[EM] RE: [Condorcet] Can we come to consensus?

[Juho Laatu]

Hello James,

On Sep 9, 2005, at 03:56, James Green-Armytage wrote:

>
> Juho, you write:
>> I'm very much in favour of trying to achieve consensus on what
>> Condorcet methods to promote in public.
>
> 	I don't see why this would be necessary. You don't need the consent of
> self-selected internet list participants to lobby for a change in the
> voting system of any particular group. Instead, you need the consent of
> that particular group, or at least a majority thereof. If you are 
> forming
> an advocacy organization, you might want to agree on a party line, but
> this list in itself is not an advocacy organization.
> 	If you want to lobby for a particular method, go ahead and do it; 
> don't
> wait until everyone on an internet list agrees with you. If you want to
> discuss the relative pros and cons of different methods, then IMO this
> list is a good place to do it.

I agree. Although I used term "promote in public" I'm not involved in 
any lobbying and I support keeping this list as a tool for neutral 
studying. I'm thus a strong supporter of "pros and cons" style 
discussion. The practical usability of various methods however falls 
within my preferred scope of the list. (I think discussing or sometimes 
even praising the properties of minmax or DMC or any other method to 
other members of this list should not be counted as "lobbying" here.)

> 	However, now that I've done my share of complaining, I would be 
> curious
> about which methods each list member "supports", or "approves of", or
> "consents to". To that end, I've created a quick and dirty wiki poll. I
> hope that it sees some use. Feel free to stretch the structure of it if
> you like.
> http://wiki.electorama.com/wiki/Method_support_poll
> 	By the way, the method evaluation poll is still ongoing:
> http://wiki.electorama.com/wiki/Method_evaluation_poll

Good, this type of activities are always welcome (although I personally 
often have problems in finding clear answers to the questions if they 
are multiple and detailed and if questions don't exactly match my 
thinking patterns :-).

>> I mean that with sincere votes the best winner
>> may be found outside the Smith set at least in some type of elections
>> (that seek for a nice non-hated consensus candidate)
> ...
>>
>> The basic example and question is if A losing to B, C and D (all 
>> three)
>> by 51-49 is worse than A, B and C each losing to one of the others by
>> 66-33. Minmax picks A as the winner but Smith compliant methods pick 
>> A,
>> B or C.
>
> 	I see no reason to think that the size of the majority against a
> candidate is an indicator of how "hated" they are. If you are 
> interested
> in consensus, I suggest that you use approval voting (or something
> similar) rather than a Condorcet method.

I don't see the connection between approval and consensus to be that 
strong. I think rankings can equally well be used to describe consensus 
- in a different way. I'd say that electing e.g. a candidate that has 
received no first place votes but is ranked second by most voters could 
be called a consensus candidate. Now when I think about this, the 
definition of Condorcet winner is actually quite close to defining a 
"consensus winner" (based on available pairwise comparison 
information).

> Condorcet is IMO essentially a
> majority rule principle rather than a consensus principle.

=> Unanimous majority in pairwise comparisons. I think it also has the 
characteristics that it tends to elect "consensus candidates" like I 
described above. I'm not a native English speaker, so I may not know 
the use of word consensus in English language well enough, but for me 
anything that talks about acceptability to or agreement by all (or 
many) (instead of dominance of the strongest one(s)) is OK (also 
approval).

> Thus I consider
> MMC-failing Condorcet methods to be awkward.

We discussed MMC and Smith in March under title "majority rule, 
mutinous pirates, and voter strategy". This time I asked only if Smith 
should always hold (since that case is easier :-). I assume that your 
earlier position is still valid and you want both MMC and Smith to be 
always respected (tell me if not).

MMC is majority related but I don't derive its validity straight from 
Condorcet. I believe quite firmly that Condorcet should hold in all 
ranking (only) based methods (both in the sincere case and most likely 
no reason to give that up for strategical defence reasons either). 
Since March I haven't made much progress in any direction with my MMC 
thinking. I tend to think that MMC is quite strongly strategy related. 
I.e. it doesn't say much about who would be the best candidate to elect 
but it says that certain voters would have the option to improve the 
result of the election (from their point of view) if they would agree 
how to vote strategically.

It is possible to construct examples where the mutual majority prefers 
candidates that beat each others badly but are still considered all 
good by the mutual majority (sort of clones or own party candidates). 
In this case it would make sense to elect one of the mutual majority 
preferred candidates. But unfortunately ranking based votes do not give 
us that information. Thus I feel that more information would be needed 
to make the judgement whether the mutual majority should be allowed to 
exclude those candidates that they mutually dislike or not (I'm talking 
about sincere utility of the community as a whole, and excluding 
strategic considerations).

MMC has some more sense than Smith, but not enough to drop minmax from 
my favourites list (of methods that provide good utility with sincere 
ranked votes at least).

>
> my best,
> James
>

Still waiting for someone to agree that Smith is not always required 
:-).

BR, Juho

P.S. For those who want, it would be possible to construct methods that 
meet MMC but fail Smith.