[EM] 3 or more choices - Condorcet

[Juho Laatu]

On 3.10.2012, at 3.35, Michael Ossipoff wrote:

> On Tue, Oct 2, 2012 at 12:55 AM, Juho Laatu <juho4880 at yahoo.co.uk> wrote:
>> On 2.10.2012, at 4.37, Michael Ossipoff wrote:
>> 
>>> A) What is it that is gained by using traditional (unimproved)
>>> Condorcet instead of Symmetrical Improved Condorcet?
>>> 
>>> The downsides of unimproved are:
>>> 
>>> .....1. FBC failure (though unimproved-Condorcet advocates speculate
>>> that people won't mind)
>> 
>> The traditional interpretation of ranked votes may well support FBC "well enough" in the classical Condorcet methods.
> 
> Yes, that's what I meant by "speculation".

It is well known that it is impossible to meet all ciriteria at the same time. Typical election methods can't  be made fully strategy free. Therefore the best approach must be to meet some criteria not 100% but "well enough".

> 
> And note that that speculation is coming from someone who likewise
> accepts Margins' failure of the Plurality Criterion. How often Margins
> fails Plurality isn't the issue. The examples in which it does will be
> widely shared with the public, by the opponents of any enactment
> proposal for any method that violates that criterion (or any other
> embarrassment-criterion).
> 
> 
> Speculation that voters won't be affected by a strategy-incentive is
> one form that unimproved Condorcet rationalization and self-deception
> can take.
> 
> But my purpose, at this stage of the procedure isn't to argue about
> the criteria and properties, so much as to establish what unimproved
> Condorcet advocates think make unimproved Condorcet so good as to
> outweigh the disadvantages that I listed.
> 
> By the way, of course it's better to call a method by the name used by
> its advocates. We have a term for the more EM-popular (unimproved)
> Condorcet versions: "Strong Condorcet". That's the term that I'll
> start using. It refers to Beatpath, Ranked-Pairs, River, Goldfish,
> Kemeny, and VoteFair. Maybe other similar methods too. Sometimes I'll
> abbreviate "Strong Condorcet" to "Strong".

I'd recommend to make the separation by using your new term "improved Condorcet" (or maybe some more descriptive name) and leave the traditional Condorcet approach as it is and how people usually refer to it, i.e. without any additional name when one refers to the basic rankings. Or maybe use generic words like "traditional" or "regular" when needed. I think you may be about to break your own good rule of using names that the advocates of the named object use.

> 
>> 
>>> .....2. Interpretation of equal-top and equal-bottom ranking is
>>> contrary to the voter's preferences, intent and wishes.
>> 
>> I guess by default the meaning of equal-top and equal-bottom ranking is to rank the candidates equal.
> 
> Wrong.
> 
> Ranking X and Y in 1st place, or ranking W and Z in last place
> additionally means that that voter prefers that the winner be X or Y,
> or that the voter would rather that the winner be someone else other
> than W or Z.

That's true already without any additions.

Juho


> 
> And, you see, that's where unimproved Condorcet --excuse me, I meant
> "Strong Condorcet"--parts ways with what the voter prefers.
> 
> I've explained this before, here at EM. I'll explain it again if requested.
> 
>> The voter may have interest to cast a stronger vote where the equal-top and equal-bottom rankings have some additional strength
> 
> ...such as being interpreted and counted in keeping with that voter's
> preferences, intent and wishes? Yes.
> 
>> , but that's another story
> 
> Indeed,  in Strong, that is indeed another story. Strong has its own
> story, and it isn't about what the voter actually prefers.
> 
> 
>> , and not the default interpretation of ranked votes.
> 
> You mean _your_ default interpretation of ranked votes. From what you
> said, your "default" interpretation of equal top ranking disregards
> the voter's preference for the top-ranked candidates over the other
> candidates. Indeed, that is a different story.
> 
> Mike Ossipoff
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info