Some additional points that I missed in my previous mail.
- The second preferences of A supporters are important. B should
maybe be friendly towards A. A strategic plan targeted against A
would lead to some sincere A>B supporters becoming sincere A>C
supporters.
- The second preferences of C supporters are important. B should
maybe be friendly towards C. C supporters are in a position where
they can decide who is the winner (neither A nor B has majority of
the votes, but C voters can make either of them win).
- The strategists have probably do not have any complete list of
ballots available (might have too if they have good relations to the
poll organizer). Making strategic decisions based on the pairwise
matrix or even smaller amount of information (e.g. the order of the
candidates) is not as easy as when knowing the opinions of each
individual voter. (Note also that the sample size in the poll should
be larger than in a typical plurality poll to get reliable
information on the number of voters with some particular voting
pattern (e.g. A>B>C).)
- One form of defence against the strategists is not to give any
information in the polls or to give false information (random or
planned) to fool the strategists. If some strategic voting will take
place in these elections it is possible that in the next elections
people/parties would defend themselves this way. This may be more
efficient than e.g. counter strategies in the actual election.
Hopefully the end result of all this would be that all parties would
recommend sincere voting. That is a good main rule for Condorcet
elections anyway.
- It is also possible to intentionally publish false polls.
One way to demonstrate that Condorcet indeed would be vulnerable to
strategic attacks in real life elections (large, public, with
independent individual opinion formation) would be to present a
plausible real life example situation (e.g. in the form of some poll
results) and then design a strategic recommendation that B party (or
its supporters, or media) could distribute via the media to its
potential supporters. Then one could calculate some possible
scenarios with varying changes in the opinions of the voters (natural
shift, reactions to strategic plans, false polls) and varying levels
of strategic voting (including counter strategies and other
strategies) among the voters.
Juho
On Jul 2, 2008, at 8:51 , Juho wrote:
Ok, election with three candidates only is the most risky from
burial point of view since then it is easy to identify the relative
position of different candidates (one winner (A), one loser but
powerful enough to try the strategy (B), one loser that is weak
enough to be used for burial (C)).
My point is that in real life large public elections strategies are
much more difficult to implement than in theoretical examples on
paper. I'll list some points that indicate why burial is not very
likely to happen or be successful.
- In most cases the set-up is not as clear as in the theoretical
burial examples. There may be more than three candidates. The
relative position of the candidates may be unclear.
- Use of the burial strategy indicates that the runner-up thinks he
will lose. In a situation where there are two leading candidates of
same strength burial would not make sense since the strategy can
not be kept secret and supporters of both candidates would use it,
and that would just elect C. I think I have actually never seen a
runner up candidate that would indicate before the election that he
will lose to the other main candidate (and therefore wants his
supporters resort to strategic voting).
- A positive campaign for one's own candidate (and potentially
negative against the competitors) is thus the typical case. Often
even the C candidate presents himself as a potential winner all the
way until the election day.
- If B is so far from victory that all his supporters may safely
give up hope of winning the election and turn to the strategic plan
he may already have too little supporters to run the strategy
successfully.
- The available polls are inaccurate.
- The opinions are likely to change before the election (when
compared to the available polls).
- The strategic candidate may lose votes due to the plot (depends a
bit on the society and if it approves or disapproves the
strategists). The campaigns of A and C will probably take maximal
benefit of B's bad morale and B's tendency to give up the hope of
victory already before the election.
- The leading candidate may gain extra votes since he is declared
to be the leading candidate before the election. On the other hand
elections are surprisingly often close to 50-50 at the election day
(which would make the strategy again unwise).
- It is not possible to control the voting behaviour of thousands
or millions of voters.
- It is possible that if too many B supporters will follow the
strategy C will win. It is impossible to control the number of
strategic voters when this risk exists.
- Many voters probably will not understand the strategy or are not
listening and therefore will not follow it.
- Also some A and C supporters may follow the strategy.
