[EM] Learning from IRV's success

[Kevin Venzke]

Hi,

--- En date de : Ven 8.7.11, Kristofer Munsterhjelm <km_elmet at lavabit.com> a écrit :
> Bob Richard wrote:
> > It turns that real live voters (including real live
> politicians) care a lot about the later-no-harm criterion,
> even if they don't know what it's called.
> > 
> 
> If true, that is unfortunate. Perhaps we would have to pick
> a better criterion that is also easy to understand,
> something like the (weak) Favorite Betrayal Criterion. But
> if we have to do that, then a lot of otherwise good methods
> go out the window.

Well, you can modify MinMax to satisfy FBC. But, you have to figure out
what to do when it returns more than one unbeaten candidate. First-
preference count is the most obvious general fix. Probably violates
Plurality. "Total support" would be better but brings in an additional
concept...

Yes, I think that if FPP didn't have such strong nomination 
disincentive, voters would be more concerned about the principle of not
having to lie about who their favorite is. As things are, who other than
a "third-party extremist" would ever need to worry about it?

It's too bad there is no conclusive, persuasive way to measure LNHarm or
FBC performance. There's certainly better and worse among methods that
fail.

Kevin Venzke