[EM] Rob: Condorcet's Criterion vs FBC. Will people favorite-bury?

[Abd ul-Rahman Lomax]

At 12:31 AM 10/4/2005, Rob Lanphier wrote:
>Normalization only works if there's not a third candidate forcing the
>distortion (e.g. candidate "C"):

That is correct. Range Voting requires the same kind of strategic 
consideration as Approval Voting, a vote is diluted if cast at less than 100%

>90 voters: A=10, B=9, C=0
>10 voters: B=10, A=0, C=whatever
>
>A: 900
>B: 910
>
>Once again B wins, but this time, the presence of information about C
>prevents normalization.  In essence, normalized range violates the
>Independence from Irrelevant Alternatives Criterion (IIAC).

Yes. Normalization only increases the relative strength so that the 
voters' vote is not diluted and becomes equivalent to a plurality 
vote for the favorite(s). (This presumes that the favorite has the 
highest range vote.)

There is, I think, a common misconception about Range, perhaps 
because of the association with sports scoring.

A sincere Range Vote is not an opinion about the candidate, properly. 
It is an expression of the voter's desire, or lack of same, to avoid 
disappointment by the election of an unacceptable candidate, while, 
at the same time, to maximize the possibility of the election of a favorite.

This requires, in Range, the use of standard Approval strategy, which 
is to give the maximum rating to the favorite, and if the favorite is 
not a front-runner, to the preferred frontrunner, and to all 
candidates equally acceptable or preferred to the front-runner. 
Failing to follow this strategy is an act by the voter which weakens 
the voter's vote. If the strategy is followed, the voter is then free 
to give weaker ratings to other candidates, especially those whom the 
voter would not terribly mind seeing elected.

The allegedly shocking example was one where the great majority of 
voters did not follow this strategy, at least not with the reported 
votes. It is correct that there might have been other votes not shown 
because they were "irrelevant," perhaps those majority voters who 
preferred A to B by a one-point margin actually did not have A as 
favorites. In this case, yes, normalization would not have helped. It 
would help in the case where voters really think all the candidates 
poor and want to express that, but still want to influence the 
outcome of the election. Or else they could simply have stayed home.

In that example, the A voters expressed a weak preference. Thus it is 
utterly unsurprising that B could win through other voters expressing 
a strong preference. As Mr. Smith might point out, this is a feature, 
not a bug. If you care strongly, express a strong preference! Range 
allows you to do that, it does not *require* you to do that, as does 
plurality and basic Approval, with binary preference allowed only.

Range is an Approval Method, and one should vote it as Approval or 
suffer a dilution of one's vote. Range *allows* the voter to make 
votes indicating a weak preference. That is an increase in the 
freedom of the voter.

It makes it appear that the method with normalization suffers from a 
failure to satisfy IIAC. Normalization considers the overall voting 
pattern, so other votes are *not* irrelevant in the same way as they 
would be in a ranked system.