Exaggerated opinions

Saari Saari at aol.com
Sat Apr 18 18:31:04 PDT 1998


In a message dated 98-04-09 07:38:47 EDT, you write:

<< What do you mean with "acceptable" or "violently opposed". To my opinion,
 you should always remember, that every voting method, that uses absolute
 preferences, is vulnerable to exaggeration. >>

Here are two counterexamples to the above claim.

1) Consider a system which allows every member to cast an opinion on every
"proposal".  Valid "votes" include single-Support, single-Oppose, and
Neither/Neutral.  Fractional votes (i.e. 1/2 partial support) are also ok.
Voting on multi-choice options occurs simply by making each option a separate
proposal.  At any given time there are a wide variety of proposals pending
(and a pre-specified criterium to determine when enough time has elapsed to
allow a possible successful decision/conclusion). A "passing score" consists
of a Support/Oppose ratio of 5:1, in other words a proposal passes if every
Oppose vote has 5 or more Support votes to counteract it.  [This is one way to
generate "rough consensus" without relying on a hierarchical dictator to
decide when close enough is close enough.]

Now suppose that for a given topic there are a dozen or so alternatives - each
as a separate proposal.  [This system never needs a "ballot" created by some
person in charge - simply a variety of proposals generated at will by the
members.  There are no amendments in this system - merely fresh proposals.]
(It is possible that no proposal will pass - meaning no decision.)  Here are
your "honest" opinions on each:
A,C,F - mediocre
B - Outstanding - you really like this one.
E - Pretty good - not great but you can live with it.
G - Horrible, rotten
H - Also bad
D - Quite good - not quite as excellent as B but still you'd be pretty happy
with it.

So how would you vote on each, given that you need to express *something*
before really knowing how well each may do?  Clearly you'd vote Full-Support
on B, Full-Oppose on G and probably H.  But what about the rest?

Some people will claim that you should logically "exaggerate" and vote either
Full-Support or Full-Oppose on all of the other choices but I believe this is
incorrect thinking.  How to vote on the "mediocre" choices depends on whether
you think "mediocre" is better or worse than nothing/no decision - situation
dependent.

What about the other "good" choices D and E?  If you vote Full-Support then
you give them maximum support but fail to distinguish in your votes the fact
that you really like B better.  But if you vote Full-Oppose (trying to push B
the most) you risk an outcome of no decision which is probably not what you
wanted either.  The solution is intelligent use of the fractional votes.  In
other words, vote something like
3/4Support on D and 1/2Support on E (or whatever other values really make the
most sense for your particular opinion - perhaps 0.99 for D and 0.98 for E).

This "smart vote" gives some support to the alternatives that you like, AND
also distinguishes that you really like B better.  The "nearly-full" votes
e.g. 0.99 carry virtually the same weight as a full vote, but they also serve
as a tie-break in the direction you favor if it comes to that.

Thus, the fact that you could exaggerate your preferences is countered by the
idea that the best way to express your feelings is to NOT exaggerate.

Example#2.  This is a method I'm exploring right now.  Same system as above,
but suppose that double-votes, triple-votes, etc. are also allowed.  But there
is a cost.  A double-vote costs $1 for each instance.  A triple-vote costs
$10.  A quadruple-vote costs $100.  All money collected goes into a common
pool which is routinely split evenly among all of the participants.  [It is
much harder to "buy" an outcome than it will appear, though.  The 5:1 passing
ratio means that unpopular ideas can be quickly squelched by a relatively
small number of Oppose votes.]

So...how do I behave?  A given proposal looks good to me, but I'm not really
much concerned about it.  Perhaps I vote Support, or perhaps I vote Neutral
which means that I will allow the opinions of the people who really care to
decide the issue.  Better not to vote "blind support" - such a behavior is
unnecessary in this system.

Now, another proposal comes to my attention.  It strikes me as a pretty bad
idea, so I vote Oppose.  In fact, I feel strongly enough that I'm HAPPY to
toss in a buck in order to get a double-Oppose vote - just to try to make sure
that it quickly dies.  But a $10 triple-Oppose is not enough additional value
- I don't care QUITE that much.

A third proposal comes up -  It looks like a winner, BUT it has one fatal flaw
clause I object to.  My vote here is mixed: Support PLUS Oppose.  The Oppose
vote carries the most weight (because of the 5:1 passing requirement) but this
mixed vote conveys that I support the idea in principle - just fix the flaw
please.

A fourth proposal is my favorite - I've been pushing and working hard to
support this idea for a long time.  I vote Support, I vote double-Support, in
fact I feel strongly enough about this issue to "waste" $10 to cast a triple-
Support vote.  (I only do this because this topic is especially near and dear
to me.)

So under this system, people will routinely vote the way they truly feel, but
will stop short of the pain threshold regarding the cost for each vote.  (If
something is worth $5 to me, I'll likely cast a double-vote for $1 but would
be foolish to escalate up to $10.)

I do not know if such a system would be viable or acceptable - time will tell.
But my point is that it is POSSIBLE to design rated decision systems which
avoid the supposedly "unavoidable" problems with exaggeration.

Mike Saari



More information about the Election-Methods mailing list