- Many B voters may not use such strategy since they feel it is not
right (also this may depends on if the society approves or
disapproves the strategists).
- Many B supporters certainly believe that B will win (this kind of
optimism is common and also strategically wise). It is not within
their interest to declare B as a loser and resort to strategic
voting and a corresponding campaign.
- The election is likely to contain all kind of votes. In the case
of three candidates that means A>B, A>C, A, B>A, B>C, B, C>A, C>B,
C, A=B, A=C, B=C. The preference strengths are also different, e.g.
A>>B>C vs. A>B>>C. And the voters may have different rationale and
morale. All this makes the system less easy to manage than in the
clean cut theoretical examples.
- For many voters the risk of backfiring may be a reason not to use
the strategy (the actual strength of C is unknown, possible use of
strategy by the A supporters is unknown etc.).
I'll skip writing a concrete example this time since there is
already plenty of text here. A detailed analysis with the numbers
is complex due to the numerous uncertainties that I listed above.
Juho
On Jul 1, 2008, at 22:19 , Kevin Venzke wrote:
Hi Juho,
--- En date de : Mar 1.7.08, Juho <[EMAIL PROTECTED]> a écrit :
De: Juho <[EMAIL PROTECTED]>
I don't entirely agree. I would rank below my
strategically-determined
approval cutoff (if I suppose the same election could
be held also
under
Approval), but I wouldn't rank that much lower,
and I don't think
other
voters should either.
Two reasons for this.
1. If you rank everybody and are predictably sincere,
burial strategy
by other voters is more likely to succeed against you.
People who
would
use this strategy need to have doubt about what
you're going to do.
Truncating at the (strategically determined) approval
cutoff is
good at
this: The main effect is that voters don't rank
all the frontrunners,
and burial strategy works basically by assuming one
frontrunner will
get support from another.
Also heavy truncation is dangerous since that could lead
even to
extensive bullet style voting (and all the benefits of
Condorcet
would be lost).
I don't suggest truncating any higher than the approval cutoff. It
should
be safe to truncate somewhat lower than that.
Anyway, whether truncation is a danger to the quality of the
results of
the method is quite a different question from whether truncation is a
danger to a specific voter. I think a voter would hardly ever be
harmed
by truncating where I suggest (you're withholding support from
options
that you shouldn't need unless your information is bad), while
they help
(over time) to create a deterrent to burial.
I also think that the probability of successful strategic
voting in
large public elections is very small in Condorcet.
It doesn't have to be successful. Backfiring strategy is a problem
too. A
given voter simply being able to imagine how the strategy could
work is a
problem.
If everyone is giving full rankings I think the latter is too
easy. Don't
worry about the Wikipedia election, just take a simple scenario
with two
frontrunners and a third, weak, undesirable niche candidate. If I
think
everyone will give full rankings, then the only reason for me to
not use
burial against the worse frontrunner is the possibility that his
supporters will be using this strategy against my candidate.
I should not be able to contemplate such a strategy just because
of the
presence of a third, unviable candidate!
In such a scenario, ranking all the alternatives would and should
accomplish nothing (unless you support the third candidate). The only
reason to ask these voters to do that is that it would be nice if
we could
say it worked. As strategy advice in this scenario I think it is
poor.
Kevin Venzke
_____________________________________________________________________
________
Envoyez avec Yahoo! Mail. Une boite mail plus intelligente http://
mail.yahoo.fr
----
Election-Methods mailing list - see http://electorama.com/em for
list info
___________________________________________________________All new
Yahoo! Mail "The new Interface is stunning in its simplicity and
ease of use." - PC Magazinehttp://uk.docs.yahoo.com/nowyoucan.html
----
Election-Methods mailing list - see http://electorama.com/em for
list info
___________________________________________________________
All new Yahoo! Mail "The new Interface is stunning in its simplicity and ease of use." - PC Magazine
http://uk.docs.yahoo.com/nowyoucan.html
----
Election-Methods mailing list - see http://electorama.com/em for list info