[EM] non-deterministic methods

[Brian Olson]

On Dec 27, 2004, at 4:26 PM, Forest Simmons wrote:

> As Jobst recently pointed out, non-deterministic methods have not been 
> adequately studied or promoted, considereing their potential 
> contribution to fairness and to strategy free voting.

They may be mathematically fair, but I find them philosophically 
unsatisfying. What's the point in voting if the result will come down 
to a random process? (I know, to increase the expected likelihood of my 
choice winning.) I suppose on the "devil I don't know" argument, a 
non-deterministic method may at least be semi-immune to electoral 
malfeasance, except that they'll probably find some way to load the 
dice and make sure the house always wins.

No, election methods are fundamentally a decision process based on 
votes, not random processes. In certain other applications, possibly in 
computer decision making and artificial intelligence, this could be a 
useful tool.

> But the determinists would say that they should be even less equal: A 
> should get one hundred percent of the probability.

Put another way, I can turn any non-deterministic procedure into an 
approximately deterministic one.
Run the ND system a million times, elect the most probable winner. If 
you've gone and added non-determinism to some method, we've now wasted 
a million minus one election calculations.

Brian Olson
http://bolson.org/