Re: [EM] my 2¢ on range voting (and other pseudomajority methods)

[Brian Olson]

On Jan 4, 2005, at 11:09 PM, James Green-Armytage wrote:

> Range voting is neither a majority rule method, a supermajority rule
> method, nor a proportional representation method. Therefore, its
> applications are very limited.
>
> Let's say that our voting scenario is a large group of people choosing 
> an
> executive (e.g. a mayor or president), and there is enough at stake 
> that
> people cannot be relied upon to vote sincerely.
>
> I don't think that range voting should be used for this scenario. Why 
> not?
> Because it is not a majority rule system. It is possible for a 
> candidate
> to win in range voting even if they are the last choice of 99% of the
> voters. Unlikely, but mathematically true.

Since you are talking about an election method that indeed should not 
be used, who here has proposed using such a method? I expect that the 
method you are striking down is subtly but crucially different than 
anything anyone was actually promoting.

So, the example in that last paragraph looks something like this, right?
99* 1.0, 0.0
1* 0.0, 100000000.0

But I don't remember any suggestions that we have such unbounded 
ballots and straight rating summation.

There has been some talk of maximizing social utility, some of it mine, 
and if the above example represents a true measure of utility, then it 
poses an interesting ethical and philosophical puzzle. Early in his 
often cited book, Kenneth Arrow asserted that you can't measure 
interpersonal utility. That's why he focused on orderings, personal and 
social. I disagree and say that we implicitly do measure interpersonal 
utility by giving everyone "one vote" and thus we give everyone equal 
utility and an equal share of the social utility.

To that end "cumulative" and "normalized" forms of ratings exist.

Of course, even a 1-100 scale can be so "abused". But if 99 people 
really vote for their favorite with a "1", and the other guy with a 
"0", aren't they effectively saying "I don't care (much)"? Why should 
we be listening to their votes more than they want to be heard?

> In contentious executive elections, I think that we should stick to
> majority rule methods. What is my operational definition of majority 
> rule?
> The winner should be a member of the minimal dominant set (aka GeTChA 
> or
> Smith). A weaker definition would be the mutual majority criterion. 
> Range
> voting passes neither of these criteria, and so I say it has no claim 
> on
> being a majority rule method.

Ah, but a different definition of "majority rule" can most naturally be 
expressed on a rated ballot. The Rule that says a majority of the 
people think that one of the choices is worth having at all. On a 
positive/negative rating scale ballot, a winner must have a positive 
average rating. If no choice achieves this, disqualify them all and run 
a new election.

> Is it somehow better than majority rule? In some non-contentious 
> election
> scenarios, perhaps it can be. But in a contentious election, it is 
> not. If
> a Condorcet winner is not elected, it does not necessarily mean that
> another candidate had a higher utility. More likely, it means that
> supporters of the other candidate pulled off a successful strategy. The
> interpersonal utility comparison aspect of range voting is just too 
> easy
> to hijack for strategic purposes, so that it essentially becomes
> meaningless in a contentious election.

I think you should justify that more in the face of something other 
than free-range voting.

> If one is concerned about the tyranny of the majority, range voting is 
> not
> the answer, because a well-coordinated majority can still guarantee
> victory for their candidate of choice. If one is concerned about the
> tyranny of the majority, one should look at supermajority methods and
> proportional representation methods, rather than at pseudomajority 
> methods
> like range voting, approval voting, and Borda.

I guess it's still a question of what we're measuring to find 
"majority". Any election method finds the choice with the _most_ of 
some quality. That quality comes through various methods from the 
voters. I guess I'm asking for clarification on what exactly it is 
about the process of these other methods that you don't like. For 
example, I have a negative gut reaction to any non-deterministic method 
and I'll back it up with rhetoric like "people should vote, not random 
number generators".

Brian Olson
http://bolson.org/