians of the Faith. They have strong (and often negative) opinions about admins collectively and the concept of adminship, and they have found in CDA an opportunity to proclaim on their particular hot-button issues. This group will be the only one which watchlists the CDA page and actively tracks new cases; they will form their own somewhat isolated, somewhat insular community around CDA. Many of them will initially be drawn from the small group of editors who developed this proposal.

Far from being an impartial and independent jury, these three groups will each have their own goals and their own particular axes to grind. The number of fully neutral, uninvolved individuals who will be willing to take the time to find the CDA nomination, review all of the evidence in detail, engage in discussion where necessary to clarify any points of confusion, and then vote (with attendant risk of future abuse from the 'other' side) is vanishingly small — and unfortunately much smaller than the three camps I've identified.

Note that this stands in stark contrast to the process by which (for example) Arbitrators are chosen.

There's no place for the admin to defend himself on the CDA nomination page.

The CDA nomination form includes no place for an admin to offer a response or defense. It's just nomination, signatures, diffs of notifications, and 'evidence in support of the nomination', followed by votes.

In other words, it's easy to drop in, see the nomination, the list of accusations, the cherry-picked evidence, and vote — all without even an inkling that the admin might have had something to say on his own behalf.

Detailed discussion and evidence will be shuffled off to a talk page.

An extended statement from the nominators – and any statement of defense or presentation of counter-evidence – will be stored on the talk page. This arrangement discourages any close inspection or analysis of evidence. Many voters who show up may not even be aware that such discussion is going on. Separating the discussion from the voting page may give the superficial impression of a smoothly-running, uncontentious process, but it discourages the openness and transparency that should be the ultimate aim.

Voting begins before discussion of or challenges to the evidence.

CDA offers no requirement or guarantee that an admin facing this process will be allowed time to prepare and present a statement in his own defense. After the nomination is filed, there is no time allotted to discuss or examine evidence before the process jumps directly into voting.

In an attempt to produce a process which is conspicuously 'faster' than the existing alternatives, corners have been cut.

There is no opportunity for compromise solutions.

Editors may agree that there are problems with a particular admin's conduct, but feel that alternative approaches (short of, or different from, desysopping) could be more beneficial. This lack of nuance is harmful. It could be that Wikipedia would well-served by imposing a topic ban or an interaction ban to keep an admin out of a contentious area. Some manner of probation subject to future review may be equally effective. If the only binding outcome possible is desysopping, it puts the community in an awkward place — do we let the defendant go free, or do we send him to the gallows? Parole, probation, community service? Not on the table.

There are too many bites at the apple.
How many times does an admin need to face the axe? There is no provision to protect admins from double jeopardy. As noted above, a CDA started as part of a larger dispute may require the same issues to be litigated twice: once in CDA, again in a follow-up RfArb. Other parts of the proposal seem to suggest that CDA can be used to desysop after a failed attempt via RfArb.

This process cannot handle sensitive or confidential information effectively.

The usual venue for handing desysopping – ArbCom – has established protocols and procedures for handing sensitive information. Under the heading of 'sensitive' I would include information that is specifically governed by the Privacy Policy, along with emails between users, IRC logs, methods for detecting sockpuppets, etc.

At CDA, the vote marches on during the whole period of time that the admin is seeking permission from email authors, participants in IRC conversations, or checkusers to refer to sensitive or restricted-access information in his own defense. Should an admin describe in detail and in public the method used to identify a recurring sockpuppeteer, just to save his own neck?

The final accounting

The way in which CDA proposes to close its discussions is flawed and prone to failure.

The thresholds are wild-assed guesses.

In technical terms, the thresholds coded into this proposal are 'wild-ass guesses', based loosely on the thresholds that apply at RfA. Given that this is a novel process with no historical precedent on Wikipedia, it seems presumptuous to assume that similar numbers with a similar 'discretionary range' should apply here as at RfA. It also betrays something of a lack of trust in the ability of 'crats to determine what constitutes 'consensus'. Are the 'crats expected to determine if a consensus exists, or just to count votes and add a veneer of respectability to the process?

The decision has a single point of failure: one individual's judgement.

