[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Geez.


Nathanael Nerode:
> The site's claim that Approval voting doesn't satisfy the Condorcet 
> Criterion is facile and inaccurate.  With approval voting, for every 
> preference order there are multiple 'cut-offs' which can be chosen by 
> voters based on strength of preference and other criteria, without in 
> fact creating insincere votes.  This isn't even analyzed.
There aren't multiple cut-offs in Approval Voting; you eiter vote for
somebody or you don't. How to decide where to do so is very difficult
and can affect the result in somewhat unpredictable ways -- if I prefer
candidates A, B and C (in that order) the situation "if I vote only for A,
C will win, otherwise B wins" requires that you strategically vote for A
_abd_ B and hides the fact that you'd prefer A from the result of the
election. As an example, consider this vote (I place a bar where
the Approval Voter would cut off):

 3 C|AB
 4 CA|B
 5 B|AC
10 AB|C

Approval vote: 14 A, 15 B, 7 C => B wins.
Condorcet: 17 AB, 15 AC, 15 BC => A wins.

That cut-off point is somewhat arbitrary and can easily change the vote's
result (just switch two of the first two voters to the second option). No
such problem exists with Condorcet.

> The Condorcet proponents at the site claim that approval voting "doesn't 
> allow a voter to express his full preferences".  For that matter, 
> neither does Condorcet, since there's no way to express "A better than 
> anyone else, B & C equal preference, but please not D".

Nothing prevents you from giving two votes equal preference with Condorcet
voting. It's a pairwise matrix, thus your refusal to rank one of them
above the other doesn't hurt their chances against anybody else.

The "please not" problem can be handled with an explicit or implicit "no
choice" option on the ballot (below which all unranked options are
automatically ranked), which is exactly what we're doing here.

> The Strategy-Free Criterion is an excellent case for Condorcet, but it 
> *only* applies to *majority* voting.  Preliminary analysis indicates 
> that Condorcet doesn't have natural sensible behavior for supermajority 
> voting, or mixed supermajority-majority options.  (Approval voting 
> handles both quite neatly.)
"Preliminary analysis indicates" is a null argument. Who analyses that,
did they consider the possible existence of a (explicit or implicit)
default/"no choice" option, and where can I read up on the results?

Matthias Urlichs     |     noris network AG     |     http://smurf.noris.de/

Attachment: pgpPEstXcFT6v.pgp
Description: PGP signature

Reply to: