[EM] ignoring "strength of opinion"

[Paul Kislanko]

> Steve Eppley wrote:
> 
> Hi,
> 
> [Rob Brown suggested I post his unintentionally private message 
> to me and my reply.  Here they are.]
> 
> Rob wrote to me:
>   On 12/1/05, Steve Eppley <seppley at alumni.caltech.edu> wrote:
>   > I agree with both of Rob's messages so far on this topic
>   > except for one sentence, which I've included in this
>   > excerpt above.He wrote that collecting some strength of
>   > opinion info cannot be avoided, but I see no strength info 
>   > in votes that are orderings of the alternatives.
>  
>   What I meant was that condorcet methods ignore the 
>   difference between, say, ranking A five positions above B, 
>   vs. ranking A directly above B. Both count as simply "you 
>   prefer A to B", while something like borda count would 
>   consider the former to mean "you *much* prefer A to B". 
> 
> I replied to Rob:
>   Right.  That's a (dubious) interpretation by Borda, not 
>   preference info contained in the voters' orderings.

There's no need to bring poor Borda into this. The "problem" relating to
"strength of opinion" can be described purely in Condorcet terms.

Voter X votes A>...>Z and voter Y votes ....Z>A. In the pairwise matrix that
is the same as those two voters not expressing any preference for A or Z. It
just seems wrong to me that a voter who has A first and Z last can have his
pairwise preference cancelled by someone who who Z next-to-last and A last.