Approval and twins

Mike Ositoff ntk at
Fri Oct 30 19:40:03 PST 1998

> On Wed, 28 Oct 1998 18:02:56   Mike Ositoff wrote:
> >> 
> >> I consider rich-party and vote-splitting to be two of the most
> >> important problems, so I think it's important to ask how well
> >> approval does in regard to them.
> >> 
> >> Certainly, approval passes GITC.  However, GITC was designed
> >> for rank methods.  In theory, plurality passes GITC.
> >
> >In no way could Plurality be said to pass the Clone-Independence
> >Criterion, if that's what GITC refers to. If an added clone
> >splits the vote, it could make the former wininer in the clone
> >set lose.
> Technically, any candidate who receives a vote in plurality is 
> not a clone, since it has been differentiated from all the other 

If the clone criterion were about preference rather than rankings,
then Plurality would fail it. Approval might fail it, if the
new clone were so much better than the others that partisans
voted only for it, just on the chance that it might be 
a frontrunner.

If the criterion is about votes, maybe something meaningful
can be said about nonranked methods if we speak about
felt preferences.

Approval's failure of it doesn't bother me, since adding a new
clone doesn't force me to abandon the others if I vote for the
new one. It shouldn't keep new candidates from entering, and
if people know what's going on it won't keep opposition from
succeeding with decoy clones.

> candidates.  Of course, this doesn't mean that vote-splitting 
> doesn't work, just that we should be careful using GITC to test 
> non-ranked methods for these problems.  In fact, as I've
> mentioned before, we should be careful about methods that
> don't allow equal rankings too.
> >
> >You say that Approval passes GITC. If GITC is designed for rank
> >methods, but Approval passes it anyway, I guess I don't understand
> >what the problem is.
> The problem is that the fact Approval passes GITC doesn't prove
> it doesn't have the rich party and vote-splitting problems.

Did someone show that Approval has the rich party problem? That
a party can ensure a victory by running lots of candidates?

If no one knows anything about who's likely to be a frontrunner,
then a party that runs lots of candidates undesirable to me
could lower the average utilility for me, so that if I vote
down to average utility, I'd vote farther in their direction.

If I vote to minimize the possible regret (I don't have the 
definition of regret closeby, but could find it), the strategy
is to vote down to the mean the utility of my favorite &
least favorite candidates, and so a party runs someone _really_
bad could cause me to vote more down toward them if I followed
that strategy. We haven't assumed that people vote to minimize
possible regret, though.

If I don't have utility ratings of the candidates, the
optimal strategy, to maximize expected ultility is to
vote for the best half of the candidates. Then, if a party
runs lots of bad ones, that would cause me to vote more
toward their direction.

But if, more realistically, I have information about utilitly
& tie probability, then has anyone shown that a rich part
that runs lots of candidates can win because of it?
Tkat's the more likely scenario, and you'd have to show
that the rich party problem exists then.


As for the split vote problem, it was recently defined identical
to the clone problem, a criteria equivalent to the clone
criterion. But if you mean that a compromise is caused to lose
because we refuse to completely abandon our favorite, that's
a problem of Margins, but not of Approval. If youl mean
a compromise is caused to lose because we refuse to rank it
equal to our favorite, that's a problem of Approval & Margins,
but not of VA.

> >> In ranked methods, a voter can differentiate between the
> >> members of a party, and still rank those members higher than
> >> anybody else.  This is not possible with approval.  In
> >> approval, a voter must decide whether to help the party as
> >> much as possible, or to differentiate between members of
> >> the party.  And if voters tend to vote purely along party
> >> lines, the choice within the party could be made by very
> >> few people.
> >Meaningless. A rational voter doesn't base his vote on
> >party loyalty. Not everyone is so party-oriented.
> Few voters would base their vote on a blind allegiance to a party
> name.  However, if the members of a party have similar policies,
> it is very likely that a voter may prefer any of them to any
> other candidate, but still have preferences within the party.

I hope you're being pessimistic about the degree of party-voting
that would occur using Approval. If the party's candidates are
indeedd almost the same, then voting for all of them is good
Approval strategy, if you consider all of them much better than
others. But hopefully not because of party membership.

Somone more like me would vote only for the best. Of the
strategies you named earlier, that more nearly described how
I'd vote, though I just call it voting for the acceptable ones.
I'd use Approval more as Demorep would, voting only for
absolutely acceptable candidates & rejecting absolutely
unacceptable ones,no matter if they're significantly less
bad than some other party. Of course the election still
has to be conducted & counted & discussed as a relative
choice, even if some, like me judge absolutely.

> Of course, this is the assumption on which the idea of clones is
> based.

Clones is about a particular special case that may or may
not be closely approached. That's why failure of it doesn't
bother me so much, and why I'm not at all impressed by IRO's
good performance in that special case, when it fails more

> >
> >> 
> >> There are two strategies that I can imagine voters employing.
> >> They will largely be psychologically determined, so it would
> >> be necessary to find out which is used by polling and
> >> observation.
> >> 
> >> Except Worst
> >> The voter votes for most members of the party, but withholds from
> >> a few of his least favorites.
> >> 
> >> Only Best
> >> The voter votes only for a few of his favorite candidates from
> >> within his party.

That's closer to my voting, but doesn't describe it exactly.
I don't have a party; none are good enough to qualify as my

> >> 
> >> If a party thinks its voters fall into the "Except Worst" category,
> >> then it makes sense to run more candidates, and thereby spread out
> >> the "No" votes.  If on the other hand they follow the "Only Best"
> >> strategy, it makes sense to run fewer candidates in order to
> >> concentrate the "YES" votes.
> >Approval has more than 2 strategies, that I've read of. The ones
> >you list aren't among the ones I"ve read of or would use.
> I suspect I would end up voting like the "Except Worst".  How would
> you balance the desire to get a party whose policies you support
> elected, with the desire to choose from within that party?

If I really liked its policies, and that were more important
than individual differences in the party, then why not vote
for them all in Approval? In practice, there's a great difference
within parties, & I certainly would be unlikely to vote for all.
I'd disqualify the ones I didn't like.

> >> If a party thinks its voters fall into the "Except Worst" category,
> >> then it makes sense to run more candidates, and thereby spread out
> >> the "No" votes.  If on the other hand they follow the "Only Best"
> >> strategy, it makes sense to run fewer candidates in order to
> >> concentrate the "YES" votes.
> >Approval gives no incentive to "concentrate" votes. A party
> >runs the best candidates it can, and lets anyone run who wants
> >to, since people can vote for him without abandoning other's.
> As I say, it depends on how people vote.  If people are too picky,
> then running more candidate will hurt a party, since the factions
> of a party will only vote for their own candidate and the vote will
> be split.

If people vote to maximize utility expectation, and have information
about utility to them & about tie probabilities, then can
you show that Approval has that problem?

It's distasteful for a party to protect a faction by trying
to shut out other factions. I doubt I'd vote for such a party.

> ---
> Blake
> -----== Sent via Deja News, The Discussion Network ==-----
>  Easy access to 50,000+ discussion forums

More information about the Election-Methods mailing list