[EM] Classifying 3-cand scenarios. LNHarm methods again.

[Kevin Venzke]

Hi Dave,

--- En date de : Dim 18.4.10, Dave Ketchum <davek at clarityconnect.com> a écrit :
> > --- En date de : Sam 17.4.10, Dave Ketchum <davek at clarityconnect.com>
> a écrit :
> >> Why IRV?  Have we not buried
> >> that deep enough?  Why not Condorcet which
> does better
> >> with about the same voting?
> > 
> > In the context that I said I wanted to use IRV, I
> wanted to preserve
> > LNHarm. It's kind of a moot point since I don't think
> it can be done.
> > I would guess that certain Condorcet methods would do
> better than IRV
> > in general, but in the symmetric scenario, it's not as
> clear to me what
> > specific behavior we need. It would be enough for me
> to have incentives
> > that keep the scenario from degrading to something
> simpler. I do think
> > IRV would be adequate for that here.
> 
> If there is a CW, any Condorcet method is required to find
> it.  Else there is a cycle and we can debate over which
> such methods do better than others.

I know that. But it's possible that we can trade some Condorcet
efficiency in exchange for other useful guarantees.

> IRV, looking at the same ballots, cannot promise to find
> CWs, for its way of discarding potential winners without
> looking at complete ballots is its major failure.

I know that.

> I have to see little value in LNHarm.
>      For A>B>C IRV will not see
> this voter's approving B and C over D until/unless seeing A
> lose.
>      For this Condorcet will see all
> that this voter says of  B>C, B>D, and C>D.

Not all LNHarm methods are IRV.

> >> Why TTR?  Shouldn't that be avoided if trying
> for a
> >> good method?  TTR requires smart deciding as
> to which
> >> candidates to vote on.
> > 
> > I didn't really advocate TTR. The main thing that is
> nice about it is
> > that there is plenty of room for three viable
> candidates and the method
> > is very simple.
> 
> If we are together, TTR is picking from what some other
> method, such as Plurality, saw as best two.  Trouble is
> either:
>      They found the best one as such,
> and we might as well quit, or
>      The other method did not include
> truly best two to hand to TTR.

Yes, that is a problem.

> > I don't think TTR voting strategy is a big problem.
> I'm more concerned
> > about TTR nomination strategy.
> > 
> >> Will not Condorcet attend to clones with minimum
> >> pain?  Voters can rank them together (with
> equal or
> >> adjacent ranks).
> > 
> > The problem is that you have to get the voters to vote
> for those clones.
> > This is easier when there is no risk to doing so,
> and/or when they are
> > allowed to do something to aid the clone set without
> having to vote for
> > all of them (such as votes against).
> 
> If the voters do not SEE clones, there is little to do for
> them.
>
> If they DO see, they should be thankful for, and use, a
> method such as Condorcet that lets them rank such together.

Many methods will let you do that. But it's not clear that you 
should always use this ability. Sometimes it can worsen the outcome.

> "for all of them"?  Two is easy to have and to vote
> for; more would seem worth less effort due to less
> likelihood.

I should have said "both of them."

Kevin Venzke