[EM] Re: direct democracy / proxy system proposal

[Bryan Ford]

On Monday 16 August 2004 10:24, James Green-Armytage wrote:
> Bryan Ford wrote:
> >What I was thinking is that the stage 2 system would be open to everyone
> >for
> >purposes of participating in discussions (e.g., reading and writing
> >messages,
> >making proposals), and anyone could even act as a proxy and thus
> >indirectly
> >wield the voting power of votes delegated to them - but only those
> >selected
> >by random invitation would actually have a "direct", baseline vote to
> >start
> >with.  That way, people (like us) who are interested and passionate
> >enough to
> >sign up on their own can serve the role of advisors in the system, while
> >still preserving some level of democratic legitimacy by ensuring that
> >these
> >self-selected advisors can only wield however much voting power is
> >delegated
> >to them by the randomly-selected representative constituency.
>
> 	Okay, I've got a good compromise. We can let everyone participate as full
> voters, but we can also distribute little code keys or something to
> people, according to whatever random sample method you want. Then, after
> the vote, we can publish two sets of results: the general result of
> everyone who voted, and the result only counting the votes of people
> selected to be part of the random sample. You like?

Excellent - yes, I like! :)

> >Can we come up with a
> >way to incorporate legitimate, secure democratic "process" into a system
> >like
> >this without making it _feel_ to users like they're being subjected to a
> >deliberative process?
>
> 	This is a great challenge, which we can go on discussing for some time.
> 	One simple idea is to simply allow visitors to the site to create new
> issues at will, which automatically creates a new page. After the creation
> of the page, there could be a certain amount of time for option
> generation, and then another amount of time for voting to take place.
> After the second period is over, the result is final. Either the length of
> the periods could be fixed, or they could be created by the originator of
> the page.

Sounds good.

> 	If you do it this way, the remaining question for me is how to arrange
> the issues hierarchically, so that visitors to the site can look directly
> and conveniently at a few of the more "important" issues (by whatever
> definition), without having to wade through a bunch of trollish or
> redundant issues.
> 	Perhaps, when you create your issue page at your own will, you are
> creating it on the "periphery" of the site, which operates in the more
> free-form manner I just described. But you could also have the option of
> entering it into the "core" of the site, in which case it will be subject
> to the rules which I described previously... but will be likely to get
> looked at and voted on a lot more.
> 	I like that idea... what do you think?

Yes, sounds like a good approach.  It seems that the meta-issue of "which 
issues get the most prominent position and thus the most attention/publicity" 
is itself an issue that should be handled as democratically as possible, 
ideally by more or less the same infrastructure that runs the regular 
deliberation process.  Perhaps, as you suggest, issues generally start out 
somewhere around the "periphery" of the site and then move toward the "core" 
as they accumulate "interest votes" or something like that.  Perhaps the act 
of casting a vote on an issue or adding it to your "bookmarks" or "watch 
list" would implicitly cast an "interest vote".  And if proxy voting applies 
to interest votes in the same way that it applies to regular votes, then 
issues proposed by widely respected people who already wield a lot of proxy 
voting weight would get an immediate head start on the "interest ladder".

> 	Also, I suggest that we might want to split the site into two different
> categories: "politics", and "everything else", e.g. favorite movies,
> favorite episode of saved by the bell, and whatever else people can think
> of.

I think the category system should ideally be fairly rich and free-form, 
allowing sub-categories of categories and so on, as long as the category 
system doesn't hinder system useability by adding too much unnecessary 
complexity.  For example, the site's home page might show sort of a 
slashdot.org-like summary listing of the current top 10 or 20 issues in the 
"interest ranking" throughout all categories; then the user could click down 
into particular categories or sub-categories to see the current top 10 or 20 
issues in that specific category, and so on.  Issues that accumulate enough 
interest votes to reach the "home page" sometime before the vote closes will 
obviously get the most widespread participation, since there are likely to be 
many users who just look at the home page or one or two of the top-most 
categories.

>>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. 
>
>        Okay. Again, I suggest a compromise: when voters fill out a proxy 
>list,
>they can be given the option of equal rankings. Thus, if they really don't
>want to make up their mind between favoring one of two or three people
>whom they fully trust, they can rank them as tied for first place... and
>the vote would simply be split two or three ways.
>        I think that it is important to give people the option of ranking 
>their
>proxies hierarchically. If they trust several proxies equally, then sure,
>they should be able to indicate that. If not, then they should be able to
>indicate that, too.

Yes, that should be doable.  And after thinking some more about how such a 
scheme might be implemented, I realized that the concern I mentioned about 
compatibility with a category system that isn't strictly tree-structured can 
be addressed in a sensible and meaningful way after all.  When deciding how a 
given user's vote should be applied indirectly through multiple possible 
proxy chains, the system first enumerates all the possible proxy chains that 
might apply, in the process eliminating "dead" chains such as cycles and 
chains leading to proxies who haven't cast votes.  Then in a separate phase, 
the system applies the user's rankings and/or vote distribution preferences 
in order to propagate the votes along those active chains in the appropriate 
fashion.  The category system only affects the first phase, and the ranking 
system only affects the second, so it should work out fine.  The upshot is 
that the user's chosen proxy ranking or distribution always applies to (and 
_only_ to) those proxies that are actually "relevant" given the votes cast on 
a given issue.  A bit complicated, but hopefully we'll figure out simpler 
ways to explain it without getting overtechnical like I did here. :)

> 	You are a CS person, eh? Excellent. Just curious: are you in Cambridge?
> I'm in Berkeley.

Yes - Cambridge, Massachusetts, not Cambridge, England.  (At the moment I'm 
actually in Geneva, Switzerland for the summer, but will be back at MIT in a 
couple weeks.)

Cheers,
Bryan