[EM] Why the concept of "sincere" votes in Range is flawed.

[Abd ul-Rahman Lomax]

At 12:31 PM 12/2/2008, Kristofer Munsterhjelm wrote:
>Are there any other ways of defining a sincere and "non-strategic" 
>ratings ballot? Direct external reference of the sort "I'd pay 
>amount Z to have X elected" fails because of income differences and 
>the nonlinearity of money. Definitions based on expected value do 
>not differentiate between strategic (planned though "sincere") and 
>non-strategic ballots.

Expected value, of course, factors in probabilities, normally. 
However, we could use expected zero-knowledge value.

I proposed something else in another post yesterday. How much would 
you contribute to the election of each candidate? The problem with 
income differences I addressed by noting that we are looking for 
sincere *relative* utilities over a candidate set. I did not examine 
this deeply, but suppose we were each allocated, say, $20 *per 
candidate* which we could donate. We don't get the money back if we 
don't spend it. How would we assign the campaign contributions?

Okay, now lets make it a bit more sophisticated. We list the 
candidates in order of preference. We have, say, $10 to distribute to 
campaign funds. But these are special funds, they are designed to 
help cause one candidate to beat another. This money will *only* 
help, for example, Kucinich beat Edwards; this presumes that Kucinish 
and Edwards are next to each other in the voters' preference list.

If candidates are clones, for the voter, the voter won't waste any 
money in that pair. So the pair are rated the same.

If the voter would put the entire $10 into one pair, that one pair 
consumes the entire range: all candidate pairs with higher preference 
would get $0 and all candidate pairs with lower preference would 
likewise get 0.

But, say, there are six non-clone candidates. The voter would put $2 
into each pair, down the list. This implies a Borda vote (clones are 
simply rated identically, that's my instant "Borda fix.")

However, I'd guess, the voter would -- without any consideration of 
election probabilities -- shift these. Two total corporate shills, 
i.e., the voter buys the Nader argument, All the money goes into the 
Nader vs (second best) campaign, nothing in the Gore v Bush campaign. 
*This is a sincere vote.*

And pretty useless; in most situations, we do *not* make decisions 
like that. We don't consider choices like Gore v Bush as both 
terribly bad, which is what this sincere vote showed. Rather, if 
there are only two real possibilities we "magnify" that portion of 
the absolute utility scale to look for finer differences.

I.e., if the election method -- and we know it -- is saying to us, 
"You silly fringe fanatic, you can vote for your silly candidate if 
you want, but, then, we will totally ignore your opinion. This 
election is going to be won by Gore or Bush, get over it. Now, if you 
want to participate, and help make that choice, vote for one of 
them." Sure. Bad method. But the underlying principle isn't bad, it's 
actually necessary. We only have, in life, the choices that are 
available to us, no matter how much we might prefer something else. 
If asked, we will say, "Yes, I'd rather have a million dollars." But 
that doesn't mean that we would hold out for it and not accept an 
offer of something less!"

More likely, the voter would take seriously bad candidates, and, I've 
suggested a standard, "Would you be ashamed to tell your 
grandchildren that you voted for X?" and group them together. They 
get no pairwise campaign funds. And really good candidates, likewise. 
If you'd be very happy with a candidate, and more than one, why not 
let the rest of the electorate decide between them. But if you have a 
clear preference, nevertheless, you should put some money there.

The trick here is that whatever you add to one pair, you must take 
from somewhere else. If you put money into the Clinton/Obama pair, 
you'll have to take it from another pair, perhaps Giuliani/McCain. 
(I'm thinking of a period before actual party polls and choices, 
purely consideration of the candidates. You'd continue adjusting till 
it looks right. (The Borda distribution, but done with positive and 
negative opinions separately, makes sense to me.)

(I've defined a "positive opinion" as acceptance or approval, a 
willingness to accept a particular candidate instead of having the 
election fail, requiring some runoff or new election. I'm coming to 
think that we really should have an explicit approval cutoff on Range 
ballots, and that positive/negative Range makes this really obvious.)