[EM] In defence of IRV

[Rob Lanphier]

Hi Robert,

Thank you for highlighting your paper.  More below:

On Sun, Nov 14, 2021 at 1:46 AM robert bristow-johnson
<rbj at audioimagination.com> wrote:
> Colin, I hope you read my paper.
>
>  https://drive.google.com/file/d/1jIhFQfEoxSdyRz5SqEjZotbVDx4xshwM/view

I've taken a closer look at your paper, after taking a look at the
Tideman paper that I believe that Colin was referring to.  I
appreciate that you have not mischaracterized your paper, and that
your paper is very salient to this conversation, and that you provided
a link to it.

I'm much more positive about the recent developments in Fargo than you
seem to be (as implied by the opening statements of your paper).
However, I believe that Ka-Ping Yee's simulations demonstrate a very
important characteristic about approval voting.  Here's a link to
Yee's work again:
http://zesty.ca/voting/sim/

Yee's work appears to show that approval voting results are MUCH more
likely to align with Condorcet-winner-compliant methods than
Hare/IRV/RCV elections.  This is a big reason why I'm pretty sanguine
(and even a bit enthusiastic) about the implementation of approval in
Fargo and St Louis.  Do we differ on this point?

Burlington's 2009 election was arguably an anomaly, which you point
out in your paper.  I appreciate that Hare/IRV/RCV elections choose
the Condorcet winner the vast majority of the time.  I also appreciate
that bridges that I drive over have routinely "done their job" and
didn't collapse beneath my vehicle.  I appreciate that almost all of
the flights that I've been on have been routine affairs, and that the
ones that weren't still resulted in my safe arrival at my destination.
We know that bridge collapses and airplane wrecks are very anomalous
events.  We don't know if Hare/IRV/RCV's failures are a 1 in 100 type
of event, or 1 in 1000, or even more rare.  I'm grateful that EVERY
commercial airliner crash receives thorough scrutiny, and I appreciate
your perspective on the 2009 election in Burlington, since my
understanding is that you participated in that election.

Prior to Burlington 2009, the IRV<->Condorcet discrepancy that I liked
to study was the Debian Project Leader election in 2003.  I should
probably provide a more accessible writeup than this one:
https://electowiki.org/wiki/User:RobLa/Debian2003

It's not only a clinical evaluation of the end result that bothers me
about IRV elections.  It's also about the candidate and voter behavior
that is elicited prior to the election.  Because IRV is subject to
"center squeeze" (as you point out in your paper), candidates are
motivated to appeal to extreme factions of their electoral base.
Centrism (and compromise) becomes vilified by activists.  Do I have
mathematical proof of this?  No, of course not.  This paragraph is a
political analysis, not a scientific analysis -- but I'm pretty
confident in my punditry on this subject.

Regardless, thanks for providing a link to your paper.  I think this
mailing list becomes a more valuable resource with relevant links.
You would have received extra bonus points (in my mind) by providing a
quick summary of the paper (or rather, the "tl;dr version", as the
kids like to call it, or an "abstract" as academics like to call it).
However, by keeping your message brief, you avoided inscrutable jargon
and you avoided posting a misinterpretation of your paper.  Kudos for
that!

Rob