Re: [Ag Meinungsfindungstool] LD

[Scott Raney <scott AT metacard.com>]

On Thu, Aug 28, 2014 at 4:24 AM, Thomas <entropy AT heterarchy.net> wrote:
> Liquid democracy as delegated voting is deeply flawed as it violates
> more or less all criteria of free and fair elections and facilitates
> oligarchy.

Well, let's not throw the baby out with the bathwater here: I agree
that there are a lot of potential problems with delegation/proxies,
but all are addressable, and in many cases completely fixable, with
the proper engineering.  Certainly having having every citizen be an
informed voter with unlimited time and resources to spend on their
government would be better, but we have to work with what we've got.
And what we've got is a lot of people with very limited time and
educational background who must nevertheless be the foundation of our
government systems.  Which means some form of delegation is necessary.
And since elected representatives are clearly out (for reasons I'm
sure everyone already knows), that leaves temporary and revocable
proxies.

> I've collected arguments against delegated voting (only German, sorry)
> http://wiki.piratenpartei.de/Benutzer:Entropy/BPT132/SMV

A very good list!  And fortunately it translates well enough into
English I can have a stab at addressing some of them (but note that
most of this is already covered in my book):

1) I'm not a big fan of transitive delegation.  For one thing it would
seem to be unstable because you're likely to get loops that could be
extremely resource-intensive to unwind.  It also makes any sort of
power-limiting features very difficult to enforce.  So my proposal is
for a single level of delegation, although there would be a ranked
list to fall back on: If your first delegate fails to vote, or decides
to delegate, the system selects the next delegate on your list, and so
forth.

2) There must be a limit on the number of proxies one individual can
carry.  The exact number is something we'll have to determine
experimentally and will depend on the size of the electorate, but my
thinking is that it would be on the order of 0.1% for a Locality-sized
election (100K to 1M voters).  Small enough so that no one individual
becomes so important that it's worth trying to bribe or intimidate
them into voting a particular way, but large enough that people won't
have to work very hard to find a delegate that is likely to vote the
way they would or have to settle for someone that isn't a very good
match.

3) Delegation only works with radical transparency.  In fact, IMHO,
*none* of these systems work without that feature because the
necessary level of security for a secret-ballot system is impossible
to achieve.  For those worried about vote buying or intimidation, a
search of the literature will show you that this isn't nearly as
serious a problem as most people think it is, even in third-world
countries.  And we can include social engineering features in the
system to discourage this (i.e., having ways of reporting attempts to
disrupt a vote, using screen names primarily and if an individual
attempts to "out" a particular voter by publishing a screen name to a
real name or address, public *real-world* shaming would be the
appropriate response).

4) Voters shouldn't chose delegates solely or even primarily based on
judgments of expertise.  Instead, due to open voting, you can find
someone who votes like you just by examining the voting record (i.e.,
it's "computer dating" for delegates!)  Of course this does require
the individual to vote directly in some cases to have sufficient data
to make these matches.

5) The solution to the "my delegate didn't vote or didn't vote the way
I expected!" problem is a to use the concept of a two-phase commit (a
common computer science technique in database systems).  The first
phase would be all the direct votes.  The second phase would be
available to anyone who designated a proxy/delegate.  They'd have 24
hours (or whatever) to examine their delegate's vote (open voting,
remember) and to override it if necessary.  This has two advantages:
 a) no fear of casting a vote you don't support
 b) increased power for those who delegate to encourage them to do so
rather than simply not vote at all.  For example, if a vote is close,
everyone who delegated would have a strong incentive to take the time
examine the actual proposal and the pro/con arguments and ratings and
cast a direct vote because theirs could in fact be the deciding vote.
Though some will argue that this means we'd be putting our fate in the
hands of the least informed and interested voters, these are *exactly*
the people we want making the hardest decisions because they are the
ones with the least bias.

> It seems German pirates are the only ones analyzing LD in more detail as
> they have most experience with it. I've found only a single English
> article criticizing some aspects of LD
> https://web.archive.org/web/20140706133158/https://arscogitandi.wordpress.com/2014/07/06/on-liquid-democracy/

Unfortunately he starts off bad, claiming there is no evidence that
people are better at judging expertise than in making judgments on
complex issues themselves.  There is considerable evidence to the
contrary.  One example I recommend, which shows that even kids can do
this, is:
Keil, F.C. The Feasibility of Folk Science. Cognitive science 34,
826-862 (2010).
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2897180/pdf/nihms-204247.pdf

> Such a delegated voting scheme appeals especially to authoritarians
> as they can easily transfer much more power to their leaders, put
> pressure on other voters to do the same, and their leaders could easily
> prevent dissenting views from gaining wider attention.

This description actually applies better to *representative* democracy
than any system that allows delegation.  With most delegation systems
the delegate can be replaced at any time either by another delegate or
by a direct vote.  Making this "power transfer" at best temporary, and
almost impossible to solidify into the kind of long-term incumbency
that is particularly problematic in representative government.

> It comes as no surprise to me that such LD systems are
> endorsed especially by clearly authoritarian (highly correlated with far
> left) personalitites among the German pirates.

Careful there: What you're probably describing is SDO (Social
Dominance Orientation), not authoritarianism.  They do share a common
tendency towards prejudice and aggression, but SDO appears on both
sides of the political spectrum whereas as authoritarianism is
primarily (or maybe even exclusively) a right-wing phenomenon (in
fact, Altemeyer even calls it RWA (Right Wing Authoritarianism), a
label I don't use because it's awkward and apparently redundant ;-)

> The original idea of Liquid democracy was a chained recommendation
> system in which experts may convince users but do not vote themselves
> http://seed.sourceforge.net/ld_k5_article_004.html

Ah, another one of those all-in-one systems.  I'm not a big fan of
these and my proposal instead is for a two-phase process, where a
proposed law or policy is worked on in a collaborative environment
with a relatively small group who refine a proposal and add comments
and questions and ratings to it, but the actual voting is done as a
separate phase where everyone participates.  There is no need for
delegation in the first phase: If you're not well-acquainted with the
issues you're not likely to make a useful contribution anyway and so
shouldn't be just throwing your weight around. Delegation is only
available (and necessary) in the second phase.

> similarly my own concept (German only)
> http://wiki.piratenpartei.de/Benutzer:Entropy/Demat
> In such systems delegation (or influence) is about information flow of
> ideas and arguments as opposed to transfer of voting power to people.

This looks similar to the features of my second-phase: The predigested
information from the first phase (all the questions, comments,
ratings, etc.) are made available to the voters to help them make
their vote in the second phase.
  Regards,
    Scott