[EM] "Proxy ranking" versus "proxy approval"?

[Bryan Ford]

In a proxy direct democracy, when a participant sets up their "standing" proxy 
lists to indicate who gets to vote on their behalf (either in general or on a 
specific topic) when the participant does not vote directly, it seems clear 
that the participant should be able to register multiple alternative proxies 
so that they can prevent their vote being wasted if their "first-choice" 
proxy does not vote on a particular issue.  (How this feature is used is a 
matter of personal choice, of course: a participant is always free to list 
only one proxy, or none.)

But if _more than one_ of the proxies on my list actually do vote, how is my 
vote applied?  In James's proposal the idea is that proxy lists are ranked, 
so my vote goes to my first-choice proxy if he votes (or further delegates to 
some other proxy who actually votes); otherwise the vote "bounces back" and 
goes to my second-choice proxy instead, and so on.  This is obviously takes 
sort of a "proxy IRV/STV" approach.

An alternative would be to take more of a "proxy approval vote" approach: if 
more than one of the proxies on my list actually do vote (or further delegate 
my vote to someone who does), then my vote is split evenly between all of 
them.  Thus in the (hopefully not too common) situation that two of my 
proxies vote against each other on some issue, my vote would be split between 
them and actually "cancel itself out", instead of favoring one proxy over the 
other.

(Note that the issue of "proxy ranking" versus "proxy approval" I'm talking 
about here is completely separate from and orthogonal to the basic kind of 
election system being used for a particular vote: for example, the "proxy 
approval" approach could be used to distribute proxy votes in an IRV, STV, or 
even simple majority/plurality election, while the "proxy ranking" approach 
could just as easily be used to distribute proxy votes in an election by 
approval voting.)

Here are the immediate implications I can think of:

* "Proxy approval" is probably somewhat more technically challenging to 
implement, since it requires splitting users' votes into fractions, which are 
likely to get smaller and smaller and get spread across more and more proxies 
as the proxy chains get longer.  Thus, there is a potential issue of 
implementation efficiency and scalability - but at the moment it doesn't seem 
to be necessarily an insurmountable problem.

* On the other hand, on the assumption that anyone serving as a proxy and 
voting on others' behalf _must_ make their votes and their own proxy lists 
public for reasons of transparency, it seems that proxies will be much more 
willing to register and use proxy lists of their own if they can do so 
without being forced to (publicly) place their political allies in a "pecking 
order" on their proxy lists, and thus perhaps show a kind of favoritism that 
they do not intend.  If instead of using a ranked list, a proxy can merely 
use an "approval list" that indicates all the other proxies they "approve of" 
to vote on their behalf without showing favoritism between them, the informal 
political barriers against proxies themselves using proxy lists may be much 
lower.

* It seems that the voting power of a particular proxy is likely to be much 
less volatile over time and across different votes in the "proxy approval" 
scheme than in the "proxy ranking" scheme, because a given proxy's "voting 
base" will consist of fractions of the votes of a larger number of users, 
rather than a smaller number of whole votes.  This difference will be 
particularly pronounced when proxies use their own proxy lists: in the 
"ranked proxy" scheme, a proxy who doesn't vote on a given issue will 
effectively transfer his whole accumulated voting weight as a unit (which 
might consist of the votes of many participants) to his first-choice proxy, 
whereas in the "proxy approval" scheme such an accumulation will tend to be 
split further.  In the "proxy ranking" scheme, I suspect that small changes 
in a popular proxy's proxy list could create large and perhaps quite 
unpredictable swings in election outcomes, whereas with "proxy approval" the 
changes in outcomes caused by minor proxy list changes are likely to be more 
attenuated.
 
Comments?  Other implications I haven't thought of?

Thanks,
Bryan