[EM] using welfare functions in election methods

[Jobst Heitzig]

Dear Stephen!

You wrote:
> Can we confirm that "welfare" is a synonym for
> "utility"?

As far as it is just a name, sure. However, on the group (society)
level, the term "utility" seems to be used most often for the *sum* of
the individual utilities, that is, for a *specific* social welfare function.

> A starting model would be: given a voting system which
> is sufficiently expressive (perhaps a range system),
> assume that everyone just votes their welfare
> directly, and see where that gets us.  [Of course this
> wouldn´t do in practice.]

I agree to the last remark. Personally, I don't believe that a reliable
measure of individual welfare is available in a real situation. Instead,
I think that these concepts are useful for the study of voting systems
in *model* situations where one *assumes* the existence of metrically
measurable individual welfare. For example, we should find out by
simulation how the various methods perform with respect to the Gini
social welfare function.

However, those who *do* believe that range style ballots give reliable
welfare information should in my opinion consider using the Gini welfare
function in their method instead of just summing up the range values.

> Also, there must be quite a literature on the
> application of these ideas to voting.  Are you able to
> point us at one?

That is surprisingly not the case. Only the concept of *individual*
utility seems to be used frequently in voting theory. I was not able to
find a single reference to the use of social welfare functions of the
kind I described. Some confusion can arise because some authors use the
same term "social welfare function" for choice rules, but that is a
completely different thing!

> Is anyone able to write an Electowiki page
> (or point us to an existing place) about the
> relationship between voting and utility/welfare?
> There seems to be nothing on the usual sites.

That would indeed be useful. Perhaps some supporter of Range Voting
could do this, since the Range Voting rationale seems to rely on the
idea of measurable individual utility :-)

Yours, Jobst