This conjoined twin of a feature request was surgically separated. The other portion of the request is now found at: Fine grained suspension
Please give the existing answers a bit of slack for addressing portions of this request that have been moved.
tl;dr version: Rather than locking people out of the house for increasing periods of time, wouldn't it be better if we had a robot that detected regular patterns of abuse and performed the banning automatically? This would be perceived as "fair", as opposed to some assertions that our current banning process is too opaque and subjective.
A significant problem that suspended users and their supporters complain of frequently is that the current ban system is not fair (being singled out, rules aren't stated, here are examples of others doing the same thing, etc)
While I hate to open old wounds, and there is an unwritten social agreement to avoid mentioning specific people who have been banned (Primarily for privacy reasons), I feel it's important that we consider some concrete examples that did happen, rather than postulate based on what could or might happen.
Threatening emails from Jeff Atwood <-- User asked privately to change their comment posting behavior to avoid suspension.
http://stackoverflow.com/users/5640/geoffrey-chetwood <-- User suspended due to numerous complaints and flags against user's comments and other interactions on the site.
Numerous other users have been banned at various times for various reasons, including extraordinary users (one of the top superusers, for instance, and that is a site that could still use a little more help). Would these cases have been better if we could remove one or another ability, rather than removing them from the site completely?
Automate the suspension system based on flags and moderator attention
- While users are told that they are being banned for specific violations, there appears to be a lot of wriggle room for the suspension trigger to occur. It would be better if, over time as problems are identified, the system is designed to detect inordinate numbers of flagging and moderator actions against a specific user for specific violations. Triggers should be set that provide the moderators/team with notices so that they can perform suspension based on a well known set of conditions. As long as the system is automated to at least this degree, then people cannot easily complain about how the system is unfairly targeting them. Further, the moderators then have clear guidelines as to what constitutes suspendable violations for some types of regular infractions.
This feature should also reduce moderator workload as no warnings would need to be given to users who receive so many flags so consistently that they lose the ability to comment, so the whole email communication aspect could be removed, and they could be encouraged instead to come here for support regarding semi-automated suspension. It may reduce the need for moderators to discuss specific cases and determine the appropriate action, and suspensions can be enacted immediately, rather than a day or more after the latest incident.
- Are there behavior patterns that we could detect that are currently not detected?
- Is it reasonable to formalize this process so once a trigger occurs it is acted upon, rather than starting discussion, or deferring the decision to Stack Exchange, Inc?
- Would this actually resolve the problem, or would banned users still be able to complain about the unfair nature of it due to some other link in the chain?