[Election-Methods] delegate cascade

[Abd ul-Rahman Lomax]

At 03:49 PM 7/22/2008, Juho wrote:
>On Jul 22, 2008, at 14:26 , Michael Allan wrote:
>
>What is btw the reason that there were no arrows forward from the two
>leading candidates in the election snapshot picture in the references
>page? Did they abstain or were their votes (not even their own vote)
>not cascaded forward for some other reason?

Representing a complete proxy structure in a two-dimensional graph is 
tricky. What was done, let's assume, is to not show proxy assignments 
that complete loops. A more full graph would show loops, but not 
those that come from what are effectively "top proxies." Consider 
this a problem to work on: show a predicted mature proxy structure, 
flat, two-dimensional. I would arrange the "voters" -- these aren't 
really votes in the traditional sense, they are assignments of voting 
rights -- in a circle, with votes who have received no proxy 
assignments on the outside, and voters with maximum proxy assignments 
being closest to the center of the circle. So distance from the 
center varies with the number of proxies collected. At the center, a 
proxy would be a superproxy, representing every member who has named 
a proxy. So if we define "proxy rank," PR, as the number of voters 
represented if a proxy votes and nobody else votes, then we can 
define radial distance equals f(PR). I would set a minimum distance 
for a superproxy,  not zero. (because there can be more than one; for 
example, suppose there are two proxies who collectively represent 
everyone, and they each name each other. Then, in the absence of the 
other, each of them represents all. Wouldn't that be a nice 
outcome!!! But, remember, this is in an FA. Representing all doesn't 
mean "controlling all." It merely would mean that you've got someone, 
or several people, who are able to voice a broad consensus pretty 
much as they see fit, with it being likely that actual voting would 
confirm that.)

So, then, the position of the members in the space can be manipulated 
so that the proxy assignments don't cross, graphically. Anyone 
inclined to work out a chart program that would take a list of 
members and a proxy table and generate the graph?



>The behaviour of voter A in the example above may be quite "sincere".
>He likes B. If B forwards his votes to some candidate that A
>considers to be worse than C then A may vote for C directly.

Sure. This is describing delegable proxy as an election method. It's 
a mistake, though, to think of it as that. Rather, think of it as 
setting up a participatory, deliberative process which allows people 
to vote directly or to assign their votes, over as many iterations as 
it takes. I.e., standard deliberative process, election by majority 
vote, say, or possibly by supermajority in some situations. As a 
"voting method" (single ballot, deterministic), it's possibly 
interesting but hardly satisfactory, for it suffers from the same 
problems as all such methods. For starters, a majority cannot be 
guaranteed unless you coerce voters (as they do with IRV and STV in Australia).

>I expect the cycles in opinions to potentially cause repeated changes
>in the cast votes (but since I don't know yet exactly how the voter
>will be cascaded I will not attempt to describe the details yet).

Actually, what I expect is that most voters will abstain, being 
content to leave decisions to their proxies, being those whom they 
have come to trust as being most informed and most likely to cast the 
best vote. So, in the end, an election decision will be made by a 
relatively small number of individuals who are massively trusted. 
It's a parliamentary system, in fact, with highly accurate 
proportional representation -- of the proxy kind rather than of the 
fixed seat, constant vote per seat, kind. (Asset Voting is a kind of 
hybrid, when used for PR with a fixed quota and vote, but if Asset 
Voting is used to create an electoral college, so to speak, with 
these electors being public voters, and having the right to vote 
directly on Assembly matters, as fractional votes based on the quota 
and votes held, it really is both systems but with a defined 
deliberative body that functions much as present bodies, only the 
voting is somewhat different (and, I'd predict, in practice, direct 
votes would not normally be enough to shift decisions. It's the 
exceptions that are important, though, and the fact that 
participation becomes full, whenever electors want that, that would 
make the crucial difference in how people perceive the government.)


>>   (I have to look at this one again in the morning.)  There's
>>a little more detail on cycles here:
>>
>>   http://zelea.com/project/votorola/d/theory.xht#cascade-cyclic
>
>Could you explain what happened in Figure 9? What are the rules that
>keep one vote at five of the candidates (red numbers) but forward
>some of the votes to the next candidate in the ring? I.e. why not
>forward all votes or keep all votes?

I would "forward" all or none, in straight Delegable Proxy. Keep it 
Simple. In fact, there isn't really any  "vote forwarding." There is 
just individuals voting as they choose. "Vote forwarding" is just a 
method of determining, then, the vote value for each vote actually 
cast. In an Asset scheme, however, used for PR, electors, now holding 
multiple votes, would assign specific votes to specific candidates. 
I've proposed that they might try to keep votes assigned in precinct 
blocks, which then has the salutary effect of allowing voters to know 
exactly whom their vote elected. Generally, with Asset, assigned this 
way, one vote elects one seat, together with all the other votes 
making up the quota. If direct voting is going to be allowed, the 
quota should be the Hare quota, i.e, if there are N seats, and V 
voters, the quota (which is exact, not rounded off) is V/N votes.