[EM] Minimizing need for insincerity

[Blake Cretney]

On Fri, 15 Sep 2000, "MIKE OSSIPOFF" wrote:

> 
> 
> I'd said:
> 
> > > In a reply to Markus I spoke of minimizing need for
> insincerity.
> > > When I say that I want criteria that measure how well a method
> > > minimizes need for insincerity, I mean that I'd like to
> > > minimize the degree of insincerity needed, and it would be good
> > > for that protection to extend to as many voters in as many
> > > situations as possible.
> >
> 
> Then Blake said:
> 
> >What these criteria generally represent is the belief that voting
> a
> >less liked alternative over a more liked alternative is an extreme
> >unacceptable strategy, but that voting two candidates as equal
> when
> >this is not your belief is hardly a problem at all.
> 
> Wrong. SFC & GSFC are about conditions under which some
> people, with complying methods, need no strategy at all.

You have a point here.  I was being overly general.  My criticism
only applies to WDSC, FBC, and SARC.  Not to SFC, GSFC and SDSC.

--snip--
 
> That being said:
> 
> Strategically needing to vote 2 candidates as equal when you
> don't consider them equal is unquestionably not as bad as
> strategically needing to _reverse_ a preference ordering. In
> fact, in a meaningful sense, the order-reversal is twice as
> bad.

That seems plausible.

> But Blake is right when he says that, as regards the social
> results of the voting systems, a method that can require you
> to vote candidates equal, but doesn't require you to reverse
> orderings, is much more than half as good as a method that
> won't require you to reverse orderings or vote equal two candidates
> whom you don't consider equal.

Note that I attributed that belief to you.  I did not endorse it
myself.

> That's because I believe it means something if people don't
> have to abandon their favorite anymore when helping a perceived
> needed compromise. For one thing, that means that it takes twice
> as many mistaken compromisers to give away an election. For
> another thing, support for those compromisers' favorite will no
> longer be completely concealed, as it is now.
> 
> To me it means a lot to say that, for the first time, everyone
> would be feel free to vote for their favorite.

You are making an emotional argument based on how you like to vote. 
If all voters felt the same way, then this might be a good argument,
but I doubt this is the case.  It seems to me that the popular
resistance to approval voting is largely caused because people don't
want to vote candidates as equal to their first choice, and are
willing to occasionally have to abandon their first choice to avoid
this.  So although for emotional reasons you think "abandoning" a
favourite is more than twice as bad as insincere equal ranking, many
people, for emotional reasons, may believe that equal ranking is
almost as bad as reversal.

--snip--
 
> >Now, I think that a reasonable test would take all these
> strategies
> >and choice limitations into consideration.  That way, a method
> with a
> >lot of the strategies Mike ignores
> 
> Blake, if there's a strategy need that I've ignored, won't you
> tell me about it?

You're right that you don't entirely overlook the problem of
insincere equal ranking, because you do have criteria that take this
into consideration.  On the other hand, you clearly have criteria
WDSC, FBC, and SARC that do not.  Certainly in isolation, you would
agree that this would be a problem.  That the methods do not measure
level of strategy as a whole, but only certain strategies.

Now, I suspect that you would make an argument along the lines that
if you use all the criteria together, you are checking for a wider
range of strategy, and that therefore, when you evaluate a method
against all your criteria, you can get a good general idea of its
level of strategy problems.

But I deny that this is true.  Consider a method like IRV.  It fails
the equal-rank favouring criteria while approval passes them.  This
is not a general view of its strategy level, but reflects the fact
that WDSC, SARC, and FBC only check for extreme cases of
order-reversal, paying no attention to lesser cases and to insincere
equal ranking.

On the other hand, just like approval, IRV fails your other group of
criteria, since these methods in effect test for the Condorcet
criterion.  By applying this battery of tests you might come away
with the impression that approval is more strategy resistant than
IRV.  Maybe this is the case, but if you look at the basis on which
each criteria makes its decision, you can't say that they provide any
evidence to this effect.

> I don't deny that a great risk of a small violation would have
> to be balanced against a small risk of a great violation.
> 
> Do you have any examples where we have to make that choice?

Clearly approval has more equal ranking problems then other methods,
but also has no order-reversal for first candidate.  These problems
have to be traded against each other.

---
Blake Cretney