[EM] Re: Range Voting

[bql at bolson.org]

On Sun, 2 Jan 2005, Kevin Venzke wrote:

> Brian,
>
> --- Brian Olson <bql at bolson.org> a écrit :
>>>> If a majority of all the voters prefer X to Y, then they should have
>>>> a way of voting that ensures that Y won't win, without any member of
>>>> that majority voting a less-liked candidate over a more-liked one.
>>
>> I think IRNR has this. I think IRNR is a strict improvement on straight
>> Cardinal Rating summation, except for the computational burden.
>
> I'm not so sure about this... What if X is the majority faction's fifth
> choice? It is easy to imagine X being eliminated in this case, only a
> little harder to imagine Y going on to win. Maybe that can't happen,
> though.

The rule asserts that there is a relatively honest vote _possible_ to 
create a certain outcome. There are also votes possible that fail to 
generate that outcome.

IRNR elects choice F for the following, and indeed E is the first one 
eliminated. And if these are honest ratings/utilities then I clam that F 
is the correct choice.

*51 1.0, 0.9, 0.8, 0.7, 0.6, 0.5
*49 0.0, 0.0, 0.0, 0.0, 0.0, 1.0
http://betterpolls.com/et?vrr=-r&if=-f&cand=6&seats=1&data=*51+1.0%2C+0.9%2C+0.8%2C+0.7%2C+0.6%2C+0.5%0D%0A*49+0.0%2C+0.0%2C+0.0%2C+0.0%2C+0.0%2C+1.0

But, the majority _can_ veto choice F by the following, where A is chosen 
by IRNR:

*51 1.0, 0.9, 0.8, 0.7, 0.6, 0.0
*49 0.0, 0.0, 0.0, 0.0, 0.0, 1.0

http://betterpolls.com/et?vrr=-r&if=-f&cand=6&seats=1&data=*51+1.0%2C+0.9%2C+0.8%2C+0.7%2C+0.6%2C+0.0%0D%0A*49+0.0%2C+0.0%2C+0.0%2C+0.0%2C+0.0%2C+1.0

> IRNR is one of the methods I've programmed using Forest's information
> on producing triangles to draw all the results for a method using three
> ballot types. One alarming thing I noticed is that IRNR doesn't always
> elect a majority favorite. Suppose the ballot types are A, B, and C>B.
> Even if the C>B faction makes up a majority, they can get C eliminated by
> giving B too high of a score.

Yes, and I claim that is a correct result. Ratings based election methods 
are more capable of finding good compromise candidates. I think a lot of 
talk of rankings assumes that there is a great difference in utility 
between every step from 1st to 2nd and so on. IRV/STV certainly assumes 
this by putting your whole vote on your first choice and only if you can't 
have that transferring your whole vote to your second choice (and so on). 
But with ratings there is more information available, and it might turn 
out that I have two or three close favorites (say, Cobb, Nader and Kerry 
or Badnarik, Buchannon and Bush). I can rate those six candidates 
something like: 1.0, 0.9, 0.8, 0.2, 0.1, 0.0 . That means it would be 
pretty much OK with me to go with my 2nd or 3rd choices.

I think it's a better result, for optimizing social utility, to go with 
the second choice of the majority that mostly satisfies 70-80% of the 
people than to just go with the favorite of 51%.

Brian Olson
http://bolson.org/