[EM] To Condorcetists:

Kristofer Munsterhjelm km_elmet at lavabit.com
Sat May 19 04:43:31 PDT 2012


On 05/15/2012 09:10 PM, Michael Ossipoff wrote:

> You continued:
>
>> Before you talk about a *need* to literally maximally help the Democrat beat
>> the Republican, consider what you have said yourself, in response to my
>> posts. You have said that the voters' overcompromise is a result of their
>> history with Plurality, not an objective *need* to, within Condorcet,
>> rearrange the preferences or the worse guy will win.
>
> [endquote]
>
> They do so because of Plurality history, yes. But their need to
> overcompromise is the result of a subjective choice, not an objective error.
>
> In fact, under certain circumstances you, too, would favorite-bury in
> Condorcet. I certainly would. (contrary to what I've said in the past, I
> admit)
>
> Suppose that it's a u/a election. The method is Condorcet. It's pretty much
> certain that Compromise, an acceptable, but not your favorite, is the only
> candidate who can beat Worse, an unacceptable. What do you do? You rank
> Compromise alone at top, that's what you do.  As would I.

In an Nader-Bush-Gore type of u/a, I would say like this:

- Okay, this is an election in the inception of a transition of the 
method. Therefore, it is u/a. If this had been some time after the 
people had got used to the new method, it would no longer be u/a.

- Therefore, there will be a few "frontrunners", perhaps two or three. 
These are the ones that will end up in the Smith set in the worst case - 
certainly none of the minor parties will.

- Since I can't push someone off the Smith set by ranking someone not in 
the set above him, and Favorite isn't one of the frontrunners, I'm free 
to rank Favorite first.

- (alternately) If Favorite and Compromise are in the Smith set and 
Worse isn't, then I should make it count: vote Favorite first so as to 
help maximally against Compromise.

- Only when all three are or will be in the set might I gain a strategic 
advantage by betraying Favorite, and I might not even need to (depends 
on u/a FBC). Here I *think* the probability would be so low that I 
wouldn't betray Favorite. However, that might sound like 
rationalization, so I'll explain at the bottom*.

Now you might say that this is cheating because I'm refining u/a further 
to a Nader-Bush-Gore situation. But consider your reasoning for a 
moment. You say you're concerned about voters favorite-betraying in u/a, 
and you say current political elections in the US are u/a. So I don't 
think I have to consider u/a elections with n-way races n>3 because by 
the time that many parties would be viable, people would have got over 
their overcompromise anyway.

> But, maybe you _don't_ know that Compromise is the only acceptable who can
> beat Worse. Maybe you have no idea which one can beat Worse. Then what do
> you do? You top rate all the acceptables.  The problem, of course, is in the
> majority of circumstances, when it's somewhere in between those two
> circumstances.

Only if they're all in the Smith set, and then only if you want to 
escalate - to bet the advantage after you push the method to weirdness 
is greater than the loss at doing so in the first place.

> You continued:
>
>> And finally, I'd give this hint: the moment it feels like the "other side"
>> has somehow acquired a preponderance of people in denial, take a more
>> Copernican view. When an otherwise sensible group holds a view that seems to
>> be silly, and to explain the silliness, a greater part of that group needs
>> to be extraordinarily blind (and very specifically so), perhaps they are
>> not. Perhaps, instead, the view is not so silly.
>
> [endquote]
>
> It's hardly rare for a majority to be mistaken. It's common.
>
> You rely too much on polling. As I said, the configuration of advocacy on EM
> is but a snapshot of something that's constantly changing.

I would hardly call the discussions on EM mere "polling". If they are 
"polling", then your discussions with the favorite-betrayers upon which 
you build most of your idea of FBC's necessity is also mere "polling" -- 
and polling with a much lesser sample size at that.

> As I've mentioned, Approval won the most recent EM poll on voting systems.
> Approval won by every method that we used. Approval was the CW, the Approval
> winner, and the Range winner.
>
> In the short list of Declaration signers, more people mention Approval than
> Condorcet, even if you count VoteFair as Condorcet. And one of the people
> who mentions Condorcet ranks it below Approval.

True enough. You keep returning to Plurality vs Approval. My point, 
however, was that if otherwise reasonable people just so seem to happen 
to have a big hole where their judgement of Condorcet vs Approval is 
located, then it has to be awfully specific (and convenient) for them to 
be thus blind.

And if favorite betrayal really is so rampant, then it is strange that 
so few other American EM participants have mentioned the need for 
absolute FBC to guard against it. You don't make it out to be a 
contested issue like say, left vs right, but rather something that is 
obvious: something where it would be easy to see that the voters will 
keep on overcompromising so nothing less than a guarantee will do. So 
where's the consensus?

(Incidentally, just so I'm clear about this, if the choice was Plurality 
or Approval, and nothing else was possible now or ever, then of course 
I'd pick Approval. Even in the worst case where backsliding causes a 
repeal of Approval, that just returns to Plurality - no worse than if 
I'd picked Plurality in the Plurality vs Approval contest. I just think 
we can get a real improvement beyond Approval too - and it's better to 
get it right the first time than have the opponents go "you got Approval 
and now you want *more*?" or have backsliding lead to problems for 
reform in general.)

>
> Mike Ossipoff

* Over here, we have party list PR based on Plurality. This is 
essentially Webster's method - there is a slight bias in favor of large 
parties, but that doesn't enter into my point here.
What is important is that there is somewhat of a wasted vote phenomenon: 
if a party gets n votes, needs k votes to get one seat and r votes to 
get two, and k < n < r, then the excess (above k) is "wasted". The 
voters could have helped another party get a seat instead.
If we were to be game-theoretically rational, we'd try to find the 
compromise party that is closest to our views while still having a 
chance of gaining more seats. However, I don't vote that way and I doubt 
other people here do as well -- at least unless their honest favorite is 
one of the radical parties who get a single seat nationwide if they're 
lucky.
Perhaps I am just an example of a voter post-overcompromise; but even if 
that is so, it shows that the sort of "instrumental vote or nothing" 
reasoning doesn't have to be pervasive. Voting mostly honestly doesn't 
even have to be unstable.




More information about the Election-Methods mailing list