Finding the probable best candidate?

[Forest Simmons]

On Sat, 16 Feb 2002, Blake Cretney wrote:

> Forest Simmons wrote:
> 
>  >In this EM archive thread, what if (for starters) we just stick to a two
>  >way race between two candidates.
>  >
>  >Wouldn't we all agree that the best democratic method is to give the win
>  >to the candidate with the majority of votes, whether or not the voters
>  >have actually based their votes on fact, fiction, or superstition?
>  >
>  >If we cannot resolve the two candidate case, then we have no hope of
>  >resolving the multicandidate case.
>  >
>  >Yet in the two candidate case the philosophical question about absolute
>  >best or just best for the voters seems to have no bearing on choice of
>  >method.
>  >
>  >Why does that question suddenly become so important when we bump up the
>  >number of candidates by one?
>  >
>  >Of course there are additional difficulties with more candidates, but do
>  >those difficulties really have anything to do with this particular
>  >philosophical question?  If so why?
>  >
> By "best for the voters" I'm going to assume you mean best according to
> the voters.  I also suspect you are treating the voters as a kind of
> group entity, the way some people think of the public.  The public has 
> certain desires and opinions.
> 
> People who take this view generally take a majority opinion by the
> members of the public to be the desire of that entity called the public.
>   With two candidates, this is a plausible way to think about things.
>   But with multiple candidates the public ends up with contradictory
> desires.  And not only in those Condorcet cyclic situations.  For
> example it may be that a majority think A isn't the best, a majority
> think B isn't the best, and a majority think C isn't the best.  But this
> is contradictory, since these are the only options.  For Demorep, we can
> imagine that one of the options corresponds to none of the above.
>   Unless one option receives a majority of first place votes (and another
> a majority of last place votes) there is a contradiction in majority
> opinion.
> 
> So, the multi-candidate case forces us to abandon the public will
> analysis.  There is no public will.  Just a bunch of people with
> different opinions.  But if our method isn't designed to find the public
> will, what is it trying to do?
> 
> Well, there's different ways of answering that question.  However, if we
> agree that we're trying to select the best candidates possible, and if
> we agree that voters have some insight into the question of which
> candidates are better, then it becomes plausible to see an election as
> using the information provided by voters as evidence toward the question
> of which candidate is best.

So you are using "best candidate" as an undefined term like the terms
"point" and "line" in an axiomatic development of geometry.


> Not 100% reliable evidence, of course, but
> evidence the method can use to make a guess.
> 
> That doesn't solve the problem.  But it does give a starting point for
> arguing about what standards are valuable, and what assumptions are
> reasonable, and then what procedure is implied from various standards
> and assumptions.


I think it makes arguments more difficult to resolve.

A definition of "best candidate" in terms of voter utilities, for example,
would make these arguments easier to resolve. 

Forest