[EM] Reply to Rob regarding RCV

[Kristofer Munsterhjelm]

On 2023-09-23 21:34, Michael Ossipoff wrote:
> I mean the former. The people who will always vote a deplorable sleazy 
> corrupt POS “lesser” evil over what we’d all like, against the hopes of 
> nearly all of us, because they’re being routinely threatened with a 
> greater evil.

Then I'd like to note that the "only" if not an iff: IRV can still have 
trouble when voters aren't timid overcompromisers.

Consider Burlington. It's reasonable to assume that the 
Wright>Montroll>Kiss voters weren't compromising; if they had 
compromised for Montroll, then IRV would have elected the CW. But they 
didn't, so it didn't.

In a way, Condorcet auto-compromises for coherent majorities: suppose 
the electorate knew who would win with the original method. Then if 
there existed a sincere CW who was not elected, a majority could have 
compromised for the CW by ranking him first, thus making him win. In 
that sense, Condorcet is good for voters who *don't* compromise.

Now one could say that voters who don't compromise don't care about 
compromise incentive, because they will never take advantage of it no 
matter how high it is. But tactical voting incentives can also indicate 
that the method is getting the honest outcome wrong.

E.g. Plurality has lots of compromise incentive. Suppose you have an 
idealized electorate of voters who think that compromising is inherently 
wrong and swear not to use it. Plurality would still be a bad method; it 
would still get 2D spatial elections like the Tennessee example wrong.

-km