[EM] Decision rings in cascade voting

[Michael Allan]

Raph Frank, in "Delegate cascade and proportional representation":
> Michael Allan:
> >  Here the central delegates withdraw their votes from an incumbent
> >  office holder (root), in order to form a loop with each other...
> 
> Well, if my rule is in effect, then candidates can start negotiating
> with each other in order to decide on a root node.
> 
> If there is a small number, maybe they would just arrange a meeting in
> a 'smoky room' and decide which one of them they will all vote for.

That would be risky.  They might lose electoral support while in the
smoky room; or soon after emerging, and announcing their decision.
The freedom to shift votes is not restricted to the central delegates.
Their own sub-delegates may withdraw support, leaving the central
delegates suspended in mid-air (a smoke ring?).  The tree would then
break into subtrees.

To prevent the structure from completely unravelling at that juncture,
the sub-delegates might form decision rings of their own, each at the
base of a subtree.  Each sub-ring would then decide on a new subtree
root, to become a new central delegate for the larger tree.  These
central delegates would then meet in yet another decision ring at the
base of the newly re-joined tree, to engage in a second attempt at
holding it all together.  (I would be better, of course, not to let it
fall apart in the first place.)

> If the rule is that the candidate who brings the most votes into the
> ring being declared the winner/ring's representative, then there is an
> incentive for members of the loop to contract the loop.
> 
> For example, if part of the loop is
> 
> ->A->B->C->D->
> 
> B and C could talk to each other (C is B's delegate after all).
> 
> They could then reorganise to give
> 
> ->A->B->D
>      ^
>      C
> 
> This means that B is now considered to to have brought in its own
> supporters, but also C's supporters.  This means that their 'faction'
> in the loop has increased in standing.
> 
> Alternatively, B could ask A to redesignate C as his delegate.
> 
> ->A->C->D
>      ^
>      B
> 
> This is the same as before, except that C is the one that remains in the loop.
> 
> Probably, which one to use would be decided by which one of B and C
> has the most votes (and also would require that A consent to use
> option 2)
> 
> Eventually, that process could reduce the loop to a single person.

But it's not much better than smoky rooms.  It does not ensure the
retention of electoral support at every step.  The electorate is apt
to reject the final decision (and decision makers), on the grounds of
exclusion from the process.

To ensure continuity, it seems better to evaporate the ring by
ejecting the *least* tributary members, one by one.  As long as the
process is predictable in its rules, and slow enough to allow for
considered responses, then the electorate will be able to participate.
They can shift their votes, and thus effectively decide who is to be
ejected from the ring, at each step, and who is to remain.

Here are the incoming votes (tribute in 1000's), to a ring at the base
of a continuous election.  If C remains the least tributary at the end
of the week, then C will be ejected from the ring (B will recast for
D).  Then C could recast for any member of the ring, and participate
freely in the decision process.

        50k   90k   30k   70k

         |     |     |     |
         v     v     v     v

 (D) --> A --> B --> C --> D --> (A)
  

The members of the ring will not participate by vote shifting.  Their
votes will be locked in place, holding the tree together at its base,
in lieu of a root.  Instead they will participate by outlining their
intentions for the upcoming term of office - communicating the
structure of power, as they forsee it.

So decision rings can serve to guide the structure of assent, and the
structure of power, into a new alignment; and to re-knit the fabric
where it was torn by the discontinuities of fixed terms of office, and
so forth.  I can forsee a few problems:

 1) Membership in the ring cannot be forced.  So collusion in smoky
    rooms may result in one member requesting early ejection, in order
    to vote support for another.

 2) A sudden vote shift at a decision cusp (just prior to an ejection
    from the ring) could confuse the results.

 3) In a subtree decision ring, the electorate from another subtree
    may shift its votes into the ring, and skew the results.

The first problem is solved by the slowness of the process, which
ensures that any collusion is effectively ratified by the voters.  The
other two problems can be solved by procedural rules and system
supports.

Can anyone forsee other problems?

-- 
Michael Allan

Toronto, 647-436-4521
http://zelea.com/