Ultimately the decision comes down to a single bureaucrat's call — yes or no, thumbs up or thumbs down. The 'crats seem to generally be good, competent people, but even good people have off days. With the exception of emergency desysoppings and the very arguable authority of Jimbo himself, we do not recognize the authority of any individual on Wikipedia to desysop. It always requires a formal, clear, final vote by the Arbitration Committee (even the emergency desysoppings are subject to ratification). This proposed process would place that ultimate authority and responsibility – for the first time – in the hands of a single bureaucrat.

I would be the first to acknowledge that members of the ArbCom are at times just as flawed, biased, sloppy, grumpy, petty and human as any other senior editors — including the 'crats. However, the requirement for deliberation and a majority vote of the ArbCom before their motions and remedies take effect means that their individual defects can (hopefully) cancel each other out. The lone 'crat in this proposal doesn't have the backstop of a committee sharing the responsibility or officially making the call. While the single Bureaucrat may engage in public or private consultion before rendering a decision, it's ultimately down to him. Moreover, the first avenue for appeal is also to that single 'crat — one editor (under tremendous pressure and intense scrutiny) is allowed to reverse a decision on his own.

The role of Bureaucrats

By far the most visible role of Bureaucrats on Wikipedia is in the evaluation of Requests for Adminship and the promotion (or not) of new admins. They have both the technical and policy means to grant the sysop bit within the framework of existing policy. What they cannot do - under policy and by deliberate design of the wiki software permissions - is remove the sysop bit once granted.

What this CDA proposal aims to do is grant bureaucrats a new power to enact desysopping decisions. This represents a substantial expansion and shift in their powers and responsibilities.

Bureaucrat selection did not contemplate this responsibility.

When editors have been granted the Bureaucrat rights on Wikipedia, it has generally been through an RfB process. This process has generally focused almost exclusively on whether or not the candidate can carry out the existing roles and responsibilities of a 'crat — largely on whether or not they will judge RfAs in accordance with the established traditions in that area.

Two current 'crats didn't even pass through RfB; they were selected as enwiki's original 'crats so that we would have a way to create admins. None of the current 'crats was ever questioned about his or her views on or potential role in deciding CDA cases. The community has not had a chance to examine them in that light, nor have we had an opportunity to endorse them (or not) to carry out such a responsibility.

Traditional 'crat roles do not involve dispute resolution, inter-editor conflicts, or significant 'weighing' of evidence.

For better or worse, the traditional role of the Bureaucrat has not' been to exercise his own judgement to any signficant extent. After subtracting off sockpuppets, the expected role of the 'crat is to count votes and declare a promotion where the fraction of support exceeds 70-75%. The post is largely ceremonial, rather like a constitutional monarch who on paper holds substantial reserve powers, but who principally rubber-stamps decisions made by raw voting.

In the vanishingly few cases where a bureaucrat has seen fit to pass a candidate with less than 70% support it has prompted an uproar. I know of no cases where a 'crat has failed an RfA candidate with 80% (or more) support. The community's expectations of our 'crats in evaluation of an RfA are generally that they will act as 'moist robots' and just count the votes. If an RfA falls into what is the only 'true' discretionary range of roughly 70 to 75%, the 'crat can close as 'no consensus to promote', or just toss a coin. The community doesn't compel detailed explanation, and the understanding is that a borderline candidate can edit heavily for a few months and then come back for his successful RfA.

Oddly enough, the current wording of this proposal (and a source of much ~~wrangling~~ voting during its development) implicitly seems to acknowledge a lack of faith in the 'crats judgement. This CDA proposal imposes specific numeric bounds above and below which Bureaucrats are emphatically discouraged from exercising their discretion. This hard-numbers approach persists despite the fact that no one yet knows what a single CDA discussion will look like, or what range of discretion might be appropriate (or peculiar) to this process.

Bureaucrats should not be required to review their own decisions.

As the members of the community who did make the final call on whether or not an RfA was successful (vote-counting based or not), the 'crats would be placed in an uncomfortable position by the CDA process. If the original promoting 'crat thought his original decision was good, he might engage in public or private lobbying of his fellow 'crats. If the original promoting 'crat felt that the original promotion was in error, his judgement would again risk carrying far more weight than it should.

The 'crats need to be able to work together and to provide a united front to the community; it helps to maintain our trust in their judgement and their ability to fairly close RfAs. CDA will be much more divisive and controversial than RfA. The proposed expansion of powers and responsibilities will encourage greater inter-bureaucrat conflict, and potentially damage the trust in (and reliability of) the RfA process.

Finally, we face the extremely unpleasant prospect of Bureaucrats called before the ArbCom to defend their decisions and participate in appeals. I can see no way that such an eventuality will benefit the project. If people are unhappy with the way CDAs are closed, are we going to have to develop a Community DeBureaucratship protocol?

Simpler approaches have not been tried

The proponents of CDA have spent a substantial amount of time and effort to construct this elaborate proposal, but we are still left with the serious flaws detailed above. More unfortunately, they have rejected any suggestion that their goals could be accomplished by different, simpler means. Indeed, I believe it is possible to achieve the goals of this process without any need to write new policy at all.

An RfC followed by ArbCom motion serves the same purpose, with greater flexibility and resistance to abuse.

This approach has been suggested by Jehochman, among others. It requires no new policy whatsoever.

In a user/admin RfC, one of the statements can simply be "Based on the evidence presented, I do not feel that JoeBlow should remain a Wikipedia admin. ArbCom should withdraw his adminship." or some variant. If this statement is heavily endorsed by the RfC participants, the matter can be presented to ArbCom at the close of the RfC. An RfC isn't on a fixed clock, so there is also the opportunity to suggest and consider alternative remedies. (Note that non-desysopping remedies may not need to be forwarded to ArbCom; they can be enacted after a suitable announcement and brief discussion at WP:AN.)

The RfC should present evidence of misconduct, an opportunity for the admin to respond, the chance for other parties to comment, and a clear statement that the community supports desysopping. If the ArbCom finds the evidence and statements persuasive, it may proceed by motion. The ArbCom can simply desysop, or can impose probation or restrictions it feels will resolve the problem(s) identified. The ArbCom will also be able to sanction other parties whose conduct has been brought up in the RfC, restoring a fundamental balance that is lacking in CDA.

Arbitrators were explicitly selected by the community to resolve these sorts of disputes. They are not self-selected, they are not bound by raw numbers recruited by canvassing, and they are required to recuse themselves where there may otherwise be a perception of bias. This process separates the involved parties from the decision to impose sanctions.

If the ArbCom feels that there are more complex issues that should be examined, they can open a full case. Much of the evidence-gathering and a great deal of discussion will be available from the RfC, so one hopes and expects that the remainder of the case (workshopping and closure) should be able to proceed briskly.

Consensus-driven change to Wikipedia:Administrators can guide Arbitration and other processes.

While some might argue that policy modification violates the 'no new policy required' standard, I offer this alternative as still being substantially more lightweight than a new full-on CDA bureaucracy and framework.

One of the major arguments raised in favor of CDA is the claim that ArbCom are reluctant to desysop admins who have not misused the tools. I do not believe this to be entirely correct, however it is certainly true that bright-line abuses of admin tools tend to lead to desysopping. Our administrator policy explictly lays out the standards of conduct we expect from Wikipedia administrators (Wikipedia:Administrators#Administrator conduct) and notes that consistent failure to meet those standards "may result in the removal of administrator status". If there is a concern that ArbCom are not applying that policy with sufficient enthusiasm, the community can propose changes to emphasize, endorse, enhance, or otherwise strengthen the admin conduct provisions of the policy.

Admins who consistently fail to meet the standards (new or existing, as the case may be) can and should be brought to the attention of ArbCom if other stages of dispute resolution fail to achieve improvement.

We have CDA already: The ArbCom just enacted a rapid, community-driven desysop request.

On 20 January, an issue was raised at AN/I regarding a history of misconduct by a Wikipedia administrator, Craigy144. [3] The misconduct did not seem to involve any use of administrator tools or privileges, however the brief AN/I discussion generally concluded that Cragy144 should not retain his sysop bit. The matter was transferred the same day to ArbCom. The ArbCom declined a full case, opting instead to handle the situation by open motion. After allowing a week for Craigy144 to offer a response or explanation for his conduct, a summary motion was filed on 28 January. The motion passed with unanimous approval on 30 January: [4].

It seems that in situations where an administrator falls seriously short of community standards – even without misusing his admin buttons – and fails to respond to constructive criticism or take steps to remedy his misconduct, the ArbCom is ready, willing, and able to desysop. From filing to desysop was less than two weeks — this process actually ran faster than a CDA would have. The onus is on the community, however, to be clear about what its standards are, and what it considers to be a desysop-worthy problem.