[EM] Fluffy the Dog and group strategy

[Dave Ketchum]

On Wed, 12 Sep 2001 01:31:55 +1000 Craig Layton wrote:
> 
> Hi all,
> 
> I'm comming in a little late, but I just wanted to clarify one or two things
> in relation to the fluffy
> example.  I don't believe that it invalidates Condorcet methods, which I
> still nominally support.  It was written a while ago and I guess it
> represents me comming to terms with the low utility condorcet winner
> argument.  I have also argued against most of the alternatives to Condorcet
> (except for the Dyadic Approval and Universal Approval methods, which I'm
> still contemplating) so perhaps I'm just a negative fellow :-)
> 
> I think Dave Ketchum argued that the fact that the majority of voters prefer
> fluffy to either candidate and they have voted thus means that there isn't
> anything objectionable about the result.  The problem with this kind of
> statement is that it sounds like a re-statement of the Condorcet criterion,
> which amounts to the argument; "the Condorcet criterion is a good idea
> because the Condorcet criterion is a good idea".

I see it differently:
     Two bitter campaigns have fought to a draw, as shown in the counts.
 Each of their candidates was rated acceptable by a large minority of
voters, and UNacceptable by a majority.
     ALL of the voters rated Fluffy as acceptable - perhaps even knowing
that that Fluffy was a dog, but expressing their great dislike of the
other human candidate.
     This goes with many methods asking voters to list those they
consider acceptable in preference order, and not asking them to list
candidates they do not like.
     Condorcet is one method that can see the above.  IRV, assuming the
above and not an exact tie between the human candidates, will see this
as a race between them and not see Fluffy's popularity.
> 
> This kind of example also brings up a game-theoretic type argument about the
> voters' best course of action.  It has been argued in the past that when
> there is a Condorcet winner, a sincere vote is always the best strategic
> vote.  From an individual agent's point of view, sitting in the polling
> booth, working out how to vote, this is indeed the case.  However, the
> supporters of the other two candidates (not fluffy) would be best served by
> getting together before-hand and working out some kind of deal, probably a
> simple agreement to truncate (only vote 1 for their favourite candidate).
> Deals between factions are possible in almost every type of voting system -
> except for systems like plurality, at least to the extent that it isn't
> possible for two groups of voters to increase both of their chances of
> winning and/or their expected utiltiy outcomes.

Truly the campaigns can encourage voters to decrease their support for
Fluffy - enough of this and Fluffy properly loses in Condorcet. 
However, Condorcet is in the business of what the voters say, not what
they might have said some other day.
> 
> The difference in the low condorcet winner scenario is that the deal is
> low-risk.  Given that each of the main factions don't have a very strong
> preference between the other two candidates, they don't have much to lose
> and the free-rider effect is minimalised.  Now that I think about it, it is
> actually highly likely that many voters will naturally vote this way anyway
> (truncate a ballot if there is no significant difference between the rest of
> the candidates), so perhaps it isn't as much of a problem for Condorcet as I
> originally imagined.
> 
> I might write more about group strategy later.  It is a real problem in
> preferential systems, and maybe a potential problem in methods like
> approval.
> 
> Craig
-- 
 davek at clarityconnect.com    http://www.clarityconnect.com/webpages3/davek
   Dave Ketchum     108 Halstead Ave, Owego, NY  13827-1708    607-687-5026
             Do to no one what you would not want done to you