Applying those criteria

[Blake Cretney]

On Sun, 18 Oct 1998 02:20:02   Mike Ositoff wrote:
>
>David--
>
>As I said, though I hadn't stated it as a criterion, the
>1st Choice Criterion, which says that no one should have to
>not rank their favorite in 1st place, is, it seems to me,
>the main goal of single-winner reform. Considering how many
>people listed sincere voting as a most important standard,
>I believe that many agree with me on that.
>
>So which methods meet that criterion?
>
>If it means that we should never have defensive strategic reason
>to not rank our favorite alone in 1st place, over everything else,
>then the criterion is met only by VA (for all pracatical purposes).
>Not by any other method. Certainly not by Margins or IRO.

45 A C B
18 B C A
17 B A C
20 C B A

   A   B   C
A  X   45  62
B  55  X   35
C  38  65  X
A wins.

But if the C voters vote

45 A C B
18 B C A
17 B A C
20 B C A  -- insincere
   A   B   C
A  X   45  62
B  55  X   55
C  38  45  X

So, the C first voters had a clear incentive not to vote themselves
alone in first place.  In fact, I don't think any ranked method,
or for that matter any method that truly has a "first place", 
fully meets your criteria.

So why does Mike feel justified in saying that VA meets
your criteria, for "all practical purposes".  Mike's pronouncements
usually have a large number of unstated assumptions.  For example,
he often says that a voter never needs to engage in a particular 
strategy, when he actually means as long as the voter's favorite 
candidate is the CW.  Of course, voters are unlikely to see it this
way.  They want their candidates to win whether they are CW or not.

As well, he tends to assume that order-reversal will happen a lot in
Margins, but never in VA.  This is because of his defensive truncation
strategy.  I don't agree that this is a practical strategy, or one
likely to be used.  We have exactly the opposite positions on
the defensive order-reversal strategy.  If you want a full discussion
on this issue, you should check past postings in this list.

Many of Mike's comments are based on the fact that VA is more truncation
resistant than Margins.  This is quite true.  If you remember how
Condorcet methods work, a candidate wins not as much by maximizing the
size of its victories, as by minimizing the size of its defeats.
When someone votes   A > B = C, they are helping A by increasing the
chance and size of B and C losses vs. A.  However, they are missing
the chance to increase a loss for either B or C.  So, in VA, a more
sensible vote would be either A > B > C or A > C > B.  Since in VA,
you don't have to worry that this may decrease the size of a loss, 
like in Margins.

So, VA reduces the effect of truncation by punishing all voters who
rank equally at the end of the ballot.  Hopefully, from this perspective,
voters will rank candidates equally anyway, unaware that they are
being punished for it.  Once they realize that random filling is the
better way, the jig will be up.  Mike believes this isn't a problem 
because it won't defeat a CW.

>And it's half-met by Approval, since at least one can vote
>one's favorite as high as anyone can vote anything--no need
>to ever vote anything over your favorite.

You can eliminate the need for any strategy by restricting the voters
expressed preferences enough.  I don't consider this a triumph.  If
you only have two rankings, it doesn't make sense to rank a less
favorite over a more favorite.  If you penalize equal rankings at
the end of the ballot, truncation becomes a poor strategy.

Remember that when Mike says something about Margins vs. VA, he may
have a number of implicit assumptions.  You can't just take it at
face value.  If you agree with all his assumptions, that's fine.  
But if you don't, you should check his statements for yourself.

Blake



-----== Sent via Deja News, The Discussion Network ==-----
http://www.dejanews.com/  Easy access to 50,000+ discussion forums