[Election-Methods] USING Condorcet

[Juho]

Hi,

Some comments below.

On Jul 11, 2008, at 2:42 , Kevin Venzke wrote:

> Hi Juho,
>
> A quick response.
>
> --- En date de : Mar 8.7.08, Juho <juho4880 at yahoo.co.uk> a écrit :
>> Ok, this is a good concrete example scenario. The votes are
>> of course
>> simplified. Surely there would be also considerable number
>> of other
>> kind of opinions than these three.
>> But let's see first where these simplified votes could lead to.
>>
>> Sincere opinions:
>> 45 A>B>C
>> 40 B>A>C
>> 15 C>B>A
>>
>> It seems that C is an extremist candidate (at the B side of
>> the
>> political map since they prefer B over A). In line with
>> this
>> explanation it is natural that A supporters prefer B over
>> C. It is a
>> bit more strange that all B voters prefer A over C. Maybe C
>> is so
>> radical that all others hate him/her. (The nature of the
>> scenario
>> will however stay quite similar even if there were also
>> some B>C>A
>> voters.)
>>
>> The A party or A proponents have a plan to bury B under C.
>
> I don't know about this. Is it necessary to have a plan? This isn't  
> like
> pushover strategy where only a certain percentage of the faction  
> should
> vote in a certain way, or risk failure. You're not going to be the  
> "one
> strategizer too many."

My thinking was that someone has to at least understand that burial  
could be attempted in this case and then plan how to communicate the  
request to vote strategically to the voters. Well, in principle it  
could be that already during the previous elections voters were made  
aware of the strategic voting options, but I think it is much more  
probable that someone will try to make the decision that according to  
the polls candidate A is now in such a position that burial could be  
attempted. This then has to be communicated (by active media or A  
promoters) to the potential supporters of candidate A. If there is no  
plan and clear communication then the outcome could be that some of  
the A and B supporters would bury and some not (i.e. voters are not  
knowledgeable strategists that could make these decisions by  
themselves).

>
>> What if the A proponents manage to get the required 89%
>> strategic
>> votes or more to support their strategic plan? Then the C
>> supporters
>> can use a compromise counter strategy and rank B first. If
>> more than
>> 5 of the C supporters will use this strategy that will
>> nullify the A
>> strategy even if they manage to get 100% of the A
>> supporters to
>> follow the strategy. Less strategic C voters needed if less
>> than 100%
>> of the A voters will follow the strategy.
>
> Nullification is no consolation though, because there is nothing to  
> fear
> from it. You may as well say that some of the C voters strategically
> vote for their last choice A in order to make A a majority favorite so
> that A voters' strategy does nothing.

Theoretically there could be a society where all voters would vote  
strategically even if they would not benefit anything out of it. In  
most societies strategic plots have however also some negative  
effects (like losing some of the crucial voters with sincere  
utilities A=100, B=99, C=0). Most societies also contain voters that  
will not try to fool the system since they consider that to be foul  
play. There is of course also always the risk of electing candidate  
C. The strategists must thus do careful analysis on the chances of C  
before recommending strategic voting (not a free ride in that sense).  
General burial is not a very good strategy (e.g. 40 A>X1>...>Xn>B, 40  
B>X1>...>Xn>A, 10 ...).

The point thus is that even if there are some voters that might be  
eager to follow whatever strategic options no matter what the outcome  
is there also also many other kind of voters. Also nullification may  
be a reason not to vote strategically to many. Also within the  
election method discussion community some people have indicated that  
they to vote sincerely (e.g. in the US presidential elections) even  
if that is not optimal from a short term strategic point of view.

>
>> If the method uses winning votes then the B supporters may
>> also use
>> truncation as their counter strategy. 30 votes or more (out
>> of the
>> 40) needed. That would be a threat to elect C if A voters
>> will apply
>> the strategy.
>
> Yes. If WV or Condorcet//Approval is used I don't believe there  
> would be
> a problem, since truncating the worse frontrunner would be both  
> natural
> and effective. (Actually, as long as truncation is allowed I believe
> voters won't rank the worse frontrunner anyway, even if you tell  
> them they
> ought to or safely can.)

