[EM] Strategy in the face of various levels of information

[Forest Simmons]

On Fri, 20 Sep 2002, Forest Simmons wrote:

> It seems to me that information levels follow a spectrum that goes from
> perfect information through partial information through zero information
> and then further down below zero to misinformation and disinformation.
>
> I believe that Approval tends to do as well as the best Condorcet methods
> in maximizing the fraction of the electorate who find the winner to be
> acceptable, at least when operating in the non-negative information zone.

Heuristic justification for this statement:

In the zero information case Approval voters tend to Approve all of the
candidates that they feel to be acceptable, which tends to maximize the
fraction of the voters that end up finding the winner to be acceptable.
The tendency for Condorcet to pick this type of winner is not quite as
strong, at least in this zero info environment.

[Note that it is possible for 100 percent of the voters to find candidate
B acceptable although sixty percent of them prefer A to B. Condorcet would
give the win to A, and so would Approval in a perfect or near perfect info
environment, but probably not in a zero info environment.]

In the perfect information case Approval will tend to give results similar
to Condorcet methods, usually picking the CW when there is one, which is
most of the time.

So Approval seems to do about as well or better (by the stated criterion
of acceptability) than Condorcet at the ends of the interval of Perfect to
Zero information, so why not in between?

>
> Its robustness leads me to suspect that it will make a good showing in
> the negative information setting as well.
>
> The existence of this negative information zone is what makes
> manipulability an issue, as far as I'm concerned.
>
> Does anybody have any insights into the relative performance of different
> methods (assuming sophiticated voters) in the face of disinformation?
>
> I assume sophisticated voters because all it takes is one respected
> sophisticated voter in each faction to effectively convert the election
> into a sophisticated voter election.
>
> There are at least two cases that interest me:
>
> (1) The experts base their strategy decisions on the basis of misleading
> polls without being aware that the polls are intentionally misleading.
>
> (2) The experts know that the polls are intentionally misleading, and can
> make educated guesses about which part of the polls consist of
> disinformation.
>
> It seems pretty obvious that the USA 2000 election was in an environment
> of disinformation, and that this is what we can expect in most big money
> public elections here in the USA.
>
> Another good reason for heeding the idea by which Joe W. has been seized!
>
> Forest

----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em