|
Re: Lansdowne
|
Blake Cretney
|
Nov 03, 2001 19:00 PST
|
On Sat, 13 Oct 2001 17:14:45 +0200
Markus Schulze <schu-@sol.physik.tu-berlin.de> wrote:
| | Dear participants,
the following paper contains ideas like "sequential independence"
|
which
| | could be interesting for the promotion of Tideman's Ranked Pairs
|
method:
Here's a synopsis. Mr. Lansdowne is writing for the military and
considers some ranked ballot methods for internal decision making.
The actual situation he envisions is where options are ranked by an
administrator according to different criteria. Each criterion acts as
a voter for the ranking it produces. In fact, a criterion may be
allowed to act as multiple voters, in order to give it greater
importance. So, he talks about "criteria" that are actually voters.
I'm not convinced that you would want to use a ranked ballot election
for this situation, but he gives some justification for why you might.
By the way, I'm going to use the word criteria with the ordinary
meaning for this post.
Lansdowne defines criteria called increasing and decreasing sequential
independence. These apply to methods that give complete rankings. If
you can always remove the candidate ranked last without affecting how
the other candidates are ordered, then your method has increasing
sequential independence. If you can remove the first ranked without
affecting the others, then it has decreasing sequential independence.
He is interested in a multi-phase voting system, where some number of
high or low candidates might be removed from consideration on the
basis of a preliminary vote. As a result, he sees these criteria as
valuable, since they would presumably make the eliminations less
disruptive and open to strategic use. If no one changes their mind
between phases, then the phases will have no effect on the outcome.
Presumably, that's what you want. I made a similar argument before,
that it would be desirable that a single vote electing x number of the
top ranked candidate should give the same result as x elections
selecting 1 candidate each time, so that the choice between these
options has no strategic implications.
Just as a side note, if one supports a method that doesn't pass these
criteria, an interesting question is, which result is better? A
single vote electing the top two, or two votes? There is little
choice but to say that the former, the election involving more
candidates, gives the superior result, and attribute this to the
greater amount of information provided to the method for making its
second choice. This is also a general answer to the independence of
irrelevant alternatives criterion.
Using his new criteria, and the Condorcet criterion, Lansdowne
considers several rank-ballot methods. Many of the methods are new to
me. Many involve linear programming, and appear to come, like
Lansdowne, from a military or managerial milieu, where linear
programming is particularly popular. He does not mention Ranked
Pairs. None of his methods pass all the criteria he gives, but Ranked
Pairs would have, so I suspect he was unaware of it.
---
Blake Cretney
|
|
 |
|
