[RangeVoting] Re: [EM] Re: Condorcet's strategy problem

[Abd ulRahman Lomax]

At 05:15 AM 9/19/2005, Jan Kok wrote:
>On 9/18/05, Abd ulRahman Lomax <abd at lomaxdesign.com> wrote:
>Please don't normalize the Range ballots.  There are several reasons
>why people might not vote the full range:

Indeed. However, we can't have it both ways. If we don't normalize, 
there are going to be objections that somehow Range voting violates 
one-person, one vote, even though what is really happening is that 
some people cast fractional votes. In Range, full rating is 
equivalent to a regular Approval or Plurality vote.

>- Truly considering the candidates to be nearly equal.  Sometimes,
>when I am in a group that is trying to make a choice, and I am asked
>which alternative I prefer, I may say, "I weakly prefer A".  What I
>mean by that is, if several of us weakly prefer A, but a few people
>strongly prefer B, then I don't mind if B wins.  I'm willing to give
>others, who have a stronger opinion, a larger vote.

Fine. However, you are deliberately casting a weak vote. And then the 
critics come along and object that there was this "vote" that was not 
equally counted, that violates one-person one-vote. Of course, it doesn't.

But the instructions, in any case, should be crystal clear. By the 
way, one way to make it explicit that one intends to cast a weak vote 
would be to note that anyone can enter a write-in vote and give that 
vote maximum rating. That vote could be anyone, or what is written in 
could be "None of the Above." This would cause normalization to have no effect.

Normalization should be done anyway in the sense that the maximum 
vote should be counted as "1".

>   Similarly, when
>I'm asked to rate something (a course, a seminar, the food in a
>restaurant, etc.), if it's not exceptionally good or exceptionally
>bad, I will give it a medium rating.  A vote such as A:3 B:2 might be
>a sincere vote.  The people who vote A:3 B:2 won't be upset if B wins.
>But the people who vote A:0 B:5 would probably be anguished if A won.

Something has been overlooked here. Let's make the vote report a 
little more complete. The weak vote as A:3, B:2, C:0

And the result of this, without normalization, could well be that C is elected.

A much better way for the voter to vote who thinks this way would be 
to vote A:5 B:4. This has exactly the same effect in the A:B pairwise 
election as 3:2, but it has the effect of causing A and B to be preferred.

However, if there is really a chance that C would be elected, the 
voter should probably vote A:5, B:5. This *is* Approval voting, after 
all. The same strategic considerations apply.

>Because of that, fine-grained Range Voting produces better social
>utility when people vote sincerely, compared with voting
>strategically/Approval style.  Why would you want to _prevent_ people
>from voting sincerely?!

Because, in a context where a significant number of people vote 
insincerely, for the ones who would be sincere to vote sincerely 
would be to grant effective control to the insincere.

It is a great error to attempt to use a voting system as a 
fine-grained expression of opinion. To do this probably requires the 
inclusion of what might be called polling information on the ballot. 
For example, consider what might look like a Range 6 Ballot:

Approved:       5 Favorite
                 4 Approved
                 3 Less Approved
                 2 Marginally Approved
                 1 Disapproved
                 (0) Strongly Disapproved.

(Or one might use a different Range granularity, Perhaps Range 3 is 
the simplest:

Preferred
Approved
(0) Disapproved.

Once again, we run into trouble when we try to think of a vote in an 
election in a partisan environment as an "approval poll." If we don't 
deal with this issue, and somehow we manage to get Range implemented 
under conditions that look anything like the present, the backlash 
will be quite damaging. It could seriously harm the future prospects of Range.

I've seen Approval used in what started as a polarized environment. 
Had a vote been taken before discussion, the result would have been A 
by a substantial majority. However, there was discussion, and a list 
of options was prepared. A poll was taken and there was one option, 
B, which was considered "acceptable" -- not preferred necessarily -- 
by nearly everyone, whereas about one-fourth of the people did not 
consider A acceptable. It was only one person, in fact, who did not 
approve B. And then a vote was submitted that B be implemented, it 
was seconded and it passed with *unanimity*. That was a group of 
about fifty people, by the way, quite diverse.

So-called "sincere Approval," and Range is simply more fine-grained 
with higher numbers of options, is great for polling, especially 
where the poll is going to advise the people who are themselves 
voting. If you don't vote sincerely, you get bad advice coming back. 
Not a great idea. But as soon as you tie a fixed result to the poll, 
i.e., it is an election with binding results, and unless the 
environment changes radically (which I'm working on, by the way. How 
optimistic should I be?), there is substantial pressure to vote "insincerely."

