[EM] ignoring "strength of opinion"

[Abd ul-Rahman Lomax]

At 05:40 AM 12/11/2005, rob brown wrote:
>On 12/10/05, Abd ul-Rahman Lomax 
><<mailto:abd at lomaxdesign.com>abd at lomaxdesign.com> wrote:
> >Voting anything other than 1 or 10 (or whatever the range is) is
> >downweighting.
>
>That is, one's vote will not have the maximum effect on the average.
>But, in fact, one's vote is strengthening the possibility that the
>candidate will have just that score in the outcome. Give the
>candidates the scores you want them to receive!
>
>Voting is not about giving scores.  It is about electing 
>candidates.  Why would anyone care what the score is if it isn't to 
>elect a candidate?

The score is used, perhaps, to elect the candidate, or is used as 
input for a further process.

But the essence of the idea is to use Range as a method of 
amalgamating the public opinion regarding candidates. It is clear 
that the method will work, quite well, thank you very much, if 
everyone votes sincerely.

Now, it's been raised that this could be the "tragedy of the 
commons." Yes, it could be. But it is far from clear that it *would* 
be. We'd have to do a lot more analysis to be able to predict this, 
and even then analysis is limited, we would need actual elections to judge.

But elections, ideally, are about selecting officers, and ideally 
they select the best officer for the job, as viewed by the 
electorate. Elsewhere we have discussed what "best" means, and we 
don't agree on that yet, as far as I can see. Range is a method of 
rating candidates for an office, and of amalgamating the ratings of a 
group of people. It works in other fields. Would it work in politics? 
If you know, you are indeed brilliant. But I don't think you know. 
You are merely guessing based on certain prejudices, prejudices which 
determine your prediction.

>Also, if a voter knows (through polls, for instance) that a 
>candidate is likely to have a score of less than 50, and he, for 
>some reason, wants that candidate to have a score of 50, his best 
>strategy would be to give the candidate a 100, which will move it 
>closer to 50 than giving him a 50.

Again, we come back to the antisocial behavior expected of voters by 
Mr. Brown. Instead of allowing the process to collect his vote, he 
wants to try to control the outcome. He wants to try to make the 
decision himself instead of trusting the society as a whole to make 
the best decision.

At some point in life, if we are lucky, we realize that our own 
opinions are limited. Many never reach this point of understanding....

>   "Give the candidates the scores you want them to receive" is 
> simply bad strategy when you are using averages.

Only if you have a specific goal in mind, *and* you have good 
information about the votes of others. If you don't know the votes of 
others, then your optimum vote is precisely the rating that you think 
the candidate deserves, in the context of the set of candidates in 
the election.

Mr. Brown continually asserts that this is "bad strategy," but he's 
never shown it. It's the core of his assertion, perhaps he should try 
to show it in a coherent way, stating assumptions and how he proceeds 
from those assumptions to his conclusion. I don't think he's done it.

Consider this. I think the candidate should be rated at 50. Yes, if I 
know that a majority of others will rate that candidate below 50, and 
I care about the exact rating, I should increase my rating to offset 
the others. However, there is a critical assumption here, and I've 
never seen it stated before. I must assume that I am in the minority, 
or, more precisely, that my views are not in the center. If I am 
incorrect about this, then my vote of other than 50 will, in fact, 
distort the outcome in the direction of my exaggeration.

This is crucial: to be attracted by that strategy, I must assume that 
the other voters are ignorant or selfish or insincere, while, at the 
same time, I'm quite willing to be insincere myself. Perhaps I think 
they are insincere *because* I'm tempted to be so.

This is an image of an isolated voter, disconnected from society, who 
thinks of the world as mostly "them." Definitely he or she does not 
think of the government as "us," unless perhaps he thinks of himself 
as in power, struggling to avoid losing control to the ignorant 
masses. That's a position to avoid if you care about your future!

>   If you are voting on numbers, for chrissake use median not 
> average ( see 
> <http://karmatics.com/voting/moose-example.html>http://karmatics.com/voting/moose-example.html 
> if you don't understand this ).  Still, using the median isn't 
> going to solve the main problem here, given that they are electing 
> candidates rather than just picking numbers.

There are many possible ways to use Range data. Averaging is one. 
Averaging discarding outliers is another which I've seen work well. 
And median is not bad at all, at first blush, but I'd want to 
consider it further. Median disregards the strength of expression of 
the extremes. That might or might not be good....

>However, yes, in an "I want my way" environment, one may wish to vote
>extremes. But there is a cost to this, which is that the ballot
>becomes an inaccurate expression of the voter's views, and that can
>have undesired consequences.
>
>
>Totally agree that it is bad to have the ballot be an inaccurate 
>expression.  But as long as you continue to insist upon a method 
>where "accurate expression of views" is diametrically opposed to 
>"optimum strategy", you can expect that.

Mr. Brown has never defined "optimum strategy" in a way on which we 
could agree.

>I put forward the DSV concept as a solution to this (I didn't know 
>it had already been invented, but it doesn't surprise me that it had).

I missed that. What is DSV?