Some more standards

[Bart Ingles]

> Bart Ingles wrote:
> > 
> > 1. Unknown candidates lose.  There should be a way to distinguish
> > between candidates who have a consensus, and those who are simply in the
> > middle because the most popular candidates are also the most hated.

Mike Osipoff wrote:
> 
> Maybe that isn't possible. Probably not. If the Democrats &

I think it's possible.  Approval, Runoff and IRO all do it fairly well,
for example, no matter how you rate them with respect to other goals. 
Approval does it by having an absolute rating system (yes/no), rather
than relative ranking.  Runoff-type methods do it by dropping candidates
who lack strong upper-choice support, and by ignoring lower rankings
until needed.

> Republicans, in '92, voted Perot second because they hated eachother
> but not Perot, then doesn't that mean they'd rather have Perot
> than the other big party? The voters are adults. They know that
> ranking Perot 2nd, if successful, can make Perot win instead
> of the other big party. We have to give those people credit
> for wanting what they vote for. Would I rank an unknown over
> a hated candidate? Sure, if I knew that he wasn't as bad. But
> if he were genuinely unknown I wouldn't know that, and he might
> be worse, and so I wouldn't do it. Say there's a John Smith
>  or a "Bob", about whom you've heard nothing. You're not going
> to put that unknown in your ranking, unless you're adventurous
> enough to try order-reversal, which is another issue.

This is more of an issue with a large number of candidates, since it is
difficult to have an unknown in a three-way race.  I'm thinking more in
terms of trying to rank all the presidential candidates of all the
parties in a single race.  When trying to cope with a large number of
choices, the natural tendency may be to rank the candidates that you
know you like near the top, and the candidates you know you dislike near
the bottom, leaving the rest in the middle for others to sort out.  Then
there is random fill, insincere extension, etc.  If there are a number
of unknown candidates, and the more well-known factions are evenly
balanced, then it might be fairly easy to have a surprise winner using
"purely pairwise" methods.

I don't know how likely such an outcome is, but I consider it serious
enough and preventable enough to rank this goal as a requirement.  This
shouldn't unduly impact other "desireable where feasible" goals such as
electing the CW, or achieving majority consensus.
> >
> > 2. If "most hated" candidates are to be excluded, there should be a way
> > to distinguish truly hated candidates from those who are ranked last
> > merely because they compete with the voter's favorite -- in other words,
> > there should be no advantage to insincerely ranking a competitor last.
> 
> Again, that would be difficult or impossible to carry out.
> There was a proposal that one could separately rank from the
> bottom. Maybe, but it carries the risk that you're inadvertantly
> ranking your most hated below an unknown who would be more
> hated by you if you knew something about him. I don't propose
> the separate ranking-from-bottom, for that reason.
>
> Mike

We seem to agree on this one.  My reason for this goal was to balance
other proposals that do seem to advocate dropping "most hated"
candidates first.  People currently vote against candidates because they
frequently have no other choice -- i.e. "lesser of two evils".  I
thought the reason for election reform was to get away from that
necessity.