[EM] Practical Democrach

[Kristofer Munsterhjelm]

On 01/25/2016 08:56 PM, Andrew Myers wrote:
>> So, the "tl;dr" version is roughly: Voters get together in groups of
>> 3, choose the best of the 3 to represent Them at the next stage,
>> selected Representatives then lather and rinse and repeat, yes?
> 
> This seems like a terrible system that will lead to tyranny. In a system
> with n levels, you need only something like (2/3)^n of the leaf voters
> on your side to win. With 7 levels that's 6%.

You can do so if you know who those leaf voters are going to be. I made
the same argument when replying to Fred way back when he first
introduced the system. However, the voting groups are picked at random,
so it'll be pretty hard to know which 6% you need to corrupt ahead of time.

My concern with the method is more that its recursive nature could cause
a concentration of power. In a simpler "three people at random choose
one among themselves, rinse and repeat" method, the concentration effect
excludes minority points of view while amplifying majorities. That is,
if I'm not mistaken, why Fred added point 1b to the procedure: the idea
is that a group of people with views held by say, 10% in general, could
declare to be a group of its own and run the procedure separately. The
final level (council) would end up having at least one representative of
the group instead of that group being whittled down into nothingness by
participating directly in the unaffiliated process.

I'm still a bit wary regarding the concentration of power, however.
Groups that are more like parties would have a coordination advantage
compared to those that are less like parties, particularly in the later
rounds. Something similar to your deliberate arranging might be possible
by statistical circumstance alone, as well: suppose that 20% of the
representatives at the next to last level are aligned with party X and
have an implicit agreement that whenever there are more than one X
member in a triad, they'll always vote for one of the X members. Then it
might end up with council representation exceeding 20% because there
aren't enough triads to go around to evenly distribute the X members.
Small towns could also have a concentration problem like that. However,
finding out whether that is indeed a serious issue would probably take
some simulation.