[EM] range voting fix try #2 by Rob Brown

[Abd ul-Rahman Lomax]

At 04:33 PM 12/9/2005, rob brown wrote:
>BTW, I'm still waiting for your response my description of how range 
>voting is subject to vote splitting, and about how Duvergers law 
>would apply to Range just as it would plurality.  (in particular I'd 
>hope you'd look closely at my 2000 presidential election 
>example)  Nor has the Range voting camp addressed why they reject 
>the basic tenets of economics and game theory (i.e. people tend to 
>independently pursue their own self interest, and systems optimized 
>for people doing this tend to be the most stable).  Till I hear 
>satisfactory answers for these, I don't expect to be joining a 
>bulliten board dedicated to one election method that I am convinced 
>is supremely broken.

Mr. Brown comes in like gangbusters to "fix" Range Voting, but isn't 
interested in a list which specifically discusses Range Voting.... Weird.

There is a series of assertions about Range which Mr. Brown has made. 
They are not well-founded. If he focused on one of them, we might 
actually get somewhere. But there are so many.... Rather than respond 
to the questions above -- which are out of context and not 
particularly easy to answer without searching back to get that 
context, or they incorporate assumptions, of the classic "when did 
you stop beating your wife" variety.

There is no Range Voting camp. There is a math professor who is 
promoting Range Voting and has done a fair amount of work with it, 
and there are a few people who have tried or who are trying to help 
the effort. There are people who want to discuss elections methods 
endlessly. And there are a few people who actually want to improve 
elections and who are acting in that direction. To act, one generally 
must choose a course of action. Some people, I suppose, live with the 
belief that their chosen course is absolutely the best possibility, 
but others merely recognize that they have to start *somewhere*.

So, instead, I'll focus on a comment from Mr. Brown's first "Fixing 
Range Voting" series.

>To me this approach would solve the massive glaring "gamability" 
>problem with Range Voting, which is that everyone's best strategy 
>would be to simply give all candidates either a minimum or maximum score.

First of all, voting Approval Style isn't gaming the system. It is 
merely deciding to sharpen distinctions. It's a choice that voters 
can make. However, it is far from "everyone's best strategy."

There is an assumption on the part of some EM people that recommended 
voting strategy is that course of action which will maximize the 
possibility of election of one's favorite, or, if not that, then 
one's next favorite, etc.

However, in my view, an intelligent voter is one who recognizes the 
importance of broad consent to government. The goal of elections is 
*not* to choose the favorite of some faction, but to create a 
government by consent of the people. Some argue with this, and, 
indeed, our present systems don't do that very well at all. Condorcet 
methods, while based on a clear intuitive concept, can only go so far 
to reach this.

Just as Approval is black and white, yes and no, and thus cannot 
express nuances, Condorcet merely extends a binary relationship to 
include all the candidates (or a truncated set).

It's true that if your goal is simply advancing your own cause, 
regardless of the effect on others, on society as a whole, you may 
wish to vote Approval in a Range system. The generally recommended 
Approval strategy is to vote for your favorite among the 
front-runners and for any candidate whom you prefer to that favorite. 
This isn't gaming the system, it's simply using the system in one 
allowed manner. If anyone is harmed by it, it is the voter, who, by 
not voting with more gradation, either over- or under-supports candidates.

Range is a system that really appeals to a whole different concept of 
the function and purpose of an election. Indeed, I'd think it good 
that elections be accomplished in two basic steps: a Range poll 
followed by a ratification vote for the Range winner, or if, perhaps, 
there were two candidates tied within a certain margin, a "choose 
one" election.

The big problem with elections is that they are not deliberative 
process. I would, in fact, do away with elections for representatives 
entirely, creating a parliament by a proxy or delegable proxy system, 
and then officers would be elected by the parliament, where a Range 
poll would be an excellent first step. The poll controls nothing; it 
merely informs the parliament of how the candidates are perceived. 
Ratification would take, at least, a majority vote.

I think we need to define the "best" strategy. What does "best" mean 
in this context? If it means that strategy which maximizes the 
expected social utility of the election for the voter, I think it can 
be shown that the optimum Range vote is nothing other than the 
expected social utility, as perceived by the voter, of the election 
of each rated candidate. This is *not* Approval strategy. Rather, it 
would resemble Approval only with respect to the most approved and 
least approved candidates.

Mr. Brown suggested -- his big "fix" -- normalizing Range Votes. He 
used a scale from a negative number to a positive one. That's one 
option that Range workers have considered. In the end, however, Range 
can be counted starting with any number and allowing a vote up to any 
other number. Normalization has been discussed extensively on the 
Range list. I proposed it there, in fact. In the end, I've been 
swayed by the argument that usually works for me: voter freedom. 
Voters should have the *freedom* to weaken their votes. Some here 
think the idea ludicrous, but, as we have pointed out, many already 
do this, in an extreme fashion, having no intermediate choice. Range 
does give voters that possibility. I know I would use it as I felt 
appropriate; those who think it silly are welcome not to use it!

The biggest problem with normalization -- which *requires* a Range 
ballot to express the extremes with respect to at least one candidate 
in each direction -- is that it requires calculation from each 
ballot. With computer systems, that's not a problem. But with 
hand-counted, pencil-marked ballots, as in my town, it would be 
practically impossible. Rather, ballot design could simply suggest 
voting the extremes if one wants one's vote to have maximum effect.

Range votes are simply summed.... They can be counted on standard 
voting machines, if the allowed numbers are relatively limited. I've 
suggested that binary-weighted ballot positions might be used, to get 
the maximum range granularity from the minimum number of ballot 
positions, but it could well be argued that this would be too 
confusing to voters:

Rate this candidate, you may mark more than one box, your rating is 
the sum of the boxes you check.

()1 ()2 ()4 ()8

This would produce Range16 (0-15) with only four ballot positions 
(levers on lever voting machines).

The big controversy, I'd say, among Range students, is what to do 
with blanks. Basically this boils down to the question of whether or 
not to sum or average range votes. If all ballots are used, then 
summing and averaging are the same, merely being different ways of 
stating the election outcome. But if blank ballots are excluded from 
the average, it's a different matter. This makes a blank ballot into 
a true abstention, and Mr. Smith, an initial advocate of doing it 
this way, noted that he often doesn't have the foggiest idea about 
judicial candidates, so why should he skew the results? And this 
argument has gone back and forth. But averaging, excluding blanks, 
would be technically more difficult and I suspect the idea, if it has 
merit at all, is not ready for prime time.

(When it was pointed out that an obscure candidate could win an 
election simply by only being rated by a few voters, suggestions were 
made that, to win, a candidate must be explicitly rated on a certain 
percentage of ballots. Personally, I think it's way too complicated....)