Unfortunately there might be other problems. The same set of votes  
that demonstrates how truncation works as a defence strategy (45  
A>C>B, 40 B, 15 C>B>A) serves as an example of a case where use of  
winning votes makes it possible for the C supporters to strategically  
bury A under B (sincere votes: 15 C>A>B).

Also the behaviour of winning votes with sincere votes can be  
questioned, as well as the idea of favouring truncation (brings the  
elections closer to plurality).

I try to study the possibility that Condorcet methods are  
sufficiently strategy proof for large public elections (with  
independent decision making) as such. If they lead to strategic  
voting and counter strategies that may already be a sufficient reason  
to use some other safer and simpler methods. I'd maybe expect all the  
main parties to generally recommend sincere voting to consider  
Condorcet to be a good method for that society (i.e. Condorcet should  
be safe enough to allow them to comfortably fall on that side of the  
fence). (What do you think of this statement as a criterion on  
whether Condorcet will work well in the society in question or not?)

I'll skip detailed analysis of Condorcet//Approval. There are some  
problems with either not allowing all candidates to be evaluated  
(with implicit cutoff) or being more complex to the voters (with  
explicit cutoff).

>
>> Theoretical examples on paper give complete information of
>> the
>> opinions and allow complete control of (uniform) voter
>> behaviour, and
>
> My claim is that the former is plausible (at least the information  
> will be
> complete to the extent necessary) and the latter unnecessary, when
> it's predictable that truncation will not be used.

On the first topic I'd expect something like 5% changes in the  
support of the major candidates to be possible before the election  
day. In some situations this may be enough, in some not. Cases where  
that accuracy is enough may be more difficult from strategy  
implementation point of view.

For this reason I recommend using concrete examples from real life  
with explicit votes when studying the vulnerability of different  
methods (i.e. general claims that rely on all details to work out the  
best/worst way simultaneously are not reliable).

On the second topic I expect most strategies to require some sort of  
central control/guidance to make them work. And I expect the voters  
to be a heterogeneous group that typically to some extent also  
objects to strategic voting and to being controlled / told how to  
behave by others. Also people with preferences of different strength  
are likely to behave in different ways. These are only some reasons  
why people tend to build their own logic and make their own decisions.

I guess you don't think that voters should generally bury or apply  
their own knowledge to decide whether to bury or not (i.e. some  
control/guidance needed in public elections to make the behaviour of  
the voters with strategic interests uniform and to make the  
strategies work even partially). If so, then the voters need to rely  
on some external source that tells them that they should bury B under  
C. Also strategy related articles in the media might be inaccurate  
and conflicting (e.g. if you would let the EM people write them :-).

>
>> (One more reason why A voters should not use burial
>> strategy is that
>> if C is stronger than expected then their strategy might
>> also lead to
>> electing C instead of B.
>
> If there's any possibility of this scenario, I don't believe people  
> would
> use burial.

In the given example that line would go at about where the opinion  
division among B supporters is 10 B>A>C, 30 B>C>A. Some B supporters  
surely prefer C over A. The strategic plot of A supporters might move  
some more B supporters to that camp. If candidate A loses some  
supporters because of the plot or for any other reasons that also  
increases the chances of C to win. Also C may gain more votes and  
thereby increase the risk. One must thus count some tolerances based  
on the uncertainty of the polls, and possible changes in opinions,  
and reactions to the plot when estimating the viability of the strategy.

If we would modify the example in such a direction that the strategy  
could be successful without requiring 89% of the A supporters to  
implement the strategy that would probably also increase the risks.  
There is thus a balance between working vs. non working strategy and  
risks vs. no risks.

My request was to identify such real life scenarios where the risks  
would be small enough and the probability of success would be high  
enough. I don't think the given scenario would still be a major  
threat in most societies. Are there others that would be more likely  
to succeed? Or alternatively, are there some general rules that could  
be used by the Condorcet voters to decide when it is strategically  
better to bury than not to bury?

Juho


>
> 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 Magazine 
http://uk.docs.yahoo.com/nowyoucan.html