[EM] IRV in San Francisco

[Eric Gorr]

At 4:03 PM -0600 11/16/04, Paul Kislanko wrote:
>Eric Gorr replied to my questions:
>>
>>  At 3:14 PM -0600 11/16/04, Paul Kislanko wrote:
>>  >No one can be added or removed
>>  >from a ballot after the votes have been counted,
>>
>>  Sure one can...just do it and recalculate.
>>
>>  >so by this distinction
>>  >there is no such thing as an IIA spoiler.
>>
>>  I believe there is.
>>
>>  Compute the winner.
>>
>>  Start removing candidates and see if you can get the winner to change.
>>
>  > If the winner changes, you have a spoiler via IIA.
>
>That's what I said below - this is an analytical tool to determine how a
>method works logically.

So, you now believe there is such a thing as an IIA spoiler?

>  > >ICC is just a weaker version of IIA, and has to be weaker
>>  because nothing
>>  >can pass IIA and meet three other desirable criteria.
>>
>  > What ICC and IIA are you using?
>
>Whichever ones you are.

Not necessarily.

There are multiple _VERSION_ of the criteria, which is why I asked.

>It wasn't ME that claimed that the difference was in
>ICC the candidates were already on the ballot, but to be precise we note
>that in your reply you said that in IIA they were on the ballot and you just
>remove them and add them back in.

You misunderstand, the ICC spoiler comes from neither removing nor 
adding candidates - they are simple there being a spoiler and one can 
see them as a spoiler without taking any further action.

For example, if a set of clones can obtain > 50% of the first place 
votes, but no candidate in the clone set wins, they are spoilers.

This, as near as I can determine, has nothing to do with IIA.

>  > On what basis do you call the ICC a weaker version of IIA?
>>
>>  Does your statement imply that if a method passes Dr. Arrow's IIA, it
>>  must pass Dr. Tidemans ICC?
>
>I make no such claim,

You clearly stated:

   ICC is just a weaker version of IIA

So, one what basis do you make that claim?

The claim is false, if you are using the non-resolute versions of IIA 
& ICC, as Markus previously showed.

>If ICC is related to IIA and a
>method can meet Arrow's other three and also meet ICC, then per force ICC
>must be a weaker criterion than IIA, else there would be a flaw in Arrow's
>proof.

Since we do not know if the resolute versions of ICC and IIA are 
directly related to one another, this statement is currently 
meaningless.

However, a claim has been make that if a ranked ballot method passes 
Dr Arrow's IIA it must pass Dr. Tideman's ICC which I believe would 
mean there is a direct relationship between them and that ICC is a 
weaker version of IIA since ranked ballot election methods can pass 
ICC, but not IIA.