From: [...] Lex Spoon Any organization has some members with clout who can use that more than the formal bylaws would indicate. Going with the reputation approach means that this not only happens, but is primary.
Indeed. And we're sending this email through a server maintained by one of those members with clout, which also hosts the main mailing list, the Web site and a bunch of other stuff. If we're going for the whole 'but I don't trust the roots' thing, shouldn't we also angle to move all of the community's resources to neutral hosting? And who should have access to those? And so on.
This may seem similar to a 'slippery slope' argument (http://www.nizkor.org/features/fallacies/ - a generally useful resource). I'm using it to illustrate a point that we probably have to trust some people, all of the time, with some responsibilities. In the case of the current reputation system, we only need one unbiased root (out of four at present) to ensure that relevant accounts are certified. By contrast, any biased root or (transitively) any of the people they trust, down to the limits of the system, can let in new relevant accounts.
Speaking entirely personally, I'm willing to accept the risk of all of the present roots being biased. I consider the chances of systematic bias across all the roots to be low in practise; I consider the consequence if they are (a reduced number of people being trusted) to be acceptable *because* I would expect systematic bias to be noticed by the community and the alarm to be raised as a result. Thus far I've not seen (to my knowledge) signs of systematic bias, and I've not seen anyone raise the alarm that this *has* happened, merely that this *could* happen. Ultimately, I'm a pragmatist; I consider that "the best is the enemy of the good" and will settle for a system that merely has few flaws.
By contrast, I consider that admitting voters by requiring them to fulfil objective criteria is brittle. The construction of the criteria is paramount. Pretty much whatever we do, I believe we're going to draw relatively broad distinctions with the criteria. As a result, we're going to get some false positive voters and some false negative non-voters if we just use the objective criteria. And if we add in a 'or anyone else the Voter Registration Sub-Committee deems eligible' clause as part of the criteria, then all the focus is on bias within that sub-committee. I believe that given typical human behaviour, this scheme has *more* flaws than using a reputation system - there are more points at which it can break. If we could draw up the criteria perfectly, I'd be happy with it; but I don't believe that we can do that within the time we have.
All personal opinion; please consider it as such! My aim in this message has been to set out why I believe a reputation system is not perfect, but is less bad than a system of objective criteria.
- Peter