[EM] Kevin, you wrote on 4 Oct '03

[Kevin Venzke]

Dave,

 --- Dave Ketchum <davek at clarityconnect.com> a écrit : 
> Let's do it slowly:

Strange that you assume I didn't understand your point.  You said you think
fractional votes are "reasonable."  I only said I thought they're "undesirable."
Then, in this message, you say you "make no sense" of my point.  So isn't it me 
who might want to explain things more slowly?

> By counting half a vote each for Tom's support for A and for B, he is 
> applying as much power as other voters are allowed to apply when they are 
> not doing equals.
> 
> I thought this would come thru clearer by noting that Tom and Dick each 
> voting A=B, should and would have the same power as for them to get 
> together and have one vote A>B while the other votes B>A.

This notion of "power" depends on IRV ideas.

This is like claiming that voters have unequal power in Condorcet, because
some voters will make more ranking distinctions than others.  One would have to
have Approval in mind to think something like that.

Or I could claim IRV has unequal voter power if voters are allowed to truncate,
because exhausted ballots have to be disregarded.

> > I see this scoring as undesirable, because by tying too many candidates in first,
> > one can cause one or more of them to be eliminated.
> 
> I make no sense of this, for the whole point to the election is to 
> eliminate all but one winning candidate.

I meant that a PARTICULAR VOTER could harm the result of the election, from his 
perspective, by tying candidates.

Fractional votes would not completely fix IRV's problem with order-reversal 
incentive, whereas whole votes would.  This is called "Weak FBC," meaning you should 
never have a reason to rank your favorite candidate below a preferred candidate.

That was my point, so you can address it, or not.

>       In basic IRV each voter is applying one unit of power to helping 
> elect one candidate.
>       In this extension the voters using it are helping two candidates, 
> and therefore should be giving each half as much help (and, if you allow 
> A=B=C, then the power should be 1/3 for each

I understand your reasoning.

I'm only saying this isn't the only way to go.  You can allow full votes for every
tied candidate without giving more "power" to certain voters.

If you don't think so, I suggest you try to create a scenario where "whole votes"
leads to an obviously wrong winner.  Now I know you are going to try:

55 A=B
8 B
37 C

B must win by any standard, but it's possible that A is a sincere majority favorite.

I agree that this is a bad outcome.  But do the 55 voters really have "more power"?
One might say that, by making fewer distinctions, they actually have LESS power.

> >>Here if Tom votes A=B, he has as much power as if he became twins, voting 
> >>A>B and B>A, giving these two an edge over other candidates.
> > 
> > This is not really so, since all voters can do this.  And if the other voters
> > bullet vote, they are denying more candidates votes as surely as Tom as giving
> > more of them votes.  That is no different from Approval, I don't think.
> 
> True that all voters could do equals, and thus have equal power.  My point 
> is that voter power should be equal when some are rating their first 
> choice candidates equal and others are not.

My point is that the IRV notion of power isn't the only one that makes sense.

> A voter ranking less than all the candidates, up to and including bullet 
> voting, has no effect on their power in IRV until/unless all of the 
> candidates they rank lose, at which time they drop out of the remainder of 
> the election.

Of course, this would still be true whether we use fractional or whole votes.

> We are still not going near Approval, for that only hears do vs don't 
> approve.  With ranking we can say best, almost best, etc.

I don't understand your point here.  How does the number of possible ranks affect
equal ranking's effect on voter power?


Kevin Venzke
stepjak at yahoo.fr


___________________________________________________________
Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com