[EM] 03/29/02 - Rob Richie Letter and Non-Monotonicity:

[Blake Cretney]

Donald Davison wrote:

>03/29/02 -  Rob Richie Letter and Non-Monotonicity:
>
>  ------------- Forwarded Letter ------------
>At 2002\03\23 07:52 -0500 Saturday, Rob Richie wrote:
> >Craig,
> >
> >I would think you would know that you can generate
> >whacky results with any system. Your example below
> >of course is political nonsense (why would every B
> >voter rank C second, but no C voters rank B second),
> >just as some of the examples that can be used to
> >mock every system you would propose.
> >
> >No, you and I know from past experience that dialogue
> >between us isn't worth the time.
> >
> >- Rob
> >
>
First, this looks like a personal correspondence between Craig and Rob. 
 Presumably Craig forwarded this to you (which he shouldn't have done), 
and you forwarded it to this list (which you shouldn't have done).  It's 
quite possible that Rob Richie would have worked harder on his argument 
if he had known that this was going to be publicly released.  But since 
the message is on the list now, I'll comment.

The reason everyone uses simplified examples, is that they are easier to 
read, write, and understand.  The idea is that if a method goes wrong in 
an obvious way in simple examples, we should have little confidence when 
the situation is more muddy.  Of course, people on this list have shown 
that monotonicity, for instance, can be violated in examples that aren't 
at all simple.

One of IRV's advantages is that it doesn't allow the vote to be split. 
 Some other methods do.  But to show this, I would have to construct an 
example where it was clear that two parties were in fact splitting the 
vote, for example, if each picks the other as its second choice.  But 
now, I can hear the Davison and Richie parallels in the anti-IRV camp 
saying that such an example is unrealistic, and in a sense it is.  No, 
the only realistic example wouldn't have people perfectly splitting the 
vote, so it wouldn't prove that a result occurs only from vote 
splitting, so it wouldn't prove anything at all.

But wouldn't an IRV supporter be right in saying, if vote splitting 
occurs even in this extreme example, where the result can only be caused 
by vote splitting, won't it occur too in normal elections.  Shouldn't 
you at least be prepared to give some argument why it shouldn't.  I 
think the IRV supporters would be right in this argument, just as 
Davison and Richie are wrong in their's.

It saddens me that Rob Richie would make the argument he does, because 
of late I have found myself defending IRV and its supporters (even 
though I don't actually advocate the method).  But how can I respond now 
when people accuse IRV supporters of being naive in their arguments?  It 
also means I'll have to update my RP web site to refute an argument that 
I previously didn't believe  was worth refuting.

---
Blake Cretney (http://condorcet.org)