[EM] Re: Condorcet's strategy problem

[Rob Lanphier]

Hi Mike,

I'm going to respond a little out of order.

On Fri, 2005-09-16 at 01:23 +0000, MIKE OSSIPOFF wrote:
> But you haven't shown that the way I would vote in a Condorcet election is 
> stupid. Saying it isn't enough.
> [...]
> Let's not sling namecalling around unless we can show that it's true.

My apologies.  I didn't mean what I said as an insult to you, but I can
see how you took it that way, and for that I'm sorry.

> [Rob] continued:
> > If I like A, I hate B, but I
> > really hate C, do I risk approving B to make sure C doesn't get elected?
> > Sounds like I might end up back in the position of approving "the lesser
> > of two evils".
> 
> [Mike replies]:
> 
> No. Don't vote for an evil, in Approval or Pluarality. But you might well 
> vote for a compromise, in addition to your favorite.
> 
> [...] If there's no obvious frontrunner, there are several other methods. Probably 
> the main one is Better-Than-Expectation:
>
> [Description of "Better-Than-Expectation" and "Threat-Promise"]
>
> There are a few other strategy methods, similar to or related to 
> Better-Than-Expectation and Threat-Promise.

That's my central objection to Approval when compared to Schulze(wv) and
other Condorcet-compliant methods.  In the vast majority of cases, a
simple sincere ranking works in Schulze(wv).  It's what I can do, it's
what I can tell others to do.

With Approval, the strategy has a lot of caveats in what could become
relatively common situations.


> Why is it stupid to do whatever it takes to maximize the probability that an 
> acceptable candidate will win, if the merit difference among the 
> acceptables, and among the unacceptables, are negligible compared to the 
> merit difference between the acceptables and the unacceptables. Under those 
> conditions, what matters is that an acceptable wins, not _which_ acceptable 
> wins. So you rank them in an order that maximizes the probability that one 
> of them will win. That probably won't be sincere order.

For people whose sincere first choice has a shot at winning, it is.  I
suspect that the vast majority of people will want one of the
frontrunners to win.  That's what makes them "frontrunners", after all.

Even if your favorite doesn't have a shot, in most situations, voting
for your favorite won't hurt.  It's certainly not nearly as risky to
vote sincerely for your favorite in Schulze(wv) than it is in Plurality.
Since I've personally chosen to "waste my vote" on a protest vote in the
past under plurality (in at least one dangerously close contest, no
less), I don't anticipate that the much reduced risk in Schulze(wv)
would be much of a deterrent for me. 

I think the only reason why favorite betrayal figures prominently in
current voters' calculations is that I think many folks are skittish
after Bush v Gore (v Nader) 2000.  I suspect that over a long period of
time without incident

 
> My objection to people needing to [reverse order] is not that it isn't safe enough, 
> or that it isn't as safe as Approval. My objection to it is that that voter 
> is falsely saying that s/he likes Kerrry better than Nader. That voter is 
> concealing their liking for Nader. That voter, in Condorcet, but not in 
> Approval, is failing to show full support for Nader. That's what I object to 
> in methods that fail FBC.

It's a matter of degrees.  Under Approval, the voter is saying that both
Kerry and Nader are equally acceptable.  For that matter, they may even
have to say McCain or Giuliani is just as acceptable if it means beating
Bush.

While that may not seem as bad as saying that Kerry is better than
Nader, in my mind, it's only a question of degree of insincerity.

> [Rob continues]:
> > We'll have to agree to disagree, because [absolute avoidance of favorite
> > betrayal] is not my goal.
>
> [Mike replies:]
> Ok, but when people are afraid to show full support for their favorite, 
> "Democracy" becomes a joke.

"Full support" means being able to express which candidate I
like /better/ than any other.

As I've said, absolute avoidance of favorite betrayal is not a goal of
mine.  It /is/ an important means, so it's only with great caution that
I'd recommend a system that doesn't meet FBC.  Given what I've seen, the
FBC problem in Schulze(wv) doesn't warrant throwing the baby out with
the bathwater.  

Rob