[EM] Fluffy the Dog and group strategy

[Richard Moore]

Craig Layton wrote:

 >>Well said. As long as the method allows the voters to
 >>express their preferences honestly (and feel comfortable
 >>doing so), and counts those preferences in a reasonable
 >>manner, how can you blame the method for making the choices
 >>the voters (collectively) tell it to make?
 >>
 >
 >
 > Let me leave fluffy aside.  I don't have my original 
example but I'll
 > provide a simple one;
 >
 > 49.2% A>B>C (sincere utilities 100>30>0)
 > 49.3% C>B>A (sincere utilities 100>30>0)
 > 00.5% B>A>C (sincere utilities 100>10>0)
 >
 > When all of the voters have expressed their preferences 
honestly, and 99.5%
 > would prefer a random dice throw between the three 
candidates over the
 > actual result, then I think this indicates that there is 
a problem with the
 > method.  The fact that this will happen in the best 
constrained preferential
 > voting methods doesn't ameliorate the fact that it is 
still a significant
 > problem.

You omitted my last two paragraphs, which gave my two 
*qualitative* conditions for a good method, so I'll repeat 
them here:

"This means that, for me, a good method is one that (a)
encourages honest expression of preferences on the ballots
(whether it allows full or partial expression is secondary)
and (b) doesn't generate unpleasant surprises in the result.
Condition (a) asks for FBC and perhaps a few other
strategy-related criteria, and condition (b) asks for
monotonicity and consistency."

"That Approval satisfies both (a) and (b) makes it a very
strong method in my book. Condorcet comes pretty darn close."

Now let me ask: Is condition (a) met by Condorcet in this
example? I would have to say it is. Is condition (b) met by
Condorcet in this example? Most people, I think, would say 
no, and that B's win is an "unpleasant surprise".

Note that I only said Condorcet comes close to meeting both
conditions (meaning it won't always do so, but it will in a 
vast majority of real-world cases), so this new example 
doesn't contradict my statement.

Approval would most likely elect C (depending on strategy
assumptions). While this is still going to be  "unpleasant" 
for many, it is certainly less of a "surprise" than the 
Condorcet winner. Condorcet isn't sensitive to the strengths 
of preferences; Approval *is*. Approval also quantizes that 
information, but as I said I think that is a secondary 
consideration.

Regardless of whether the method is sensitive to strengths
of preferences, we still don't know (from the numbers) if
the low-utility candidate is an unpopular centrist or just
somebody's dog. Election methods won't tell us things 
outside the scope of election methods.

In this case the real problem (for the voters, not for the 
mathematicians) is that we need better candidates. But we 
*do* need to use methods that will satisfy condition (b) 
when there *are* candidates that have sufficient support 
from sufficient numbers.

With a good basic method, additional rules such as Demo's 
ACMA (?) can be added to correct for occassional failures 
due to lack of any candidate who is acceptable to sufficient 
numbers (forcing a new election). I consider these rules to 
be supplementary to the method itself; i.e., there is a 
higher level process that says "Hold an election using 
method X; then if the winner meets acceptability criterion Y 
then stop; else, select new candidates and repeat." Method X 
and criterion Y should be chosen according to needs.

I'll ignore the missing 1% in this new example. Maybe there
were hanging chads.

Richard