[EM] RE : Re: Range voting, zero-info strategy simulation

[Kevin Venzke]

Hi,

--- Abd ul-Rahman Lomax <abd at lomaxdesign.com> a écrit :
> At 04:42 PM 10/30/2006, Kevin Venzke wrote:
> >I wrote a simulation to measure the utility of zero-info Approval
> >strategy in zero-info Range elections.
> 
> Range Voting is designed to maximize overall utility to society, not 
> *expected utility* for the individual voter. 

I'm not sure why this would be important when determining individual
strategies under Range.

> This study seems to 
> replicate the common criticism of Range Voting, which is that voters 
> may attempt to maximize their personal expected benefit by voting 
> strategically, in this case by voting Approval style, with a 
> particular strategy:
>
> >These results suggest to me that in the zero-info case, if there are
> >thought to be more than about 2 other voters, Range should be voted
> >as in Approval.
> 
> This is utterly unsurprising. Venzke has simply confirmed the 
> obvious.

I suspected the outcome I got as well. That doesn't make the experiment
worthless.

> >D. Zero-info Approval strategy. The voter gives a 10 to every candidate
> >at least as good as the average value of all candidates, and gives a
> >0 to the others.
> 
> This is truly zero-info strategy. You don't have to have the foggiest 
> notion of what everyone else thinks about the candidates. Useful to 
> people who've been living in ... gosh, can't think of any place where 
> media don't exist. Perhaps you have been on a round-the-world cruise 
> on a balsa wood boat and your radio broke down.
> 
> This imaginary voter has somehow managed to determine values for the 
> candidates, *all of them*, while remaining utterly innocent of how 
> everyone else feels. If the election of one of the candidates would 
> cause a revolution, he hasn't got the foggiest idea. I find this 
> assumption internally contradictory.

What is the point of this? An unrealistically ignorant voter does not
have to actually exist for it to be true that this voter would already
have a viable strategy.

> It is thus a bit like the assumption that voters will vote 
> strategically because (1) their sincere preference is A slightly 
> preferred to B, and (2) they want A to win so much that they will 
> distort the expression of their true preferences in order to defeat 
> B. They both don't care much and care very much at the same time. And 
> it is precisely this thinking which is behind this most common 
> assertion that Range is "vulnerable to" or "encourages" strategic voting.

If we're talking about ideal strategy, if the voter only slightly 
prefers A to B but is giving A the top score and B the minimum score, 
this would have to mean that A and B are probably the only serious 
candidates.

> Of course, the statement that the voter "should" vote Approval was 
> made without stating the criteria which would determine how voters 
> "should" vote.

My assumption is that the voter, at the time of voting, has a ranking
of candidates in mind such that the first candidate in this ranking
is the candidate that this voter would most like to see elected. And
the last candidate in this ranking would be the candidate that this
voter would *least* like to see elected. Etc.

If at the time of voting the voter *does not know* the identity of
the candidate he would most prefer to have win, then indeed the way
one votes in an Approval election might not be something that can be
translated for a Range election.

> A Range election is designed to maximize collective value. If voters 
> want to generate the greatest good for the greatest number, they will 
> sincerely express what the election of each candidate means to 
> *them*. If they, instead, effectively lie to the system in order to 
> gain some expected extra personal benefit, they may gain a small 
> increase in *expected* value, but with two costs: (1) they may lose 
> value if their insincere vote results in the loss of their second 
> favorite, for example, because they devalued this candidate in order 
> to maximize their own favorite candidate's winning. 

That is not a cost, since a sincere vote runs the same risk. And 
according to you it is "utterly unsurprising" and "obvious" that this
risk is *higher* with a sincere vote than even a zero-info strategic
vote.

> and (2) they risk 
> winning at the cost of collective dissatisfaction with the election, 
> which can easily harm the "winners," in more than one way.

Well, I would have more confidence in incentives that are built into
the method.

> I think we need to be very careful about assuming that voters will 
> commonly distort their votes in an attempt to manipulate the system 
> to produce the election of their favorite. 

I don't understand this choice of words. It seems to me that if we
want to be "very careful" then we should assume that voters will be
trying relentlessly to exploit every mechanism available.

> The idea that this is a defect in Range is caused by continued and 
> often unexamined assumptions that voting systems are supposed to 
> elect the pairwise favorite, the Condorcet winner, if one exists. 
> This, again, is rooted in an utter neglect of preference strength, 
> because the Condorcet winner could be far from being the winner who 
> will maximize benefit. I have given examples showing how very bad 
> choices can be made by assuming that a whim is equivalent to a strong 
> preference.
>
> Since these analysts presume that the pairwise ranking winner should 
> win, they think of strategic voting as "bad," and "not to be 
> rewarded." It is old thinking.

I'd rather say that these analysts think of strategic voting as bad,
and that is why they think the pairwise winner should win. Electing
the pairwise winner eliminates a lot of need for strategy.

> The strategic voter -- which in this 
> context means a voter who lies about true preferences in order to 
> gain some immediate advantage -- may somewhat increase his immediate 
> satisfaction by bullet-voting or Approval style voting, with some 
> strategy as was described -- but only when large numbers of other 
> voters don't have a strong preference regarding the pairwise election 
> in question.

Personally I appreciate the need to harness voter strategy under 
Approval and Range. I think it makes good methods out of them.

But I should say that I find Approval's strategy guarantees to be
quite good (FBC, MD as I define it; an Approval ballot filled out
based on one's gut is probably already the optimal vote).

> Basically, if you care enough to lie to the system, 
> well, you care enough to outweigh the mild preferences of other 
> voters and your alleged mild preference wasn't.
> 
> The paradox behind this, I don't think has been noticed by other 
> writers before. I'd appreciate comment on it, particularly from Mr. 
> Smith, the Range Voting expert.

Comment:
It assumes that it requires more than a mild preference to be willing
to take full advantage of the method when other voters (voluntarily)
are not.

If everyone else is choosing not to eat the cake, I'll go ahead and
eat the whole thing; doesn't mean my heart was set on doing so.

> I think that real Range data from real Range elections will show that 
> this alleged "defect" of Range Voting does not exist, or, more 
> accurately, that it does little or no harm.
> 
> If the majority allegedly deprived of their allegedly rightful 
> victory -- by the election of their second favorite due to "betrayal" 
> by the strategic voters -- really care about this, if it is a serious 
> loss to them, well, then, it is also true that their alleged mild 
> preference was really stronger than stated in all these scenarios.

My concern is more that people who want to vote solely in their own
interests might not realize how to do so effectively. I like a method
to be rather foolproof in this respect.

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 
http://fr.answers.yahoo.com