[EM] Fair and Democratic versus Majority Rules

[Kristofer Munsterhjelm]

fsimmons at pcc.edu wrote:
> How did this thread get side tracked to Proportional Representation?
> 
> Proportional Representation only works for multi-winner elections.
> 
> Of course, everybody knows that PR is the way to go in multi-winner elections. 
> And why is that?  Because it solves the tyranny of the majority problem in that
> setting.  
> 
> So why cannot we see that this same problem exists in single winner elections? I
> suggest that the analogous remedy in the single winner setting is proportional
> probability.  The simplest method that accomplishes this is random ballot.  But,
> as I suggested, there are better stochastic methods that yield probability
> distributions with less entropy while still solving the tyranny of the majority
> problem..

An assembly picked by PR stabilizes itself because the differently 
positioned candidates balance each other out and actually meet and 
discuss. If the assembly rules have supermajority rules, that may also 
make it more likely to reach a consensus rather than oscillating between 
extreme positions.

On the other hand, in a single-winner election, there is only one 
winner. That winner usually won't have an "inner assembly" to balance 
himself. Thus, I think that the reaction against nondeterministic 
methods arise from the thought that if we can't be sure who will win, 
the candidate who wins might win simply by the luck of the draw and be 
unsuitable - and then we're stuck with him without other council members 
to moderate him.

Abd's 10% example is an example of this. If 10% think [disastrous 
policy] is really good and votes accordingly, then by using Random 
Ballot, you'd get that disastrous policy 10% of the time. While Random 
Ballot may find a brilliant policy that only 10% knows of, it can't 
discern between that and a horrible policy that (rightly) no more than 
10% support, and the loss from the latter would more than outweigh the 
gain from the former.

Or so one would argue.

In a low-entropy method, as far as I understand it, there still has to 
be a random component that encourages the voters to find a compromise. 
If the fallback component is too close to a deterministic method, then 
the majority won't care to try and find a consensus because they'll 
benefit more from the fallback. On the other hand, if it's too random, 
it could be bad indeed if the different groups truly can't find a 
compromise.