User:Bovlb/Karmic voting

This brief essay is written as a reaction to some recent discussions proposing changes to the process of granting administrator rights. Herein I propose a significant change to the way that Wikipedia grants user rights. This proposal not only makes radical changes to community processes, but also require significant software enhancements; I therefore acknowledge that it is unlikely ever to happen.

One of the issues with the current RFA process (and the non-existent process for administrator recall) is that votes happen rarely, within a narrow time window for each editor, and in an ill-frequented part of the project. Another is that voting is done in the form of a discussion, and can become quite heated. This proposal makes voting commonplace, unrestricted in time, part of the regular interface, and not directly discussable.

There is a new user rights group "voter", members of which are allowed to assess other editors. The assessment of an editor by a voter is a simple (unranked) "positive" or "negative". Assessments are valid for some length of time, say one year. Assessments can be renewed, reversed, or retracted at any time. Assessments are accompanied by an optional log message, and the logs are public.

It is possible to see the total number of (extant) positive and negative assessments of an editor at any time. A gadget might be able to colour or otherwise indicate the assessment status of editors in log entries and signatures, both with respect to the logged-in voter and the community. Another gadget might indicate your own assessment status at the top of the screen, with highlighting of changes.

Making an assessment is a subjective matter that is up to each individual voter, but in general it is trying to answer the question, "Does this editor make a net positive or negative contribution to the encyclopedia?". The community might develop more extensive guidance on how to assess. An explicit property of this system is that it does not distinguish between or impose a weighting on different types of competence (for example, content creation, template editing, bot creation, wiki-gnoming, conflict resolution), nor does it attempt to determine whether an editor is acting in good faith.

Assessment might be performed as a normal part of reverting, new page patrol, recent change patrol, vandal patrol, reviewing article history, interacting on article talk pages, and reading Wikipedia noticeboards. There is a risk that certain types of contribution might be more likely than others to attract assessment because of their inherent visibility, or that errors will attract more negative assessments than good work attracts positive. This could be mitigated by replacing the (somewhat defunct) editor review process with a request for assessment.

Certain user rights (e.g. voter, rollbacker, reviewer, autopatrolled, autoconfirmed) are associated with a threshold test in terms of positive (P) and negative (N) assessments. For example, autoconfirmed might be assigned when P > 1 and P/(P+N) > 50%; the voter right might be assigned when P > 100 and P/(P+N) > 75%. To reduce thrashing, there is another, lower threshold for losing a status (a form of hysteresis). The community will probably want to allow administrators to override the results of this voting, but this should be applied only in exceptional cases.

Similarly, there could be a (fairly high) threshold for editors to be eligible to be administrators, but it would probably be best to retain a human gateway (e.g. a bureaucrat) in the system to catch any problems. Similarly, there could be a lower threshold at which administrators might be recalled. It would not be appropriate to block users automatically on the basis of negative assessments, but administrators would be able to access lists of editors accumulating negative assessments to see whether they arise from actions that require administrator intervention. IP users are not assigned user rights, so voting on IPs is only useful for such administrator review. It might be especially useful when considering rangeblocks.

The voter group is intended to be composed of those editors who have a thorough understanding of Wikipedia's fundamentals, and what it takes to make an encyclopedia. To bootstrap this process, the right might initially be assigned to administrators, but it would quickly spread beyond that (and perhaps exclude some administrators!).