I also think even an experiment can be dangerous, because really, there's no such thing as an experiment here. [ ] And the only way to stop the experiment would be for Jimbo to make a very divisive decision to stop it.
Suggestions:
The experiment could run on meta. I expect less people will get emotionally attached there, given the nature of its content.
Perhaps a more refined scheme might lessen the confusion about what a person's rating is based on. People might want to compliment one another for sheer number of contributions, reliability of content/factual knowledge, courtesy/diplomatic efforts. An expert who writes only highly valued articles in his/her particular field of knowledge might score low on etiquette.
Doubts:
I'm not sure whether it will work (ref wikimoney) or whether it is a good idea anyway. It might lead to a meritocracy where people with a high rating throw their high wikistatus into an argument, or may even unintentionally intimidate others with a different opinion, where now the merit of each edit is judged on its own.
To avoid confusion and lessen the meritocratic side effects, the experiment might be confined to the latter criterium: personal behaviour/wiki etiquette. This is probably what Jimbo was thinking of in the first place, since he gave potential usability for (de)sysoping as a possible benefit. Then again, when a debate gets heated people will have stronger opinions about each other. Will it be helpful when people resort to the statistical equivalent of name calling? Will people with a negative account become stigmatized, meaning their future actions will be prejudged based upon their etiquette rating?
Erik Zachte