[EM] Responses to some of Forest's ideas

[Richard Moore]

Forest Simmons wrote:

> 
> On Tue, 10 Jul 2001, Rob LeGrand wrote:
> 
> 
>>Forest wrote:
>>
>>>3. It eliminates the occasional low utility Condorcet Winner.
>>>
>>I question the advantage of this . . . advantage.  ("Allow myself to introduce
>>. . . myself.")
>>
> 
> 
> Here's an example of this (zero info):
> 
> 45 A >> B > C
> 45 C >> B > A
> 2  C > A >> B
> 4  B > A >> C
> 4  B > C >> A


One thing to remember is that no election method can 
distinguish between the case where A and C are extremists 
and B is a moderate compromise (the electorate being highly 
polarized), and the case where A and C are the moderate 
representatives of opposing viewpoints and B is the village 
idiot with 8 close friends or relatives.

In the latter case, the advantage of eliminating the LUCW
is clear, but it's not such an advantage in the former case.


> Does it bother you that after the results are announced, the first faction
> (A's supporters) might regret not having (insincerely) approved B ?


I think minimizing regret is an important theme, perhaps not 
so much in picking between methods as in setting individual 
strategies. Certainly it is important in Approval voting if 
one is not using a purely numerical method (1): If you 
prefer X > Y > Z then vote XY if and only if you would 
regret Y beating out X by one vote (yours) less than you 
would regret Z beating out Y by one vote. Since we are told 
this is a zero-info case, I think members of the first 
faction should be happy with their vote being the best vote 
possible under those circumstances.

Richard

Note 1: Of course, the numerical solution for maximizing 
utility expectation also shares the goal of minimizing 
regret, assuming that utility and regret are opposites.
But perhaps by "regret" we mean something that is not 
quantified?