[EM] I eat my words (but not wholly) 1

[Kevin Venzke]

Warren,

--- Warren Smith <wds at math.temple.edu> a écrit :
> But I do not buy some of his abuse-o-gram.

You think my message was an "abuse-o-gram"?

>  Fact is, a lot of voters want to express
> maximum information (natural human drive) and do not want to truncate ballots.
> Also in a lot of natural situations voters feel they DO know about eveyr candidate in the
> race.  You argue it is  "joke" that I could imagine that.  I disagree.

Even if it's true (that all voters can sincerely rank all the candidates), it
is not so meaningful under methods like MDDA and ICA, since truncation has specific
meaning in these methods. Voters, and candidates, will have to understand what
method is being used.

And even if the voters still do rank all the candidates under MDDA or ICA, at
least they will have the sense not to rank the last candidate, in which case
the approval scores will mirror the last-preference scores, which are not likely
to be tied.

> Also, I do not buy the abuse of my "DH3 pathology".   The strategies he considers
> "stupid" for voters are not stupid.  They can in fact be "strategically forced"
> under the right assumptions about what the others are going to do.  Far as I can see.

Why don't you show that, under plain ICA or MDDA, voters can be "strategically
forced" to bury the worse frontrunners beneath dark horses? I don't really believe
you can do this, and this is why I say DH3 deals with what stupid voters can do
to themselves. Your use of the word "pathology" gives me this impression also.

> And I still do not understand some of this jive about "risk free burial"
> and still woudl like explicit election examples so I can understand.
> It it too handwavy at the moment.

I should say, it's not entirely risk-free (using an explicit cutoff, I mean).
But it's still not very good, when it can be made virtually always too risky.

On a philosophical level: ICA and MDDA disqualify candidates who have pairwise
losses. Then approval decides among those left. When you can place your
approval cutoff wherever you want, you can rank A above B for the sole purpose
of disqualifying B. You're not obligated to approve A. The only risk of electing
A is if the B voters decide to rank A above your guy (C) for the sole purpose
of disqualifying C. But as long as they don't do this, you can use candidate A
in this way, and it will either move the win from B to C or do nothing.

Example:

35 C|A>B
40 B|
25 A|

Under MDDA, C wins. DH3 voting (burying) pays off.

Under ICA, the approval winner could attempt to steal the election from the
Condorcet winner:

40 A>B|C
35 B|C>A (insincere)
25 C|

B wins. If the A voters don't actually approve B, then A wins and the burial
was harmless. If the B voters believe that the A voters won't vote A>C>B, then
this is a good way for the B voters to vote.

So I don't suggest a placeable cutoff for these methods.

Kevin Venzke



	

	
		
___________________________________________________________________________ 
Appel audio GRATUIT partout dans le monde avec le nouveau Yahoo! Messenger 
Téléchargez cette version sur http://fr.messenger.yahoo.com