[EM] Combined elections

[Juho]

On Dec 4, 2009, at 6:48 PM, Raph Frank wrote:

> For example, if there was a tie
>
> A>B>C>A
>
> and only B and C were found, the B wins.

Yes. There are also voters/parties that support alternative A more  
than B. They have an interest to find the A solution.

Hopefully already the official process will find A with high  
probability (the conditions and preference rules should be simple  
enough to allow this) and the private calculations would be done just  
to double check, to see that there is no fraud and to check that the  
official process is efficient and correct.

(And of course in simple enough elections we can use algorithms that  
will find the optimal solution with certainty.)

>>> I think it might have some strategy problems.
>>
>> Maybe inherited from the ratings side?? What would that be?
>
> Not 100% sure :), it might be OK since condorcet is used.

The last phase was Condorcet based and in principle voters could even  
give directly ratings to each winner combination. The method should  
from this point of view be as strategy free as Condorcet methods are.  
The summing procedure (=combination preferences counted from  
individual candidate preferences) could introduce some more details to  
this but so far I'm not aware of anything interesting.

> If a majority really hate an option, then all wining sets with that
> option will have a really low rating.

Yes, as in Condorcet really bad alternatives will be eliminated with  
very high probability even in the presence of strategies. (maybe one  
can say "with certainty" if there are no strategies)

Juho