[EM] Why Not Condorcet?

[Dave Ketchum]

On May 15, 2010, at 7:17 PM, Kevin Venzke wrote:
> Hi Dave,
> --- En date de : Sam 15.5.10, Dave Ketchum  
> <davek at clarityconnect.com> a écrit :
>> De: Dave Ketchum <davek at clarityconnect.com>
>> Objet: Re: [EM] Why Not Condorcet?
>> À: "Kevin Venzke" <stepjak at yahoo.fr>
>> Cc: election-methods at electorama.com
>> Date: Samedi 15 mai 2010, 17h34
>> On May 14, 2010, at 7:08 PM, Kevin
>> Venzke wrote:
>>> Dave, by the way,
>>> --- En date de : Ven 14.5.10, Dave Ketchum  
>>> <davek at clarityconnect.com>
>> a écrit :
>>>> We can dream of value in details as we sit here
>> and
>>>> debate.  Real-life voters need a way to
>> express their
>>>> most serious thoughts with reasonable effort:
>>>>       To vote for more than
>> Plurality's
>>>> one - which even Approval offers.
>>>>       To vary their approval
>> according
>>>> to their amount of liking - Condorcet and Score
>> offer this.
>>>>       To ask for only reasonable
>> effort
>>>> from the voters - see Condorcet.
>>>>       Score demands more.  A
>> voter
>>>> thinking of A>B>C>D has no trouble
>> offering min and
>>>> max ratings to A and D.   With
>> Score the
>>>> voter is expected to diligently assign the
>> available rating
>>>> space among A>B, B>C, and C>D.
>>>
>>> I notice that all of your arguments have to do with
>> the expressiveness
>>> and simplicity of the ballot (except when you
>> criticize IRV).
>>>
>>> Some objections to Condorcet could be:
>>> 1. It is not expressive enough (compared to ratings)
>>
>> Truly less expressive in some ways than ratings.
>>      This is balanced by not demanding
>> ratings details.
>>      And more expressive by measuring
>> differences between each pair of candidates.
>
> That is probably the argument I would make more often. Ratings has
> no interest in pairwise contests, which makes its behavior (and ideal
> voter behavior) very different from Condorcet methods'.
>
>>> 2. Offensive strategy potential (absent in IRV,
>> ratings, Bucklin)
>>      How is IRV different?
>
> In IRV your lower preferences are not regarded until all higher
> preferences are eliminated. This means it is completely impossible to
> make your favorite candidate win by lying about your lower  
> preferences.

AND, by your choice of higher preferences in IRV, you can prevent your  
lower preferences from ever being seen.
>
> Perhaps more importantly: It's also impossible to try to do this and
> fail so badly that you elect a candidate no one likes.
>
>>> 3. Lacking guarantees (e.g. FBC or LNHarm)
>>      Isn't this standard among methods
>> - each with different details?
>
> Yes, but different people will value different guarantees.
>
>>> 4. Too complicated to explain, or propose (a
>> conceptual hurdle with
>>> Condorcet is that we leave the actual ballots for the
>> pairwise matrix
>>> right away, making it hard to understand how voting
>> different ways
>>> could change things)
>>      Some Condorcet methods of handling
>> cycles are truly complex - I recommend choosing a method for
>> which cycle explaining is doable.
>>      Counting into the matrix should
>> class as understandable.
>
> It's possible. I do think it would be helpful if Condorcet could be
> defined in terms of how a single ballot "goes through the process." In
> essence Condorcet sucks all the data out of the ballots like a  
> vacuum and
> finds the best winner without thinking about which ballot said what.  
> This
> makes for a pretty good method but it's also what means Condorcet  
> provides
> relatively few concrete guarantees to the individual voter.

It is true there is no record, except for the ballot, as to which  
voters contributed to each pair.

In Condorcet every pair of candidates competes.  From each ballot  
every candidate ranked gets counted among the pairs:
      Among ranked pairs, if ranked higher than its competitor.
      Against other candidates, simply for being the one ranked of the  
pair.

Usually one candidate will win, as CW, for winning (I am assuming  
margins, and assuming no ties) in each of its pairs.  This is partly  
that some candidates are liked more than others, and partly that  
Condorcet treats all said on each ballot as having equal value.

Next possibility is a 3-member cycle among the best liked  - each  
member beats one other and is beaten by one other - deciding which  
wins is based on which solution method is used.

More complex cycles are possible, but expected to be rare.
>
>>> 5. Not thought to be politically acceptable (third
>> place in FPs can win)
>> You seem to be complaining about newness - a problem for
>> any new thought until/unless accepted.
>
> I'm not complaining at all, I'm suggesting reasons of others. The  
> problem
> could very well just be newness, but IRV for instance has this  
> problem a
> little less than Condorcet.

IRV has backers SELLING - some of us complain that they do not require  
truth as a basis for their selling
>
> Kevin Venzke

Dave Ketchum