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

[Abd ul-Rahman Lomax]

At 01:32 PM 12/2/2008, Jonathan Lundell wrote:
Good point; you're quite right. My claim might be right in the context
>of zero polling knowledge, but not otherwise.
>
>Which is all the worse for Approval.

I responded to Mr. Bouricius. His example was misleading, in fact, 
because the "illogical" vote was a moot one. He clearly did not 
consider X to be a serious challenger to Z, his favorite, and he 
voted for X for reasons other than trying to help X win. He did not 
want X to win, and the only reason he voted for X was because it was 
"harmless." Clearly, he'd have ranked X lower than Z and Y, in IRV, 
though, I'll note, his vote probably would not have been counted, 
since he had, I presume, a candidate who would have not been 
eliminated before X was executed, er, eliminated.

In Approval, his vote counts. All the votes are counted.

The example is actually one of Approval working. The voter set an 
approval cutoff *for the real election* between Z and Y, and the X 
vote was dicta, a statement made to show some kind of irrelevant 
opinion. This is a voter who would have voted for Nader in Florida, 
same arguments, really, except that he had the *option*, had it been 
Approval, of an additional vote for Y.

Now, did this voter know that his vote only for Z might allow W to 
win? Same effect here, with Approval, as was the case in Florida. I 
assume that any reasonably informed voter would know this.

Range methods, and Open Voting (Approval) is one of them, only allow 
the voter to exercise one full vote over the entire candidate set. 
I.e., if you rank the candidates, and place some fraction of the vote 
in each candidate pair, the sum equals one. Same is true for 
approval, except that the vote in the pairs is limited to zero or 
one. In Range, it may be fractional votes. The vote shown by 
Bouricius is insincere, but it is clear that the voter only votes 
this way because the voter considers the vote will not affect the 
outcome. If the voter is mistaken, presumably he might regret it (the 
preferences certainly show that, his preference for Z over Y is quite 
strong, enough that he will risk the election of W in order to 
express it, so Y must be rated, I'd assume, sincerely, probability 
below 50%, zero knowledge, and knowledge would increase the rating, 
if he cares about outcome. So an outcome of X would be quite bad. 
Clearly, he voted this way -- it only makes sense -- to make a 
statement that X, or X's party, has some valuable contribution to make.

Note the mix of sincere and strategic in the description of why he 
rated them the way he did. He stated that he was voting against Y 
because he didn't want Y to beat his favorite. He must think (1) that 
this is a reasonable possiblity -- he said that it was -- and (2) he 
cares very much about that not happening, and, rationally, cares 
about that more than he cares about beating W. Therefore, in fact, 
his preference for Z over Y is strong, and I'd suggest that he quite 
sincerely votes against Y. Normal strategic considerations would 
suggest, unless this is a three way race, he approve Y. Borda would 
have him, effectively, approve Y (vote greater than 50%)

So all we have here is a sincere vote for Z and against Y and all the 
other candidates, with X being an exception for reasons not stated, 
but, whatever the reason for the vote was, the voter only cast the 
voter because it was considered "harmless." Maybe he wanted to tell 
someone that he voted for Badnarik also....
It has nothing to do with the election, unless he was way off in his 
assessment of possibilities, and it elected Badnarik. He was willing 
to risk that.

This is not a problem for Approval at all. Approval does better than 
Plurality in this situation, and if we are going to face off election 
methods, I'd face off Bucklin against IRV, not pure Approval. While 
Bucklin does not comply fully with Later No Harm, it is not likely to 
be seen by voters that way, the non-compliance only takes place when 
a majority is not found, unlike Approval, which only has one round. 
And what goes around comes around. Prohibiting LNH violation allows 
the voter to not hurt his candidate with a lower ranked vote -- 
unless a majority is required! -- but it also prevents other voters 
from helping the candidate similarly. I'd *never* choose IRV over 
Bucklin. Unless I've missed something big!

To summarize this, IRV prevents, unless a majority is required, a 
voter's lower rank vote from "harming" the first rank candidate. 
However, it does this by preventing other voters, similarly, from 
"helping" that same candidate win. We would never respect this in 
deliberative process, or simply talking things over with neighbors to 
make a common decision, we would see a refusal to reveal second 
choices as selfish or partisan, not considering the group welfare, 
non-cooperative.

Neighbors are getting together. Some say, "I want A, but B would be 
okay if others feel strongly."
One says, "I want C!"
When asked of anything else is acceptable, he says, "No! Not unless 
it is certain that we won't choose C. Then, maybe -- I won't say -- I 
might have a second choice."

What would you think? Any method which complies with LNH is enforcing 
this as a rule, upon everyone. You *cannot* express your second 
choice until that choice is eliminated.

It is no wonder that the referee, when reading Woodall's original 
paper that defined LNH, expressed disgust, as Woodall points out in 
the paper. It is a *highly* controversial criterion. Like a number of 
them, it can sound good to someone who has not considered the 
implications, and especially to people soaked in politics, where 
people fight like cats and dogs for their favorite, when the 
candidates that they are fighting for -- and against -- are pretty 
much the same. Our team vs. their team.

It's actually bad politics, and it has to go, and eventually it will. 
It's fine to make refined choices, that's not the problem. It's when 
the overall perspective is lost that we start to see damage.