[EM] Districting Quests: Quixotic vs Practical

[Ernest Prabhakar]

Hi Joe,

On Jan 21, 2004, at 7:05 PM, Joe Weinstein wrote:
> In my opinion, this absolutist goal is anyhow overly ambitious, and is 
> not required for a practical acceptable process of districting.  To my 
> thinking, Altman’s extended argument against HIS picture of ‘automated 
> districting’ just beats a long-dead horse.  Whether or not districting 
> is ‘automated’, we never need nor expect to get a provably optimal 
> plan, nor even a provably near-optimal plan, in an absolute sense, 
> i.e. considering ALL possible plans.
>
> Instead all we really need is RELATIVE optimality:  a districting 
> process which is open to submission of sufficiently many - but not 
> impossibly many - independently crafted well-defined plans, among 
> which the finally chosen plan P1 must be provably optimal or 
> near-optimal.

I agree with you that absolute optimality is silly.  The only real 
important question in my perspective is whether we can define an 
objective 'figure of merit' Q, such a plan with a large Q is always 
better than one with a smaller Q.  If we can do that, then there's 
numerous techniques for easily maximizing Q (I use simulated annealing 
in mine). Of course, those are local maxima, since it is not possible 
in general to prove a global maxima (as Altmans rightly, if 
irrelevantly, points  out). However, in most cases it also should be 
trivial to calculate Q for each of them, no matter how many different 
submissions are being evaluated.

-- Ernie P.