[EM] Why I Think Sincere Cycles are Extremely Unlikely in Practice

[Juho]

On Nov 12, 2010, at 9:55 PM, Kevin Venzke wrote:

> --- En date de : Ven 12.11.10, fsimmons at pcc.edu <fsimmons at pcc.edu> a  
> écrit :
>> If sincere cycles are extremely
>> unlikely in practice, then the best Condorcet method is the
>> one that most
>> effectively discourages artificial cycles.
>
> But this is a huge statement. What if the Condorcet method that most
> effectively discourages artificial cycles, also discourages  
> nomination?
> Or gives politically unacceptable results? Or has a clone problem?
>
> What if it were the case that *no* Condorcet method, in practice,  
> suffers
> from artificial cycles?

I agree that performance with sincere votes is also critical. If one  
assumes that in some election strategic voting is marginal, then the  
best Condorcet method is the one that picks the best winner. In this  
situation all the strategy related criteria need not be considered.

On the other hand we know that all Condorcet methods are vulnerable at  
least to the burying strategy. In many/most elections there is no  
working strategy available, and typically the dynamic nature and  
uncertainty in the elections and opinions make efficient and rational  
implementation of strategies impossible. But still there are cases  
where some strategy might improve the utility of the expected outcome  
from the strategists' point of view. In that case successful use of  
some strategy could be more harmful to the society (especially if one  
can show that the sincere winner was changed by a rational strategy to  
a less wanted winner) than the possibility that the election method  
does not pick the best winner (possibly due to using a method that is  
optimized to defend against strategies and not to pick the best  
sincere winner).

In such cases (or if people start using some irrational strategies  
extensively) one should apply sufficient defensive means to make the  
method work in the given (strategic) environment (despite of making  
the performance with sincere votes worse). It is also possible that  
one has to give up use of Condorcet methods altogether if the problems  
are too big.

Unfortunately Condorcet methods have not been used widely in  
(typically political) competitive elections and in societies with lots  
of potential for strategic behaviour. In the less competitive  
environments where Condorcet methods have been used I'm not aware of  
any problems with strategic voting. Condorcet methods are in general  
very good against strategic voting since there are no obvious tricks  
that voters could always or often apply to get "better" results. The  
best strategic recommendation is in most cases (in practice maybe  
close to always) to vote sincerely.

For these reasons I'd expect Condorcet to work very well in most  
societies also in political elections. In order to prepare also for  
the situations where strategic voting does not pose a significant  
threat, also serious discussion on the performance (benefits,  
problems) of different Condorcet methods with sincere votes would be  
very welcome, in addition to the analysis of the different strategic  
weaknesses and strengths of the different Condorcet methods. The point  
thus is that there is a wide range of environments where the election  
methods can be used. In the case of Condorcet methods one might often  
need also the other end, i.e. the one where the voters (or sufficient  
majority of them) tend to vote sincerely.

Juho