[EM] 06/13/02 - FWD - From: Saari at aol.com Wed Jun 12 17:01:21 2002:

[Donald E Davison]

  ------------ Forwarded Letter ------------
From: Saari at aol.com
Date: Wed, 12 Jun 2002 16:58:21 EDT
Subject: Re: 06/08/02 - Sixty Percent is Considered a Landslide:
To: donald at mich.com

In a message dated 6/9/02 2:42:45 AM Pacific Daylight Time, donald at mich.com
writes:

<< 06/08/02 - Sixty Percent is Considered a Landslide:

 Dear Mike,

 Thank you for writing - (I think?).

>>You're welcome (I guess).

 If I understand you correctly, you would discard the following single-seat
 election methods: Plurality, IRVing, Condorcet, Bucklin, and Borda, all
 methods which regards any candidate as the winner that receives fifty
 percent plus one on the count of the first choices.

>>Recap: Candidate A: 60% rate as"excellent", 40% rate as"awful".
Candidate B: 100% rate as "very good".
Yes, any system which *automatically* chooses "A"  in this scenario, assuming
all voters vote honestly within the system parameters allowed, is the wrong
outcome.

<< If Approval Voting is your method of choice, it may not be the choice of
 the sixty group.  Those voters may not equate `verygood' as being equal to
 `excellent' and have no desire to vote 60 AC.  If so, I agree and I don't
 see them voting 60 AC.  I say they will `bullet vote' 60 A.

Approval is somewhat better than ranked votes but still falls short.  The
purpose of any voting system is to allow voters to express their views as
fully as possible, tally the results, and determine a winner.  So one measure
of the potential effectiveness of any system is how many "bits" of
information get transmitted in the voting process.  "Vote for one" is pretty
"lean" data.  Ranked votes are better but still too-simple.  If I rank four
candidates A>C>B>D, you still have no idea whatsoever what I think of
candidate C - love it, like it, tolerate it, or hate it.

Approval voting gives me one "bit" of data per candidate to express.  I want
more.  The system I'm developing lets you cast one, two, three or more
"tokens" for or against each candidate.  This allows a rich, robust degree of
"opinion expression".  To tally, take the ratio of "for" tokens over the
"against" tokens.  In a "choose one of several" election, the highest "score"
wins.  In a "Should we do XXX?" vote, any proposal receiving more than the
predetermined "passing score" (e.g. "5:1 is a winner") is thus a winner.

But even if simple single-bit Approval Voting is used, I don't agree with
your assessment.  Let's make the scenario more realistic - assume there are
other candidates as well but all are considered "poor" by everybody.  I'm a
member of the 60% group (A=excellent, C=v.good, PQRS=poor).  How would I vote
under Approval?  Assume that I do NOT have accurate intelligence as to the
likely outcome.  (This is a good assumption.  If there is enough intel that
people think they know the outcome and begin to try to "game" the system - if
the gaming is successful then the outcome changes - meaning the outcome is
really unpredictable all alone.)

How do I vote?  Certainly a "+" for choice A, probably a "-" for PQR and S.
But a choice to NOT vote for C might easily backfire if A happens to lose.
So I might want to play it safer by voting for several good outcomes, not
just one.

A "multi-vote" system makes the dilemma even easier.  Simply vote "+++" for A
and "++" for C or something similar.  This is your best overall strategy.

 `The more things change, the more they remain the same.'  The forty group
 will also `bullet vote', which reduces your election down to a Plurality
 election, not exactly what you are wishing for.

<< Don't worry about the forty group being `pissed', they'll get over it.

But why get people pissed if it is unnecessary?  Outcome "C" is liked by
everybody - but outcomes like "A" give democracy a bad name.  The concept is
good but the implementation sucks if it chooses upsetting candidates over
very nice ones.

There's a nice analogy - your company is going to enter someone into a race.
Candidate "A" is an outstanding runner but also a real prima donna. 50-50
chance he'll be unavailable on race day, but if he enters he's almost a sure
winner.  Candidate "C" is strong and reliable.  Which one to choose?

My point is - the answer should be determined by the actions of the voters.
Sometimes "good enough is good enough", other times "only the best will do -
even if it is risky".  If you let voters say what they think about each
candidate, then they can use their own judgement and expression with each
one, weighing the candidate against the "job requirements" and making an
appropriate assessment.  But with a simple "ranked" vote that entire
opportunity for self-expression is lost.  All I am left with is "I like A
better than C." (but you have no idea if I like C or not, ditto for A).

<< Forty group should keep their head, it's not going to be that bad.  We
 should be more concerned about who makes the laws, how we are going to
 elect them, and how are we going to control them.  We need a bit more
 Direct Democracy, but I digress, you are not asking me about Direct
 Democracy.

>>When our election methods choose extreme, polarizing candidates over
simple, well-liked ones, I become concerned.  When the laws are created (in
Congress) using a "Proposal/Amendment" process which encourages strange
behavior, and where a simple majority is enough to reach a decision but the
hierarchy controls when or IF a topic even makes it to a vote in the first
place, I observe a result which seems less than desirable.

Maybe I need to get my web site up.  I've been working for a long time to
define a workable system - it's time to let some people know about it...
Thanks for your efforts.

Here's a puzzle question.  How can a group of 20-50 people make collective
decisions on various subjects - WITHOUT the use of a chairperson (benevolent
dictator) to "control the meeting", decide when is the "right time" to call
for a vote, etc.?  The current methods are stifling to the various members
who have good ideas but can't bring them to a vote.  I believe it can be done
and done well - but the current system as currently re-defined needs to be
field-tested to know for sure.

Mike




----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em