Wikipedia:Tool apprenticeship

Tool apprenticeship is a proposed supplement to the current Wikipedia:Requests for adminship (RfA). Currently it is being proposed to run on a trial basis (details below).

In tool apprenticeship, a user who has a need for a particular administrator tool or tools, such as deletion or protection, makes a request to receive that tool. The user is judged according the the following criteria:

• The user must be in good standing;
• The user must have an ongoing need for the tool and be able to begin using it right away;
• To the greatest extent possible, the user should be active and have sufficient experience in the area in which they plan to use the tool.

Satisfying these and following a consensus discussion, they receive the tool on a trial basis for a limited period (currently one week). When this period expires, the tool is automatically revoked. The trial period is subject to probation (tool revocation in case of misuse), and may involve voluntary mentoring.

After the end of their trial, the user can file a request for a new trial (also of one week), based on their performance during the trial period, which will be granted if the user substantially used the tool and exercised good judgement. If the request is denied, the user will be given extensive feedback on their usage and will have to demonstrate through their actions a clear understanding of that feedback before requesting a new trial.

Request for trials would be subject to a 7-day consensus discussion at Wikipedia:Requests for tool apprenticeship. Successful requests are closed by a bureaucrat, while an unsuccessful request can be closed by any experienced user in good standing. This discussion would be short and straightforward in most cases (similar to AfD) because less scrutiny is needed than for a full RfA: it's only granting a subset of tools for a limited period subject to probation. Irrelevant discussions that are not related to the tool or area under request are discouraged and may be hidden if they occur. "Conditional support" opinions, where a user agrees to support if the candidate agrees to their terms, help to avoid polarization.

Objectives and comparison to RfA

The primary objectives of tool apprenticeship are:

• First and foremost, to get important administrative work done on the project, reducing backlogs and freeing up administrator time to work on other important tasks;
• To allow users to learn to use administrative tools in compliance with policy, via hands-on experience and regular feedback;
• Facilitating a more informed evaluation of candidates at Wikipedia:Requests for adminship by allowing candidates to demonstrate real experience with administrative tools in a meaningful but restricted setting.

Compared to the RfA process, tool apprenticeship:

• Only gives tools to users who need them for the types of tasks they perform.
• Involves much less discussion of requests - it is expected most requests will be quickly approved or denied with little discussion.
• Gives users feedback on an ongoing basis.
• Reduces the stress of requesting tools, since discussion is more limited and far less scrutiny is needed for a temporary trial of a subset of tools.

On the other hand, tool apprenticeship is intentionally limited: it cannot assign all tools, cannot assign tools permanently, and cannot assign tools without the restriction of probation. These are reserved to RfA. Apprentices are also not permitted to perform actions restricted by policy to only be performed by administrators, such as closing contentious deletion discussions.

Proposed process trial

Currently tool apprenticeship is being proposed to run on a trial basis for three months, subject to the following restrictions:

• It will run alongside the existing RfA process (see below);
• Tool trial duration will be limited to one week;
• Tools that can be granted will be limited to at most two of: delete, protect, block, editprotected;
• Users can only do one trial at a time (may involve multiple tools, but only for a single task);
• Tools will be granted to at most 10 users at one time (in total). If there are 10 active apprentices, no more requests will be closed until one of their trials expires.

This process trial will enable us to monitor and evaluate the process and monitor the affected users. A backlog of requests is expected during the trial. Tools are not yet being assigned permanently so that the effects of the trial do not outlast the trial, but when users return to request a new trial they will have their actions reviewed in the same manner.

In the event that many apprentices abuse their privileges in such a way as to cause widespread damage, and/or the system is ineffective at preventing or limiting damage, the process trial may be terminated prematurely.

After the trial

One week before the end of the trial, bureaucrats will immediately cease to close requests or give out any tools. Because anyone who received a tool during the trial will have the tool expire one week after they received it, all tools given during the trial will be revoked by the end. The trial will not last longer than 3 months, and will not be extended. A report will be commissioned from those who monitored the apprentices during the trial detailing problems, confusion, and successes. The process will undergo revision and discussion to address problems, and the next stage of evaluation (or full deployment) will later be conducted pending another Requests for Comments and community consensus.

Does this supplant or supplement RfA?

This proposal is a supplement to RfA. Users may choose freely which process to pursue at a given time. Some users may choose to only acquire a few tools to do specialized work in a particular area without ever pursuing adminship, while other users may first acquire a few tools through the tool apprenticeship process and, after acquiring sufficient experience, later request adminship. This would substantially improve the ability of voters at RfA to evaluate their ability to use dangerous tools responsibly. As noted above, only RfA can grant tools permanently or grant the full collection of tools.

Probation

Apprentices receive tools under probation, meaning that serious misuse can lead to immediate tool revocation. Misuses can be reported at Wikipedia:Bureaucrats' noticeboard, and a bureaucrat will examine the situation and remove the tool if justified. The user can later issue a new request, where discussants will evaluate whether they have overcome the issue leading to revocation.

Example scenario

Suppose User:NewPagePatroller is a new page patroller who has for some time been appropriately tagging new articles for speedy deletion. They request the deletion tool to allow them to speedy delete new articles. The request is granted for a period of one week, and they immediately begin using it during their patrolling. After one week, they make a new request for another one-week trial. Their deletion log is reviewed and it is determined that their deletions were responsible and consistently in line with Wikipedia:Criteria for speedy deletion, so the request is granted.

At the same time, another request is made by User:VandalFighter. User:VandalFighter has a history of giving correct and timely warnings to vandals, suggesting appropriate protect and block actions on talk pages and WP:AIV, and in some cases has even reformed vandals who expressed regret. They are requesting the protect and block tools to assist them in their vandal fighting. The request is granted, again for a trial period of one week. However, User:VandalFighter becomes overzealous and makes some blocks that are too long or too broad according to the Wikipedia:Blocking policy, blocks that are later modified by others. They later request another one-week trial, but this request is denied, and their errors are explained to them. They spend another month discussing more blocks on WP:AIV, and begin to demonstrate a good awareness for appropriate blocks. The next time they make a request for a one week trial for the protect and block tools, the request is approved.

Responses to anticipated objections

Some tasks, such as vandal patrolling, require multiple tools and it doesn't make sense to request them one by one.

• If multiple tools are genuinely required for a single task, a user can request multiple tools at once. They can expect to face further scrutiny when requesting more tools, or more dangerous tools, but because it is still a trial, the scrutiny would be less than for a full RfA.

If we trust a user enough to give them one tool, we trust them enough to give them all of them.

• This is a common fallacy. Trust in one area does not necessarily transfer into other areas (consider the aphorism "I trust my bank with my money and my brother with my children, but not vice versa"). Trust with using tools should be based primarily on how a person uses those tools, which depends in turn on their experience with the tools and/or the tasks in which they are applied, which varies widely from one tool to another. Someone can be a trustworthy person and still not be ready to capably use every tool. This is already reflected in our practice: users of rollback, although they may be trustworthy people, may not pass RfA, for a variety of reasons; and likewise administrators are not guaranteed to pass a request for checkuser or oversight rights, although they have been granted many other tools.

If tools are granted so easily, doesn't it invite abuse?

• A user who is on their trial is also considered on probation and may have the tool revoked for blatant misuse. Such a user is less likely to be granted a trial in the future, or may be granted a trial subject to a shorter time period or mandatory mentoring. Additionally, the restriction to a subset of tools and the automatic time limit limits the scope of the damage a user abusing a tool can cause before their actions are reviewed.

What if a user makes a mistake and can't fix it because they don't have the proper tool? For example, restoring an article they deleted by accident?

• A user can request assistance from others for tools they don't have access to, just as regular users do. In the case of a simple accident, such a request is expected to be processed quickly and with little overhead.

