[EM] Survey of Multiwinner Methods

[Greg Nisbet]

I don't doubt that Asset has some desirable properties, but trying to
compare it to "conservative" methods like STV is difficult. In particular,
conservative multiwinner methods are more amenable to Bayesian Regret-type
simulations than methods with districts, parties, or candidates capable of
making decisions as well as to more traditional analyses with specific
criteria that a given method either passes or fails. It's hard to apply
either approach to comparing such different methods against each other,
even though it's relatively easy to compare methods within a class (e.g.
methods with parties, methods with agent-like candidates).


On Tue, Jan 8, 2013 at 10:35 AM, Abd ul-Rahman Lomax <abd at lomaxdesign.com>wrote:

> At 03:24 AM 1/8/2013, Greg Nisbet wrote:
>
>> There's some definite motivation for writing the list of criteria to
>> exclude parties, districts, and relying on candidates making decisions.
>> These sorts of mechanisms are not always available (for instance, picking
>> pizza toppings or locations or something of that nature). That's not to say
>> that these methods are never useful or not "competitive" with party-less,
>> district-less, candidate-decision-less methods in some sense ... it's just
>> harder to compare their merits in some sort of "universal" sense without
>> considering very specific factors about the community associated with the
>> election. In that sense, limiting focus to methods that don't impose these
>> sorts of additional structure or requirements on the candidates or voters
>> helps to simplify things rather dramatically.
>>
>
> Sure. It *also* can be expected to produce results that are less than
> optimal.
>
> "Relying on candidates" to make decisions is a narrow view of Asset
> Voting. Rather, Asset Voting delegates decision-making, not to
> "candidates," per se, but electors. Because we are so accustomed to voting
> for "candidates," we overlook this when we think of Asset.
>
> In the Pizza election, obviously the Pizza doesn't decide (but that
> Mushroom-Olive pizza makes a mean argument!). But one could perfectly well,
> in a complex multiple-choice "election," choose, not the result, but who or
> what process makes the choice.
>
> However, there is a classic method for doing this that works perfectly
> well, and it is simply supermajority rule, even consensus process. (If the
> majority feels that the group is taking too long, "We're hungry, dammit!"
> the majority can simply make its own decision, and invite others to join, o
> not.) The real point is that classic decision-making process is
> *interative*. It is not a simple one-shot amalgamation. In classic process,
> someone may say, "Let's get a Pepperon Pizze!" and the group shouts "Yes."
> Supermajority. But then someone says, "I really can't eat that!" And the
> majority will, in a functional social group, back up.
>
> Range voting can be used to speed this up, but the "speed up" only will
> occur with fairly large groups, and especially groups that cannot meet in
> person. In person, Approval is the departure from traditional formal
> process that makes sense. The final key to showing a decent decision is a
> *ratification,* if there is any doubt. I've seen an apparent 97% A, 70% B,
> under Approval, turn into 100% A, no exceptions, when presented for
> ratification, but this was in person. This whole process can take less time
> than setting up and adding up the votes for a Range poll.
>
> Asset can pick a result that would be *completely* unexpected, something
> entirely outside of initial consideration. Asset can be understood very
> simply: instead of making decisions only from ballots, delegate the final
> decision process to representatives, for efficiency. With the pizza
> election suppose that it develops that all but one member of the group
> *really wants* pepperoni, and one can't eat pepperoni. The problem with
> buying separate pizzas is that a smaller pizza, or by the slice, is more
> expensive. However, the group, if it wants to accomodate the minority
> person, could in fact, decide to by two pizzas, one small, and could
> present the contribution to be put in to be sufficient to buy both pizzas
> (if we assume equal contributions, say). Or someone could decide to
> contribute a bit more. Balancing all this, discussing it, etc, could be
> difficult, but if two people end up discussing what to do, it's quite
> possible that 100% consensus could be obtained on a result. Everyone would
> agree that the result is fair.
>
> The universe of possible "candidates" is large!
>
> The complexities of various voting systems arise from the limitations,
> generally, of single-poll pure amalgamation processes. Asset is a *generic
> solution,* and it works without party structures or districts. (Asset can
> work for district elections, but the universality of representation is
> damaged, that's all.) Short of Asset, various reformed versions of Majority
> have been suggested, with the compromise that a runoff can complete with a
> Plurality result. I've shown how a maximum 2-poll system can satisfy the
> Condorcet Criterion, for example, with an appropriate first-poll ballot.
>
> (Esentially, range polls collect the most detailed voter preference
> information, and, with decent rules, can be interpreted as ranked ballots.
> Hence I highly recommend range ballots, which would include Bucklin with
> balanced preference options (standard Bucklin was Range 4, but with a
> missing disapproved rank). Range resolution can be increased for larger
> candidate sets. And Asset could actually be used as a tweak to increase
> majority completion rates.)
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20130108/d50cfa29/attachment-0004.htm>