This issue does not really come up with basic Approval. It is clear 
in that method, where you only have two options, Yes or the implicit 
No of not marking the ballot, that you are rating candidates based on 
a criterion that you have yourself set based on the set of candidates 
running for office. You pick a set to Approve and a set to 
Disapprove. In doing this, you are choosing to aid the election of 
all candidates in the set you have approved, and to act against the 
election of those you have not approved.

This is an action, not an emotion or an expression of some kind of 
absolute approval. Frankly, we should not call the method "Approval." 
We should just call it something like "Alternative Voting." In 
Approval, votes are votes quite like what they are in plurality: an 
action. Not a speech.

In Range, the situation has not really changed; the only difference 
is that *intermediate* actions are possible. Again, these are 
actions, and they are expressions only of relative approval. Not of 
absolute approval. So when you fail to vote the full approval for at 
least one candidate, you are partially abstaining from voting. Many 
voters will fail to understand this unless it is made very, very 
explicit in the instructions.

>- As you say, "some voters would not understand that they are half-way
>staying home if they don't vote the maximum range".  Well, if a voter
>doesn't understand that, and the voter ignores the directions of his
>favorite candidate to "Vote 5 for Favorite, and 0 for all others", do
>you really want to normalize his vote so it has the same power as
>other voters?

Yes, if it is necessary to do this to counter the objection of 
violating one-person, one-vote. If the votes are not normalized, it 
will have happened that some voters are casting a full vote, and some 
are casting less than that, perhaps without understanding it. This 
*is* a violation of 1P1V. It could be argued that the violation is 
not harmful, but, definitely, it will make implementation more 
complex politically.

The 1P1V objection is already the number one objection we see to 
Approval and Range.

>- The voter may fail to give the highest rating to a candidate because
>of some sort of error.  For example, punching the wrong hole in a
>butterfly ballot.  Or not punching the hole out in a punch-card
>system.  Consider what would happen if a voter intended to vote A:5
>B:1 C:0, but failed to punch the 5.  Normalizing the ballot would give
>A:0 B:5 C:0.  It's unfortunate that A lost that 5-vote, but promoting
>B from 1 to 5 just makes the problem worse.

Grasping at straws.... a ballot error can result in a miscounting of 
a vote, no matter how you slice it. It is misleading to talk about 
this as a 5-vote. Each voter casts at most one vote. The vote is 
actually for an option that creates a rating. It is the options that 
will be reported.


> > The raw ballots would
> > still be available for informational purpose, so if someone was
> > voting, for example, their "favorite" as, say, 2, and the rest as 0
> > or 1, their statement would not disappear. But the ballot would be
> > normalized, i.e., the votes would be counted in this way:
> >
> > 800 votes:
> > A: 5
> > B: 3.333
> >
> > 200 votes:
> > A: 0
> > B: 200.
> >
> > Totals:
> > A: 4000
> > B: 2666
> >
> > Looks better, doesn't it?
>
>No, I don't think so.

Ah, I can see that Mr. Kok would not do well in an Approval system. 
Given the election totals, and with anything like present electoral 
conditions, the result was obviously lopsided. You have an 
environment where 80 percent of the voters prefer (by a substantial 
margin, by the way) A to B, and then we have 20% of voters who claim 
to completely reject A and prefer B. But that kind of polarization is 
almost certainly not sincere. So the result of B winning is actually 
horrific. I didn't make up the example. And examples like that will 
quite certainly be proposed by the opponents of Range.

I think it is necessary to understand something about politics: you 
can have the best argument in the world, but if your opponent comes 
up with a plausible argument that looks devastating, you are on the 
defensive. If your response, even if it is logically airtight, takes 
more words than the objection, you've probably lost. You will look 
like a sophist, trying to confuse people with complicated arguments. 
Think about how the arguments are going to look to unsophisticated 
voters (or legislators).



> > ...To exercise a full vote under
> > Range, you must rate at least one candidate at maximum.
>
>And at least one candidate at minimum!

Yes.

>Some people are working to implement Condorcet in Washington State.
>Rock Howard did some work to try to implement Approval Voting in
>Texas.  I'm willing to leave them alone.  I'd like to stake out a
>claim on Iowa and New Hampshire for Range Voting.  That leaves 46
>states up for grabs.

The Condorcet movement in Washington State is actively considering 
other options besides Condorcet, at the moment. There is some problem 
with the moderator of the Condorcet list, who, it seems, is trying to 
some extent to impose a relatively narrow focus. It is unclear how 
much of this is him, and how much it is the preference of the 
legislator being advised. I do know that specific advice regarding 
how a legislator might use the results of extensive deliberations 
such as take place in the list, without being overwhelmed with the 
traffic, was rejected by the moderator....

Gatekeeping is a very important function, and it is one position 
which I think should only rarely be entrusted to a single person, 
except where the rules are quite explicit and clear.