[EM] Summability

[Forest Simmons]

On Thu, 24 Jul 2003, Adam Haas Tarr wrote:

> Kevin wrote and Markus responded:
>
> >> I'm surprised to read this.  I thought "simple strategy" was a
> >> virtue for an electoral method.  Surely runtime isn't considered
> >> a serious issue for summable methods...?
> >
> >No! It is a desirable property that there is no simple way to
> >manipulate the result of the elections.
>
> Well, I don't think anybody here would deny that.  But the simple fact
> that one can more easily get a sense of how the polls will translate into
> election results, does not mean that the method is more manipulable in
> reality.

I agree.  Here's a method that gives horrible sincere results, but whose
transparent cause and effect relationship makes it easily better than IRV:

Method X:  The Ranked Pairs Loser is the winner of Method X.

As soon as everybody understands the method, they realize that optimal
zero information strategy is to reverse their sincere preferences.

Of course, a few well meaning voters will believe that they are voting
their conscience by marking their sincere preferences without taking into
account the actual effect of their vote, while not even realizing that
their expressive intent will be misinterpreted along with the perverse
instrumentality of their ballots.

Note that perfect information strategy and zero information strategy are
as likely to agree with each other under Method X, as they are under
Ranked Pairs, hence much more than under IRV.

If a method has this kind of agreement, and is also simple enough that the
experts cannot confuse the public with bad advice, then the method is
relatively immune to two of the most dangerous kinds of manipulation.

Forest