[EM] Why the concept of "sincere" votes in Range is flawed.

[Juho Laatu]

--- On Wed, 28/1/09, Kristofer Munsterhjelm <km-elmet at broadpark.no> wrote:

> Juho Laatu wrote:
> 
> > A = 1000
> > B = 200
> > max_support = 100
> > C = 50
> > approval = 30
> > D = 1
> > min_support = 0
> > E = 0
> > F = -100
> > max_preference_strength = 10
> > 
> > Approval interpretation is A=B=C>D=E=F.
> > Range interpretation is A=B=100, C=50, D=1, E=F=0.
> > Rankings interpretation is A>B>C>D>E>F.
> > Rankings interpretation with preference strengths is
> A>B>C>D>(0.1)>E>F
> > 
> > I'm not sure if anyone really wants
> > all this. The first practical
> > implementation might be a Condorcet
> > election that covers also ratings
> > data for polls.
> 
> If you're going to do Condorcet and Range comparisons,
> a simple Range ballot would suffice.

Yes, the rest of the information
is not needed.

> Then, if candidate a
> has a lower score than b, in the transformed ranked ballot
> (to use for Condorcet), b > a.
> 
> It seems quite hard to defend the request for Range type
> information if such information isn't actually going to
> be used.

Yes, that may mean some extra work.

Collecting Range type information
that is going to be used in the
election may be even more difficult
(because of strategic interest not
to give sincere opinions).

> Therefore, unless either Range or cardinal weighted
> pairwise is to be used, I agree with you: it would only be
> of polling use.

CWP could collect sincere ratings
data even in competitive environments.


One more observation. If we want to
collect sincere non-normalized
ratings data it would make sense to
define some reference points.
For example in range [0,1000]:

900 = best possible candidate
800 = excellent candidate
700 = good candidate
500 = average candidate
300 = bad candidate
200 = terrible candidate
100 = worst possible candidate

Juho