[EM] IRV in San Francisco

[Bart Ingles]

Eric Gorr wrote:

> At 8:16 PM -0800 11/14/04, Bart Ingles wrote:
> 
>> Eric Gorr wrote:
>>
>>> At 7:44 AM -0800 11/12/04, Justin Sampson wrote:
>>>
>>>> On Fri, 12 Nov 2004, Eric Gorr wrote:
>>>>
>>>>  > Well, it will cause IRV to fail the Independence of Clones 
>>>> Criterion and
>>>>
>>>>>  thereby be subject to a spoiler effect again.
>>>>
>>>>
>>>>
>>>> Doesn't IRV suffer from spoiler effects anyway?
>>>
>>>
>>>
>>> Depends.
>>>
>>> The method itself passes the ICC, so spoilers cannot come from there.
>>
>>
>> Take a hypothetical IRV variant where candidates are eliminated at 
>> random, until a "majority" winner emerges.  Would this method still 
>> pass ICC?
> 
> I cannot think of a case where random elimination of a candidate tied 
> for least votes would cause IRV to fail ICC.

Actually, I wasn't thinking of ties, or of a system that would actually 
be used.  I was thinking in terms of how far you could distort the 
method and still pass ICC.  In other words, completely replace the usual 
"fewest first-choice votes" elimination rule with random elimination, or 
even eliminating the candidate with the most first-choice votes short of 
an outright majority.  As far as I can tell, such a system would pass 
Mutual Majority; I just wondered if it would pass ICC.


>> If so, is the method spoiler free?
> 
> Depends on whether one wants to consider spoilers in the context of IIA.
> 
> There is always the possibility that spoilers can come from directions 
> other then ICC and IIA.

What would be an example of a spoiler (ICC or other violation) which is 
NOT an irrelevant alternative?

Bart