[EM] IFNOP Method (was Re: Question about Condorcet methods)

[Dave Ketchum]

On Thu, 19 Oct 2006 06:43:37 -0400 raphfrk at netscape.net wrote:
> From: davek at clarityconnect.com
> 
>  > Given: 
>  >   35 A>C 
>  >   33 B>C 
>  >   32 C 
>  >
>  > I see: 
>  >  IRV will discard C, letting A win. 
>  >  Condorcet will see 65 C winning over 35 A 
>  >
> 
> The point is that the later no harm property means that there is no
> incentive to truncate your ballot.

There is a limit to adding new complications to an admittedly contrived 
ballot (what we often do to make a point).

Condorcet and IRV usually agree as to winner.  My final vote described C 
being liked MUCH better than A - a nearly 2:1 preference.

If a couple B voters had stayed home, IRV would have seen the C 
preference.  Since their presence made IRV declare A (who was least liked 
by them) the winner, they could join the crowd demanding Condorcet.

I started with 35 A - which gave the same results of the 2:1 preference 
since Condorcet would have seen A>B>C>A, a cycle in which A>B and B>C are 
near ties while C>A is stronger.
> 
> If A's supporters changed to
> 
> 35 A
> 33 B>c
> 32 C
> 
> The result becomes
> 
> A>B:  35 > 33
> A<C:  35 < 65
> B>C:  33 > 32
> 
> Thus, A supporters have shifted the vote from a C win to a tie (which
> is an improvement, unless B wins the tie-break).  Thus, it is in their
> interests to truncate, if their ranking is A>>>C>B.

I mentioned this vote above - that it does not matter here what the A 
voters do about truncating.

If enough B voters truncated, A would be the agreed winner.  It is their 
indicated preference of C over A that Condorcet sees in declaring a win for C.
> 
> The end result may be that voters will start truncating/bullet
> voting, getting us back to plurality.

The ability to rank has MUCH value - it permits ranking TRUE first choice 
as first, while usually protecting against worst lemons winning.  Even 
with the ability, voters will often find bullet voting to fully express 
their desires - forcing such voters to express more opinions than they 
have can easily provide useless noise.
> 
>  > I LIKE Condorcet. What does this mean below?
> 
> I was saying that to preserve the later no harm property, you must do
> something like IRV.

If "later no harm" is an argument against reading all that the voter says, 
then I DO NOT see it as a valuable property.
> 
> This results in a loss of access to some of the ranking information, but
>  at least it doesn't create an incentive to not rank candidates.

As some of the above shows, IRV often ignores ranking.  Condorcet always 
reads it, sometimes thus more accurately responding to voter desires.
> 
> The other option is to make it compulsory to rank all candidates.

TWO results:
      Annoy all the voters.
      Get only destructive noise when demanding more expression by voters 
than some of them have any interest in expressing.
> 
> Actually, would randomly completing uncompleted ballots create an
> incentive to fully rank all the candidates even in condorcet ?
> 
Another way to introduce destructive noise from a combination of voter 
response and machine generated noise.
>  >
>  > DWK 
>  
>  
> Raphfrk
-- 
  davek at clarityconnect.com    people.clarityconnect.com/webpages3/davek
  Dave Ketchum   108 Halstead Ave, Owego, NY  13827-1708   607-687-5026
            Do to no one what you would not want done to you.
                  If you want peace, work for justice.