[EM] strategy and method complexity and the advantage of minmax methods

[Juho Laatu]

Hello Anthony,

On Jun 7, 2005, at 08:06, Anthony Duff wrote:

> The pertinent question is whether people here have wildly exaggerated
> the importance of strategic voting, and whether simple minmax
> methods, such as PC or MMPO are good enough.

This is a good question. Strategic voting may be a big risk if there 
are easily implementable strategies that make someone with clearly 
lower utility than the best candidates to win. But in cases where the 
risks are marginal and strategies practically impossible to apply, it 
may be better to focus on other aspects of the voting method (like 
simplicity, understandability and picking the candidate that has the 
best utility value (as determined by the community for this election)).

I think the discussion at this mailing list focuses quite heavily on 
the strategic vulnerabilities of different methods and on how to fight 
against them. Very good. But of course one should every now and then 
also look at the complete setting from a practical viewpoint, i.e. what 
and how big the risks are in actual elections (e.g. in large public 
elections).

> I believe that the difficulty in organising the public at large into
> voting insincere preferences so as to generate an insincere cycle is
> too great to be realistic.

Hope so. Maybe someone here has a good collection of bad examples of 
artificial loops and other strategy problems in practical elections 
(with estimates on how probable they are in real life).

It would be also interesting to get thorough analysis of different 
strategies with respect to how much information of the expected votes 
is needed, possible need to coordinate large number of voters, 
probability of successful outcome, probability of causing harm to 
oneself when applying the strategy, probability of people understanding 
when and how to apply the strategy etc.

> It is not in the nature of
> legislators to refer to external definitions, such as your own.  They
> are going to reword your definitions, as they understand them.

They might also trust a uniform voting method science community telling 
them that some certain method is the best one. This is however maybe 
the biggest problem of the Condorcet community - no agreement on which 
method is the best. Legislators might btw be happy also with rewording 
only the surroundings and leave the (too mathematical) core of the 
method intact :-).

> I think that the principle of condorcet, of a full pairwise analysis,
> is simple enough for most people to appreciate, and that abandoning
> the pairwise principle of analysis in order to perform the rare
> completion is not simple enough for most people.

I support electing the simplest method and most natural utility 
function (for the election in question) that is still sufficiently 
strategy free. But I guess people on this list will have very different 
opinions on what that method would be (and which threats/strategies are 
the important ones). For many it surely must be the ultimate most 
strategy resistant method.

Best Regards,
Juho


P.S. I'm also still a bit unclear if people think that all single 
winner elections should use the same (their favourite) voting method or 
if they think there are different needs (utilities) and different 
voting methods correspondingly.