[EM] Letter to Discover Magazine

MIKE OSSIPOFF nkklrp at hotmail.com
Fri Nov 3 23:55:48 PST 2000

```

Joe Weinstein wrote:

Bart's example is intended to make the case for 'approval' - i.e. a voter's
pass-fail grading of all competing options (i.e. candidates or measures).
In fact it makes an even better case for higher-resolution grading, where
there are more grades than just 0 (fail) and 1 (pass). Using today's (or
anyhow next year's) computer, there seems every prospect that elections can
allow voters to use a higher-res grading scale, say the the scholastic
5-grade scale (F=0, D=1, C=2, B=3, A=4) or, better yet, the 0-100 grade
scale.

You're quite right: In a Utopia where everyone would rate sincerely,
and trust everyone else to, where no one cares about optimizing their
outcome--as can be done by insincere voting in such a system--in
that kind of a Utopia, I agree that the method you suggest, which
has been called Flexible Points Assignment, and Cardinal Ratings,
would be the best, if we want the greatest good for the greatest
number. If we want to maximize social utility, which I think would
be a good goal under those conditions.

Under more realistic conditions, voters will vote strategically,
good method, even under real conditions, even when people vote
strategically. What's the best strategy for Cardinal Ratings?

Give maximum points to everyone you'd vote for if it were Approval,
and give minimum points to everyone whom you wouldn't vote for in
Approval. In other words, if we're allowed to assign 0-100 points,
then we give 100 to everyone we'd vote for in Approval, and zero to
everyone we wouldn't vote for in Approval.

In other words, Cardinal Ratings is strategically equivalent to
Approval. Under existing conditions, it's effectively the same method.

So we might as well just use Approval. No need to complicate the
balloting. The only change needed is: Where the ballot now says
"Vote for 1", it would instead say "Vote for 1 or more".

Is there still a case for proposing Cardinal Ratings? I'd say yes,
because people are used to rating things from 1 to 10, or from 0 to
100. Maybe people's greater familiarity with Cardinal Ratings would
outweigh its more demanding balloting requirements. Someone said
that Olympics judges rate contestants from 1 to 10, making that
an extremely familiar voting system.

I propose Approval, because it's such a small change from what we
use now, but it could turn out that Cardinal Ratings would get
better acceptance, due to people's prior familiarity with it. If so,
I'd advocate Cardinal Ratings. People would soon catch on to the
strategy.

Weinstein continues:

two-option contest to be LOST by the option that is preferred
(qualitatively) by the majority.  That is actually quite OK:  the
sufficiently fierce preference of a sufficient minority OUGHT to outweigh
the mild preference of the majority.

That's true under the Utopian conditions that I mentioned. With Approval
or Cardinal Ratings under existing conditions, in a two-option
election, of course the majority winner would win every time.

The members of that majority, in Cardinal Ratings, would give maximum
points to their preferred alternative, and minimum points to the
other alternative. If it's 0-100, they'd give 100 to one alternative
and zero to the other.

That isn't a bad result. We have no reason to expect otherwise,
people being as they are.

It turns out that only a very few of the very best rank methods are
as good as or better than Approval. The others all have serious
strategy problems. They retain our lesser-of-2-evils problem
to a greater degree. With Approval or Cardinal Ratings, no one ever
has any incentive or need to vote anyone else over their favorite.

Mike Ossipoff

Unconstrained high-res grading would seem to combine the best features of
and of artificially constrained high-res grading (e.g. ranking schemes).

Joe Weinstein
Long Beach CA USA

>From: Bart Ingles <bartman at netgate.net>
>To: "election-methods-list at eskimo.com" <election-methods-list at eskimo.com>
>Subject: [EM] Letter to Discover Magazine
>Date: Sun, 29 Oct 2000 22:30:21 -0800
>
>
>Dear Editor:
>
>Dana Mackenzie's article [May the Best Man Lose, November] may be the
>best treatment of the subject I have seen in a popular publication.  Of
>the featured voting systems, Borda could well be the method of choice
>for engineering applications where the voters are automata incapable of
>varying preference levels.
>
>Not so with human voters.  Suppose the voters rate candidates on a scale
>of 0-10 (ratings in parentheses):
>
>45%   Clinton(10)   Perot(1)   Bush(0)
>25%   Perot(10)     Bush(1)    Clinton(0)
>30%   Bush(10)      Perot(1)   Clinton(0)
>
>Here Perot is the Borda winner, even though 75% of voters strongly
>dislike Perot.  The runoff winner (Bush) is despised by 70%, while the
>plurality winner is only rejected by 55% (and seems the best choice in
>this example).
>
>Of course, if either Bush or Perot were more highly rated as a second
>choice, then Runoff or Borda would choose the better winner.  No
>deterministic voting system can be correct in all cases where the voters
>are either strongly for or strongly against each candidate.
>
>Approval voting does adapt to all of these scenarios, if we assume the
>voters are rational and willing to make modest compromises in exchange
>for large gains in probable outcome.  To me it seems more important for
>a system to behave reliably in such highly polarized situations, than to
>worry about which solution is best when the voters themselves are
>ambivalent.
>
>Bart Ingles, Jr.
>

_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.