How will users in search of assistance find users who have the tools needed to help them?

• A simple category system on user pages could be used to group users according to the tools they possess (or the software could be augmented to do this automatically). As an additional benefit, tool apprenticeship allows users in search of aid to distinguish users who are inexperienced with tools (on their trial) from those who have more experience.

Won't this just lead to more bickering over which tools each person should get?

• No. A person requests specific tools, one or two at a time, for a specific purpose. The discussants do not have the ability to grant tools that were not requested, or to grant full administrator rights (only RfA can do that). If a person requests a trial with multiple tools, they are accepted or rejected as a group, since presumably not having all the tools they need would be useless (unless it turns out that some of them are not needed for the task at hand).

What about overspecialization? Some issues crossing areas require several tools to deal with. The overhead of coordinating response to these among multiple users would be costly.

• It is true that granting less tools results in more coordination overhead for certain tasks. However, overall efficiency is expected to improve for three reasons: 1. most work is confined to limited areas using a subset of tools; 2. the overall number of partially-empowered users is expected to be larger than the number of administrators, so they can absorb the communication overhead; 3. RfA will continue to supply admins who can work on this type of cross-cutting problem.

The overhead of processing tool requests may be costly.

• Most of the overhead of processing requests (e.g. reviewing logs, weighing various concerns) is done by the discussants in request discussions, who do not require any special abilities, only a basic understanding of the process and requirements (and such users are plentiful). Likewise, unsuccessful requests can be closed by any experienced user in good standing. If there are too many successful requests for bureaucrats to close, more bureaucrats can be recruited for this purpose. The 10-user limitation during the process trial implies that the bureaucrat load will be low.

What about viewing deleted revisions? It's impossible to review use of that tool, and it has privacy concerns.

• This tool won't be available during the trial, postponing this question. In the future, software changes might allow auditing of viewing deleted revisions.

If this process becomes established, RfA will make it a requirement or it could become a de facto requirement, making it more work to pass RfA.

• While the proposer of this process would oppose any attempt to make apprenticeship an RfA requirement, it might occur. However, this is just one of many unnecessary requirements that may be levied against RfA candidates, and addressing these systematically is outside the scope of this proposal. To make an analogy, it would be senseless to eliminate the featured article process (which gets important work done on articles) simply because promoting featured articles has become a de facto RfA requirement.

Experience at tool apprenticeship will be given too much weight during subsequent RfAs. Mistakes will be seized upon, and problematic users with clean apprentice records will pass too easily.

• While tool apprenticeship is expected to be considered at RfA, it should not be given excessive weight, and good candidates who learn from their mistakes should pass. As above, this is just one of many factors that may be given excessive weight by the RfA process, and systematically addressing this poor decision making is outside the scope of this proposal.

Questions about design choices

Why is mandatory mentoring of apprentices not a requirement?

• Mentoring of every participant would be ideal, but the workload for mentors would severely limit the number of participants, and apprentices may resent mandatory mentoring. The probation and review system provides an incentive for apprentices to voluntarily seek out mentors and ensures that their use of the tool will be reviewed, during follow-up request discussions. This component may be amenable to change.

Will there be specific criteria for users to make a request, like a required number of edits?

• The consensus discussion that processes requests is expected to take factors like these into account. Such criteria can be developed later if the volume of requests becomes too large to manage. By that time, the outcome of past requests will provide guidance for what effective criteria should be.

What is needed to implement it?

Only simple configuration changes in LocalSettings.php would be required to enable assigning individual tools to users (creating one user group for each user right; for the trial, only four groups would have to be created). Technical changes to the MediaWiki software would be required to make them automatically expire after the trial period. Pending such changes, bureaucrats could implement this mechanism manually. If for whatever reason bureaucrats fail to remove the tool by the end of the apprentice's trial period, the user will be instructed to cease using the tool or else be blocked until the tool is removed, which can be enforced by any administrator.