[EM] Finding the probable best candidate?
Blake Cretney
bcretney at postmark.net
Tue Feb 12 19:32:04 PST 2002
Forest Simmons wrote:
>Pragmatically speaking, in a democracy what do we have besides voter
>utilities, intuitions, and hunches to measure the goodness of a
candidate?
>
Well, voters may have sound reasons. But you're right that we have only
the voters to go by. But if there is such a thing as a voter being
correct, and if voters have some tendency toward being correct, it
would appear that the probable best candidate is a reasonable goal.
>
>The utilities don't have to be "utilitarian." They include the
altruistic
>values held dear by the voters as well as their economic interests.
>
But if you're hoping to use the method to find the candidate that will
cause the greatest utilitarian benefit, then you don't want people to be
altruistic. You want people to accurately give their personal benefit,
in the hopes that when the method sums the personal benefits you'll get
societal benefit. At least assuming all members of society vote.
For example, let's imagine there are three faction's the A's, the B's
and the C's. The following table shows the utility of each candidate to
each faction. They don't add to 100.
Actual Utility Total
A B C
24 A 1 .2 0 31
36 B .3 1 .1 46
30 C 0 0 0 0
C doesn't really benefit anyone (at least as far as the A and B voters
can see), but the C voters intend to vote for C alone. The B candidate
increases utility by the most, but the A candidate increases A utility
by the most. If the B voters want to increase total utility, they would
see that A's utility is much better than C's, and approve it as well.
If the A voters want to maximize their own utility, they will see that
B does much worse, and approve only A.
In this example, A would win. So, by voting in the common interest, B
actually reduced the total utility of the outcome, from their
perspective. Of course, with information and strategy, there might
be a different result.
---
Blake Cretney
More information about the Election-Methods
mailing list