[EM] Re: river, ROACC

[James Green-Armytage]

>You would be right if I was to admit that there is anything like a
>"utility function" -- which I do not: I believe that it is absolutely
>misleading to think the utility of an option could be measured in any
>way. 

	I think that even if you don't accept such a strict and tidy utility
function as I provided before (which is obviously a convenient
artificiality for the sake of example), you should at least accept the
notion that voters might feel much, much more strongly about the
difference between some candidates than the difference between other
candidates.
	So, if you replace the utility math of my example with the more earthy
notion that for the Bush > Kerry > Nader voters, the Kerry > Nader
preference is utterly insignificant in importance compared to the Bush >
Kerry and Bush > Nader preferences, the logic holds. The Bush voters have
virtually no incentive against order-reversal (Bush > Nader > Kerry), and
they have a very well defined reason in favor of it. (Probability of Bush
being elected skips up from .1 to about .4 
	(I messed up the calculation before; it's ((1/3 x 9/10) + 1/10), rather
than just 1/3, I think... since there is still a .1 chance that Bush will
be a CW.)

>If it could, people would always have complete preference orders,
>which they don't. 

	Yes, clearly, cyclic individual preferences are incompatible with
defining preferences in terms of cardinal ratings, but I don't want to get
into this again. Even if you stick to your belief in cyclic preferences,
I'm hoping that you can recognize the potential benefits of a cardinal
ratings interface.

>However, assuming there was something like utilities, I still don't
>think the above example is as severe a problem as others for two
>reasons: First of all, as you realized also, there is a
>counter-strategy. 

	Ah, but if the Nader voters prefer the (.3, .3, .4) chance of getting
(Nader, Kerry, Bush) to a (0, .9, .1) chance, then they might not decide
to exercise the counter-strategy.

>Secondly, it is quite improbable that any Bush voter
>will reverse his/her Kerry/Nader preference as you suggest, because that
>will result in his worst option (Nader) with quite a large probability
>(1/3). Do you really think voters will take that risk? I don't.

	That's just my point. If they barely care about the difference between
Kerry and Nader, the risk is insignificant, and there is no great reason
not to take it. There is not much to lose, and plenty to gain. So yes, in
that case, I think that they would be likely to do it.

my best,
James