Mitigating IR) & Plurality

Mike Ossipoff dfb at
Fri Nov 22 19:56:19 PST 1996

First I want to emphasize that the mitigations that I'll
describe here are "patches", and that it's better to use
a good method, then to use a bad method with a patch. Nevertheless,
it might be a good thing to have these patches ready in case
they're needed as a last resort.

Tom mentioned Hallett's method in a letter that I'm still trying
to reply to, trying to overcome some mailer incompatibility problems
(when I "group-reply", my mailer sends a copy to "distribution
list at", or "distribution list at", and then
, somewhere along the line, the entire transmission gets aborted
because of that incorrect address. In principle I could copy
all the reciepients' addresses in manually, but if I make even
1 typing error (I have no backspace when doing that), the
whole thing gets aborted because of that 1 mistyped address.
I've made about 3 or 4 attempts so far, and I'll keep trying,
and maybe eventually I'll type all the addresses correctly,
or there will be a way for me to reply, instead, to a letter
with those names in its Cc: line, which is what my mailere
can deal with.

Anyway, my reply to those letters will be along as soon
as possible.

Anyway, Hallett's method, as Hallett explained it in his book,
amounts to choosing the Beats-All alternative if there is one,
and othterwise, to use IRO (Hare-Ware).

This is a patch on IRO. Not a completely effective one, since
the method still fails GMC, and the basic democratic principle
on which GMC is based.

So why not add another patach? How about saying:

The method is IRO, but:

Any alternative that is pairwise beaten is disqualified before
the count unless every alternataive is pairwise beaten. And any
alternative that is majority-rejected is disqualiflied before the
count unless every alternative is majority-rejected.

That paragraph avoids the problem of the method failing
Condorcet's criterion & GMC.

I'm not saying that makes the resulting method as good as
Condorcet. It doesn't. But it patches up IRO's worst


Similarly, those same patches could be put on Plurality, by
the same wording.


Another possibility, completely different: Re-voting based
on favorite count rule:

Say, again that the method is IRO. Then we could say:

The method is IRO, but there will be 2 counts. The 1st count
will be conducted using all the methods in a (reasonably long)
list of methods favored by some voters.

In that 1st count, a voter may specify that s/he wants the winner
according to a certain count rule (specified by hir) moved to the
top of hir ranking, for the 2nd count, which will be by IRO.

Any voter could do that. The 1st count would actually be several,
possibly many, counts, a count using each method that somenoe
specifies (again, up to a reasonable number of count methods).

So, in the 2nd count, by IRO, each voter would have the winner
by hir favorite count rule at the top of hir ranking. Even
IRO can't hurt the alternative that you have at the top of
your ranking.

Again, for that matter, the method, instead of IRO, could
be a Plurality count. The voter would specify that the
alternative that won the rank-balloting, by hir favorite
method, shall be the one has hir vote in the 2nd count,
by Plurality.

IRO advocates will hate this, because it takes away IRO's
chanced to deny people's wishes.


Again, it would be better to just use the best method, rather
than this method, but this method tremendously mitigates
even the worst methods (the ones mentioned above). I call
this "revoting using favorite method". 

It's would be pretty difficult to argue against. As long as
the lousy method that someone likes is the official count
methd, how can they object if the voter is given a chance
to choose how s/he will vote based on good information?

Tom, would you do me a favor & forward this to that distribution


Oh, one more thing: The voter could also have the option to
specify that everything but that 1 alternative shall be
dropped from hir ranking (if the official method isn't
Plurality--because that's automatic in Plurality).

In fact, especially for the IRO advocate, there could be
a special rule, if Condorcet is the official count method,
for the 2nd count, that the voter has an additional option
to have dropped from hir ranking the alternative that is
1st choice of the fewest. All three options I've suggested
could be offered, and used singly, or in any combination.

Those options are, to repeat:

The voter can indicate that:

1. That the winner by a method specified by the voter shall
   be moved to the top of hir ranking for the 2nd count.

2. That every other alternative shall be dropped from hir ranking.

3. That the alternative that is the 1st choice of fewest voters
   (in the initial raw rankings, before voters have modified
   their rankings) shall be dropped from hir ranking. (This
   3rd option is especially for IRO advocates).





More information about the Election-Methods mailing list