Thanks Juho, for working to make this dialog more useful!
DWK
On May 21, 2012, at 7:36 PM, Juho Laatu wrote:
[Note: Michael Ossipof's message was not a reply to a mail on this
list but to an offline discussion.]
On 21.5.2012, at 23.13, Michael Ossipoff wrote:
I don't know what you mean by "all regular voters".
I requested strategy descriptions that would be intended for real
life elections and normal voters (not EM experts) in such elections.
That means that the strategy shall be clear enough so that normal
voters can implement it.
The above strategy is for voters who perceive the above-desecribed
conditions.
If you want a general strategy for Condorcet, none is known.
Also strategies that do work only under specific conditions are ok.
You just have to write the strategy description so that a regular
voter can see when that strategy can be applied and when not. I
don't require that a "general" strategy should be used (=modify your
vote) in every election. It is enough if there is a strategy that
can be applied reasonably often, and that will clearly improve the
outcome from that voter's point of view. I encourage you to rewrite
the strategy so that it clearly indicates when a voter should use it
and how the vote should be modified. A working strategy for public
elections must be such that regular voters can successfully
implement it (or get strategic guidance e.g. from his party and
implement that strategy).
I have no idea what examples you're referring to.
You identified two of your examples by giving their "characteristic
numbers", 27,24,49 and 33,32,34. I found and commented the latter
one in a mail (http://lists.electorama.com/pipermail/election-methods-electorama.com/2012-May/030400.html
).
If you want to say that one of them isn't good enough, then you
need to clearly specify it, and then tell what's wrong with it, and
why you think so.
I hope the mail was clear enough. Maybe you did not notice that mail
since you did not comment it yet.
Juho says:
That would make the strategy a working strategy (although not
necessarily a strategy that would work often).
[endquote]
Then it wouldn't be a general strategy, would it.
I used term "general" just to indicate that the voter can refer to
that strategy description in all elections and check what it says,
not that it would always lead to a modified vote.
Ok, but which Approval article?
[endquote]
The one that I posted to EM. The one that is at Democracy Chronicles.
Google gave me this: http://www.democracychronicles.com/2012/05/06/problems-current-voting-system-plurality-voting/
. It seems to be the correct one since it talks about Approval
strategies at the end.
Do you mean that you only want your favorite to win? Then, in
Approval, approve hir only.
This doesn't sound like a good strategy. You know well that there
are better Approval strategies.
If you want to maximize your expection, I've told Approval strategy
for that purpose.
What I'm missing is a description of the strategy in an exact format
that can be used by regular voters. (But I think I got one for
Approval at least towards the end of this mail. Only Condorcet
strategy still missing.)
But if you're questioning the assumption that people wouldn't
strategize in Approval, I merely suggest voting for all whom you
like. If you want to, you can strategize. Suit yourself.
Yes, that is what I meant. And I'm still confused with your idea
that people would choose between those two options in a competitive
election.
“Would I rather appoint him/her to office than hold the election?”
What should I provide? I'm willing to be more concrete if you tell
me what you want.
[endquote]
In general, what you should provide is the specifics of what you
mean. You never do that, and no doubt you never will. That's why
talking to you is a waste of time.
In particular, in this instance, you speak of focusing on concrete
practical strategic vulnerabilities. I suggested that you specify
and focus on one.
You say that Condorcet is vulnerable to strategies. I say that it is
not enough to sow that in theory some modification in the votes
would give a better winner to the strategists to prove a practical
vulnerability. I say that in order to prove that there is a
practical vulnerability one must be able to give practical
guidelines on how some strategy can be applied successfully in real
life elections when the voters have some poll information available.
You can pick any vulnerability type that you think is easiest to
take benefit of. If I'd pick one strategy that would be a limitation
to you.
On the Approval side you say that Approval works fine. I say that
there are situations where Approval fails in the sense that the
voters don't have any reasonable strategies. You should pick the
strategy that works in all situations. I'm to point out the
vulnerabilities based on that given strategy, e.g. by providing
examples of situations where that strategy does not work well enough.
I provided at least one problematic Approval example to you although
I did not yet get yet any exact definition of a recommended Approval
strategy that voters could use when planning how to vote.
The direct implementation of better-than-expectation could just
consist of approving the candidates who are better than (or maybe
exactly as good as) the result-merit that you expect from the
election.
I said "(or maybe exactly as good as)" because approving a
candidate who is exactly as good as your expectation doesn't affect
your expectatation. It doesn't matter whether or not you approve a
candidate who is exactly as good as your expectation. You can flip
a coin, or yes, go by how you feel that day.
Ok. Random or free choice in the case of a "tie".
Juho says:
Does "result-merit that you expect" mean the value of the (single)
guessed winner or maybe the weighted average of potential winners?
[endquote]
Answer to both questions: Yes, if that's what you feel that you
know, or have a perception or feel about.
People are not going to determine their expectation in the election
by multiplying the win probability of each candidate by hir
utility, and summing the products. But you can do that if you want
to.
But you know how good a result you expect from the election.
I take that to mean that voters are expected to estimate (at some
level) the value of each candidate and the probability of each
candidate to win, and then estimate what the value of the outcome is
likely to be. They need not do that mathematically. It is enough if
their "general feeling" to some extent approximates the mathematical
formula. (Actually I think that the equation is more detailed since
people have different weights e.g. for uncertainty and risks. But
let's forget that for now and consider this to be a good enough
equation.)
Juho says:
Does the expectation refer to the sincere opinions or does it
include the expected strategic voting too (much more complicated
and cyclic)?
[endquote]
...if you want it to.
Ok. I take that to mean, if the voters take that into account in
their considerations.
It seems that this is finally an exact definition of an Approval
strategy. It is not quite user friendly since I don't expect normal
voters to understand much about expectation values. Actually the
article (see above) contained one classical definition of Approval
strategy that says pretty much the same thing and that voters might
understand better (although you had your doubts about it too). It
was “Would I rather appoint him/her to office than hold the
election?”. Also this may be too difficult for many voters, which is
not good for Approval (many voters will not follow the strategy).
But I'll take this strategy (with either definition) as your
recommended strategy for voters in competitive elections. I think
I'll come back to that in another mail to see how well this strategy
works.
If you have a similar exact strategy description available for
taking benefit of the claimed Condorcet vulnerabilities, that would
be welcome, and actually needed to make the strategy credible.
Juho
----
Election-Methods mailing list - see http://electorama.com/em for list info
