[EM] Markus & Manipulability

[Forest Simmons]

On Wed, 6 Feb 2002, Blake Cretney wrote:

> Well, I can't tell anymore whether we disagree or not.  Here's my point. 
>  If some people are able to get more influence by a greater 
> understanding of the method, or better guesses about how other's are 
> voting, I say that is a bad thing, although to some extent inevitable. 

I agree with you on this, that's why the more transparent the strategy the
better. If strategy is opaque, then the sophisticated have an advantage
over the naive.

>  Some people would say that the better informed have a right to whatever 
> greater influence they can get, and it would be wrong to frustrate this 
> natural process.

Not I.

>  Others would argue that the only important thing is 
> that the "right" candidate wins (possibly the sincere Condorcet winner), 
> so strategy is only good or bad in so far as it helps or frustrates that 
> goal.
>
 
Here's how I think about that. Usually, if a method tends to elect high
social utility candidates under sincere strategy, then the method will not
do too bad under optimal perfect information strategy especiall when it
also happens to be the case that optimal strategy is close to sincere
strategy.

[Also, it seems to me that the nearer that optimal perfect information
strategy is to optimal zero information strategy, the harder to manipulate
the method with falsified polls, because the less sensitive the strategy
is to information. Here's the area in which Demorep's ACMA shines.]

If a method doesn't do too well on social utility under sincere strategy,
then usually the social utility of the winner can be improved if the
voters have access to understandable strategy and to accurate polling
information.

But in this case falsified or inaccurate pools can be disastrous.

If someone gives me a rifle with the sights out of whack, I can compensate
with "Kentucky Windage" after one or two practice shots.  If someone gives
me a rifle with a fancy sight that is electronically controlled to take
into account the speed, distance, etc. of the target, and the fool thing
is not calibrated correctly, then I'm just lucky if the target gets hit
occasionally.

I'll just summarize my thoughts about manipulability and strategy this
way:

If a method is manipulable, then it is important for the voters to
understand basic strategy for self defense. This is easier if the strategy
is transparent.  Therefore it is desirable that manipulable methods have
transparent strategy.

Transparent strategy and manipulability go together, but the relation is
not cause and effect (implicit assumption in Nurmi's definition).  Rather
it is problem and partial compensation. 

Forest

> But I say that although other goals may be important too, the 
> possibility that some voters can get a better outcome by outhinking the 
> rest of the electorate is a bad thing in and of itself, as it frustrates 
> the goal of equal influence.  It is also bad because the result may be 
> less stable, as it can change without a change of opinion.  As far as I 
> can see, this is the position taken by those much-despised academics.
> 
> So, what say you?  Is strategy as I think of it (voting based on extra 
> information) a good thing or a bad thing or neutral.  And if you start 
> by saying that all voting is strategic, I'm not going to respond ;-)
> 
> ---
> Blake Cretney
> 
>