[EM] Rob Brown's "repair" of range voting

[Abd ul-Rahman Lomax]

At 12:30 AM 12/7/2005, Warren Smith wrote:
>If you insist on this sort of fix, then what you would want would be
>a renormalization via a 1D linear transofrmation intended to cause
>the max score to be 100 and the min score to become 0, among the
>scores the voter offers.
>
>My feeling is this "fixed fix" is worse than unrepaired range voting
>since
>   * more complicated
>   * no longer will work with today's voting machines
>   * prevents voters who intentionally want to downweight their 
> vote, from doing so
>
>But it does have some adherents.

Normalization is automatic with Range2 (i.e., ratings of 0 and 1). 
The question is whether or not it should be used with 
higher-granularity Range. My thinking has gone back and forth on 
this; my ultimate conclusion is that normalization should not be 
automatic. Should there be a ballot option?

Probably not. But the ballot should make it clear that to have 
maximum impact on the election outcome, one should include, in the 
series of votes for a single office, the extremes.

Preventing "weak" votes actually defeats, at least partially, several 
of the benefits of Range, which is precisely the allowance of the 
expression of weak preferences.

If you don't care which of A or B are elected, you can vote the same 
rating for A or B. If A and B collectively are your favorites, then 
you can give them both the maximum rating.

There is no need for elaborate Vote Dilution options. I'm not even 
sure what that would mean, with the suggestion given. Your vote would 
be multiplied by the dilution ratio? Why?

Range allows the direct expression of Utility values. If the voter 
simply votes at least one option as maximum, and one as minimum, 
their vote has as much impact as anyone else's. (What counts in each 
pairwise race is not the absolute value of the vote but the 
difference between the vote for each pair.)

If we had an enlightened electorate that always voted according to 
its perception of social value, Range would be an ideal amalgamative 
method, taking advantage of the distributed intelligence of the electorate.

But people, under present conditions, will exaggerate. It is enough 
to counter this, I think, that voters know to vote the extremes when 
it matters to them. So you wouldn't, in Range10 with a two-candidate 
election, vote 1,2 (9 would be the max vote). Unless you trusted most 
of the rest of the electorate to vote honestly. So you would vote one 
as 9 and one as 0, to maximally express the difference between the 
two candidates.

However, it is always interesting to ask the question, "What if 
everyone else does as I do?" After all, more often than not, they 
will. (Not necessarily for you as an individual, but for most people. 
And we tend to be more like others than we often think.)

If everyone votes Range in a balanced way, sincerely after thoughtful 
consideration, not exaggerating, you end up with an optimal election 
method, actually.

If everyone votes extremes, you end up with Approval, which is better 
than what we have but which is not very good as a method for 
amalgamating intelligence, the bandwidth is far too narrow.

As I've mentioned, ballot design could make self-normalization the 
normal behavior. There is nothing wrong with that.

It is what happens with intermediate candidates that is interesting....