[EM] Responses to some of Forest's ideas

[Forest Simmons]

As usual we rely on Craig to come up with all of the interesting examples.

I'm willing to bend my stance a little on the basis of comments and
examples from Rob, Craig, and Demo, as well as my belief in Universal
Approval as a good standard. 

The examples and discussion around them show that deciding an approval
cutoff can be an exasperating experience, especially if we just know the
utilities but don't know how the voter feels about the candidates.

Demorep's method would take some of the pressure off for getting the
approval cutoff just right. A non-sophisticated voter might appreciate
this feature, as Rob and others have argued in different words. 

And as I mentioned before, the general population would probably agree
that the winner in the (zero-info) example below should be the CW, and not
the Approval winner: 

55 A > B >> C
35 C > B >> A
10 C > A >> B

Universal and Dyadic Approval also give the win to A, but just barely, so
I would agree as well, on the basis of my favorite methods.

But playing the devil's advocate for a minute, I will state the case for
B:

Let's use Demorep's definition of a yes vote as our definition of
approval.  A voter is supposed to approve a candidate if that candidate is
"tolerable."

In the above example B is intolerable or totally unacceptable to only ten
of the voters.

At what cost do we go from candidate B to candidate A? 

The cost is a net increase of 25 in the number of voters that have to be
represented by someone totally unacceptable to them. Now 35 voters are
represented by someone intolerable to them.

What do we have on the positive side to balance out this distressing
change of fortune? 

Well, now 55 of the voters that would have been represented by a merely
tolerable candidate get to have someone they like better, someone that
will better cater to their interests.

The additional luxury of the 55 compensates for the abject misery of the
25 dropouts.

This might make sense in a society where the benefits get spread around
evenly "from each according to his abilities and to each according to his
needs" but in a society of "every man for himself" this is a cause of
concern.

The fallacy of Expected or Average Utility as a measure of aggregate
utility is the assumption that one man's gain can compensate for an equal
loss by another man.  It might work in some utopia, but not in any country
or state that I know of (where the only form of welfare without stigma is
the giant corporation feeding at the public trough).

This is why I suggest filtering the sincere utilities through the square
root function (for example) before adding them to get aggregate social
utility in future simulations.

One more argument in favor of the sincere zero-info approval winner B:

Democracy implies "consent of the governed." 

Which of the candidates A or B would govern with the consent of more of
the voters?

I submit that the members of the A faction could consent to B as their
governor without too much difficulty, but how could the 35 members of the
CB faction consent to someone that they find totally intolerable and
unacceptable?

For the 55 first faction supporters to willingly use their majority status
for personal gain while forcing 25 additional citizens into abject misery
would require either a callous lack of compassion or an arrogant
paternalistic attitude of, "We know what's better for them than they do." 

So either the greed or the arrogance of 55 percent deprives the body
politic from being governed by maximum consent.  Approval = Potential for
Consent.

In summary, we don't have to argue for adoption of Approval over Condorcet
only on the basis of simplicity, Favorite Betrayal Criterion, etc. 
Approval gives more hope than any other (multi-candidate single winner)
method to the disenfranchised minority against the "tyranny of the
majority." 


Forest