[EM] IRV vs Plurality

[Jonathan Lundell]

On Jan 14, 2010, at 9:53 AM, robert bristow-johnson wrote:

> On Jan 14, 2010, at 12:43 PM, Jonathan Lundell wrote:
> 
>> On Jan 14, 2010, at 9:34 AM, robert bristow-johnson wrote:
>> 
>>> do you mean their 2nd choice is not counted because their first choice loses in the final round?  that goes without saying, but that's the dumb IRV rules.  that is an *arbitrary* threshold imposed upon IRV, that 1st choices count and *anything* below the 1st choice makes *no* difference until it gets bumped off and every other choice gets bumped up.
>>> …
>>> but then, one might ask, by what reasonable measure would we eliminate Wright or Kiss over Montroll?  i heard one guy suggest that they should count *both* 1st and 2nd choices for evaluating which candidate is the weakest and eliminated.  but that's just another made-up threshold that someone pulled out of their butt.
>> 
>> Looking only at the current top rank is necessary to preserve later-no-harm. You may not like it, but it's not arbitrary.
> 
> elevating the later-no-harm criterion over the Condorcet criterion (or some other criteria) is arbitrary.  who's idea is it that later-no-harm (to the non-Condorcet winner) is superior to Condorcet?

We know that we can't have a system with all the properties that we might independently desire. Consequently, we compare systems overall, looking not just at their list of properties met and unmet, but at the implications for voter behavior, nomination, campaigns, and so on. 

Those implications have been widely discussed on this list, and I won't try to repeat those discussions. Suffice it to say that to elevate a single criterion, CW or LNH or other, to the sole criterion by which we judge methods just doesn't cut it.

What's not arbitrary, in the case of IRV, is that we only look at top choices. LNH is not, in itself, a self-evident must-have criterion, but some of us are persuaded that the implications of LNH for nomination, candidate and voter behavior make it highly desirable.