[EM] Cycles in sincere individual preferences

Paul Kislanko kislanko at airmail.net
Mon Sep 6 16:16:49 PDT 2004

See  below.

-----Original Message-----
From: election-methods-electorama.com-bounces at electorama.com
[mailto:election-methods-electorama.com-bounces at electorama.com] On Behalf Of
Adam Tarr
Sent: Monday, September 06, 2004 5:50 PM
To: election-methods-electorama.com at electorama.com
Subject: Re: [EM] Cycles in sincere individual preferences

At 10:34 PM 9/6/2004 +0200, Jobst Heitzig wrote:

>Just a summary of my opinions (in case anyone misunderstood):
>1. Incomplete or even individual prefs are no problem for most methods.

>Most Condorcet methods, yes.  They are pretty meaningless for IRV, Borda, 
and most other non-pairwise methods.<<

No, Condorcet methods that don't allow pair-wise input for the voters are
meaningless. Borda, IRV and most non-pairwise methods do not make the
mistake methods that depend upon the pair-wise matrix makes. 

>2. Whether some set of individual prefs is rational or logical doesn't

>>I disagree.  Social choice methods are (again, excepting unanimity) about 
disagreement and ruling that some opinions matter more than others.  I 
would choose to implicitly disregard self-contradictory preferences.<<

Well, since Condorcet methods can have cycles, they are self-contradictory.
I agree completely, they should be disregarded.

>3. We should not restrict freedom of preference expression without need.

>>Aside from my opinion about the value of cyclic preferences, they 
needlessly complicate the ballot.  Yes, I know they can be but one 
option.  It still complicates the ballot.<<<

They simplify the ballot if done correctly. What COMPLICATES the ballot for
a voter is making them pick from 120 possible ranked ballots when you're
only going to count 10 pair-wise comparisons.

>>>4. Methods based on pairwise comparisons should ask the voter for:
>pairwise comparisons, of course.

Only if you think this gives more information than a simple ranked order, 
which is only true if you think that nontransitive individual preferences 
are meaningful.<<


If you're going to use a "simple ranked order" to infer information, why not
let the voters just PROVIDE that information so you won't always be wrong in
your inferences?

More information about the Election-Methods mailing list