[EM] No evidence that IRV doesn't fail. Reasons why it must

[Paul Kislanko]

-----Original Message-----
From: Eric Gorr <eric at ericgorr.net>
To: Dgamble997 at aol.com <Dgamble997 at aol.com>;
election-methods-electorama.com at electorama.com
<election-methods-electorama.com at electorama.com>
Date: Friday, January 23, 2004 3:44 PM
Subject: Re: [EM] No evidence that IRV doesn't fail. Reasons why it must

>>somebody said something

>Yes, I can invent numbers that show just the opposite.

This is the most absolutely true statement ever posted to the EM list. All
examples purported to be counterexamples to an assertion include "invented"
parameters that support the position of the poster. But there's something
worse going on...

For every example that meets such-and-such criterion there is a fairly easy
way to come up with a counterexample that shows a failure to meet a
different criterion, and what most people on this list do is switch criteria
from a post to a reply to an objection to their post. For example, a poster
suggests rather strongly that the lowest-ranked alternative on a ballot can
be assumed to be the least-desired alternative, but three replies later the
same poster can say that such a suggestion is untenable if we allow that all
dis-approved alternatives may be treated equally.

The proof that there's no self-consistent way to determine a group
preference from individual preferences includes the mechanism for providing
counterexamples to methods that purport to meet some criteria, provided we
allow the counterexample to use different criteria than the method claims to
support.

See this thread, where I claimed that a conclusion based upon fully-ranked
ballots was incorrect and my assetion was "refuted" by someone who used
truncated and equalities-allowed ballots to support the original assertion
about fully-ranked-ballots with no truncation or equalities allowed. The
original claim is not made valid by the subsequent use of truncated and
equality-allowed ballots, since what I said originally was that you'd need
to use those to draw the conclusions made in the original post.

In my original post, I said the conclusion in the post to which I was
replying deprended upon some implicit assumptions about the method being
used.  In the reply to my reply, the implicit assumptions were made explicit
by the examples using truncations and equalities. But in the orginal post to
which I was replying, the assertion was that the last-ranked alternative was
as important as the first-ranked alternative.

In all of the replies to my post, the responders pointed out all the ways
that the last-ranked alternatives should be given less regard than the
first-ranked alternatives. Which was exactly what David Gamble originally
said, and with which I agree. The poster who claimed that there was
information to be gained from the last-ranked alternative on any specific
ballot didn't support it in this thread, and even gave examples that would
support the contrary.