User:Timotheus Canens/On sanctions

The Arbitration Committee authorizes administrators to impose discretionary sanctions in the most problematic and contentious topic areas on Wikipedia. Unfortunately, figuring out what sanction to impose is not an exact science. It would have been nice if we could simply plug a set of factors into a set of rules and get the magical "best sanction". Alas, there is no such thing. This page attempts to outline the factors I consider and do not consider in coming up with what I think is an appropriate sanctions. Other admins may well disagree.

The preliminary question: authority
One basic principle is that administrators have no independent authority to impose discretionary sanctions; any such authority must come from explicit authorization by community consensus or by arbcom. Therefore, sanctions can be imposed only if any prerequisite conditions are satisfied; most of time, all that is needed is a prior warning, but sometimes a more specific warning may be required.

However, in some cases of serious disruption, a block may be appropriate under the blocking policy independent of discretionary sanctions. Those blocks do not require a special warning, since they are not placed under the authority derived from the discretionary sanctions provisions (and, for the same reason, are not subject to the usual protections attached to AE actions), although it is often advisable to deliver the discretionary sanctions warning when imposing the block.

The factors
The fundamental goal of any sanction is to "ensure the smooth functioning of the project". To that end, sanctions should be sufficient, but no greater than necessary, to end misbehavior and prevent future disruption.

I find the mantra of "preventative not punitive" to be not very helpful, for there is no bright-line distinction between the two. It's not black and white. Even a purportedly punitive sanction is most of the time preventative in the sense that it deters future misbehavior, both by the sanctioned user, and by other parties.

Somewhat counterintuitively, this is the factor that actually determines the starting point for a sanction, because it is the most objective factor. An occasional slip-up may only warrant a short sanction or even just a warning. Chronic misbehavior means that an extended break or restriction is necessary.
 * The starting point - history of recent sanctions, and past sanctions for the same conduct

I normally do not consider sanctions that have expired more than a year ago, except to show a pattern with more recently imposed sanctions. When an editor has behaved reasonably well for a full year, I don't think it's appropriate to treat them as different from any other editor simply because of something that happened a long time ago.

If the user has been sanctioned previously for similar misconduct, I normally use twice the length of the most recent previous sanction as the starting point, since the previous sanction has failed to curb the misconduct, and a stronger one is needed.

Otherwise, I use the length of the most recent sanction, if there is one - if that sanction was successful in holding back misbehavior, at least for a time, then it has a good chance of being successful again here, too.

If there is no sanction in history, then I usually use 31 hours for blocks and 3 months for other, non-blocking, restrictions. Blocks are good for violations of long-term restrictions, which are used in turn for continuing misconduct. A one-off incident normally need no more than a warning.

This is very hard to quantify. The following is a non-exhaustive list of sub-factors that I normally consider:
 * Nature of misconduct
 * 1) The willfulness of misconduct. Willful misconduct (where the user violates a restriction knowing that the act is a violation) suggests that a stronger sanction is needed to induce a change in behavior.
 * 2) The disruptiveness of the misconduct. Highly disruptive misconduct needs a stronger deterrent.
 * 3) Whether the user was baited. If a user "explodes" when baited, that's certainly not ideal, but it is much better than a user who "explodes" spontaneously. In those cases, a lesser sanction may well be sufficient to prevent future misconduct.
 * 4) Whether the user recognizes the inappropriateness of the conduct. If they do, and commit to avoiding such misconduct in the future, then a lesser sanction, or sometimes no sanction at all, may well be all that is needed. This is not absolute, though.

These factors are either already considered in another form above, or should not (in my opinion) be considered at all.
 * What I usually do not consider
 * 1) The "goodness" ("non-disruptiveness") of the edit. Misconduct is considered to be disruptive per se, and if the edit is particularly disruptive, it is a reason to increase the sanctions from the starting point. The lack of particular disruptiveness does not therefore require a reduction.
 * 2) The "correctness" of the edit. It's dangerous to allow questions of content to influence conduct decisions. Allowing sanctions to be based on whether the admin agrees with the editor on the content question is the fastest way to disaster, especially in highly charged topic areas.
 * Note, however, that edits that are unambiguous violations of our content policies may suggest that a lengthy break is needed to protect the project.
 * 1) "Good work" by the sanctioned editor. Good work does not excuse misconduct. However, it may be appropriate to design the sanction in a manner that would minimize the collateral damage.
 * 2) The action of another editor. Except for baiting or similar cases, the action of user B generally have no bearing on the severity of misconduct of user A. Misconduct by B is grounds for sanctioning B, not refraining from sanctioning A.

But it's all black magic, really
The above, I hope, provides some insight into my thought process. However, it is not, and does not purport to be, complete. It is difficult, if not impossible, to articulate every single thing I consider when I come up with a sanction, and there are enough subjective factors in there to make the outcome different from person to person even under the simplified scheme outlined above. Perhaps it's all just black magic, after all.