[EM] RE : Re: RE : Hay Voting

Kevin Venzke stepjak at yahoo.fr
Sat Jan 20 09:25:26 PST 2007


--- Peter de Blanc <peter at spaceandgames.com> a écrit :
> On Sat, 2007-01-20 at 17:19 +0100, Kevin Venzke wrote:
> > What I understand from this is that you will draw one ballot, but may
> > still pick any candidate as the winner. What prevents the voter from
> > just revealing his first preference and assigning nothing to anyone
> else?
> It's not possible to assign all of your voting mass to one candidate.
> Your ballot starts out with all of the voting mass already allocated
> evenly between candidates. You spend your voting credit to buy
> "transfers" of voting mass from one candidate to another. With N
> candidates, there are N choose 2 possible pairs. You pay voting credit
> to transfer a certain amount of voting mass between each pair.
> The more you spend on one particular pair, the less you're getting per
> marginal voting credit (the cost is the square of the mass transferred).
> So in your optimal strategy, whenever you prefer a candidate A over a
> candidate B, you will spend some credit transferring voting mass from B
> to A.
> In fact, this pricing rule is such that your optimal strategy (in
> expected utility terms) is to make the size of each transfer
> proportional to the difference in utility between the two candidates.
> You could spend all of your credit in order to double the voting mass
> allocated to your favorite candidate, but this is not the most efficient
> use of your voting credit because you would be able to transfer *more*
> mass if you spent some of your credit on other transfers.

I will say it's pretty impressive if you're truly created a method
under which a voter is bribed to reveal lower preferences (and accurate
utilities) in exchange for more voting power in favor of his top 

> Deterministic Hay Voting has the same allocation rules as standard Hay
> Voting, except that the candidate with the most mass is always elected.
> This changes the incentive structure of the voting system, so voters
> will no longer vote honestly. However, I was wondering if it still might
> be possible to recover the voters' true utility functions by analyzing
> their votes. We can pay the voters (in votes) to perform some of this
> analysis for us (ie showing us their probability distribution for how
> the other voters will vote).

I suspect this (which sounds like a two-part ballot) is more work than
would by justified by what you could get out of it. Losing the ability
to trust that voters' orderings of the candidates are sincere, is 
too big of a problem...

Kevin Venzke


Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! 
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses 

More information about the Election-Methods mailing list