[EM] a response to Andy J.

[David L Wetzell]

On Mon, Oct 31, 2011 at 3:20 PM, Jameson Quinn <jameson.quinn at gmail.com>wrote:

>
>>> I would elaborate on Jameson's sentiment here.  I think this e-m list
>>> will be very willing to discuss your method, but most of us will probably
>>> end up not supporting it in the end.  That's just the law of averages,
>>> since the vast majority of methods ever designed have serious problems and
>>> we're pretty good at picking holes in methods here.  We're also biased
>>> toward simplicity.  And we know that hybrid methods have a particularly bad
>>> track record.  If you did get some of us to support it, it would probably
>>> take months of light discussion and constant revisitation to do so.
>>>
>>
>> I'm a bayesian, not a frequentist.
>>
>
> ...Then you should be used to having to translate from frequentist to
> bayesian terms. When Andy said "the law of averages", he meant that our
> prior distribution for your intelligence is the intelligence distribution
> of people who propose a new method in their first post here. We have seen
> many smart people fail to make a good method on their first public try.
> Perhaps you have good reasons to believe you're smarter than those people,
> but we don't.
>

Well, when you have different info, you are likely to have different
conclusions.

But I'm not on my first try.  I've been studying and dialoguing on
electoral reform almost full time for more than three  years now.  I have a
PhD in Econ and studied under an absolute genius, Warren J Samuels at Mich
State University.  Among other things, I know about the limitations of
rational choice methodologies from him.

>
> In a less-snarky tone: your method is indeed better than IRV in some ways,
> including honest BR and summability. But it is worse in others, including
> simplicity of description,
>

1. It is a little less simple, but not that much.


> strategic incentives,
>

2. Whether one ranks a second or third choice could only back fire if there
was a close race between your first choice and one of your other choices
for being the third place candidate in the first stage.  This isn't likely
and even then, it's safe to say that the third/fourth place candidate in
the first stage has a less than 1/3rd chance to be the winner in the second
stage.

and LNH.
>
3. It's no longer 100%, but it's up there.  Similarly, there'd still be a
chance of non-monotonicity, but it'd be relatively low and thereby unlikely
enough not to affect voters strategies dysfunctionally.


> If your goal is to come up with an IRV-like system which is acceptable to
> IRV activists yet improves on IRV's flaws, I encourage you to try again;
> though I've found this goal too hard for me, you may find differently.
>

dlw: The truth is out there.  For IRV to be used in "big" elections with
lots of candidates, as was tried out in VA election last year, there needs
to be a way to cut down the time it takes to count the vote.  There may be
additional ways to modify IRV to deal with the problem, but I would wage
good money that they're not as simple as the IRV3/AV3 hybrid.

>
>
>> My rule is simple.  2 stages: a limited form of AV to determine three
>> finalists then IRV with only three candidates, which is the case when it
>> typically works best.
>>
>
> This is more complicated than either IRV or approval, since it essentially
> includes both of them.
>
dlw: yes, it is technically a little more complicated.  But it's not a lot
more complicated.  It still counts as a relatively simple election rule,
compared to lots of the other options out there.

>
>
>> On the other hand, I think you would have a very hard time getting IRV
>>> supporters to even consider this method.  They don't seem very open to ANY
>>> changes to IRV at all.
>>>
>>
>> I'm on pretty good terms with Rob Richie of FairVote.  I respect his
>> organizational leadership and their skill at marketing electoral reforms.
>>
>
> So do I.
>
>
>>  He has indicated an openness to my idea.
>>
>
> I am sure he would change his opinion if he realized your method violates
> LNH.
>

I would likewise be willing to wager a good deal that the diff between
circa 90% LNH and 100% LNH is not a good deal-killer, especially for a
bigger election with potentially a large number of candidates.

>
>
>>
>> Maybe they too are biased in favor of simplicity??
>> And I don't think the Condorcet criterion is *that important*, as I
>> think in political elections, our options are inherently fuzzy options and
>> so all of our rankings are prone to be ad hoc.
>>
>
> Criterion-based arguments might be brittle in the face of noisy input
> data. BR-based arguments, including those which attempt to account for
> strategic incentives, are not. In other words: arguing that options are
> fuzzy does not give a license to ignore the flaws in your proposal.
>

dlw: aye, and I do not rely on that arg alone.

For BR-based args, some of my key args I rely on are: the over-reliance on
cardinal utility assumptions and the import of the number of effective
candidates in an election.  I believe based on my understanding of how BR
works that the fewer candidates you got in a BR model that the less extreme
are the diffs between AV and IRV.  IMO, 7 candidates is too much for an
election model, because there really only tend to be at most 4 serious
candidates in most single-winner elections.  It's because of the cost of
campaigning is too high relative to the chance of reward when too many
serious candidates are in the mix...

>
> Jameson
>
> ps. While I spend a lot more words on where I disagree with you, you seem
> like a smart person and I bet I agree with you >> 80%.
>

since you likewise are a smart and sincere person, I'd bet I can get that
to be >>90% in a week or so...

dlw
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20111031/6db4fe65/attachment-0004.htm>