Maps for ClaraT

[arendt]

Trying to post a JPEG file

map2XGA.jpg (sorry - looks like same as before) I don't know how to get
non-website hosted images into DU.

And here's that file you didn't get.

N-K GOVERNMENT
by arendt

1. We need a new paradigm for structuring our government

1.1 Miserably incomplete and lame summary of NK model

The NK-model is fundamentally a spin-glass model. Consider a system of
N components. Let each component be affected by K self+other components, where
K < or = N. Now try to maximize some "fitness" function on the N components.
In the case where K=1, the fitness function is simple: maximize the fitness
of each individual component. The globally most fit system is the one with
the largest sum of component fitnesses.

ASIDE: This is like money economics, where everything can be reduced to one
measure (i.e., money), and where a thoroughly selfish player is free to ignore
his impact on all other players. Classical economics "proves" that this kind of
behavior creates the invisible hand. And, if you accept the initial hypotheses
about single metrics and non-interaction, you are trapped in the ideology of
laissez-faire.

But, as K increases, the fitness "landscape" becomes much more complicated
and non-monotonic. It is easier to become trapped in local maxima and miss
the global maxima. Worse than that, Kaufmann proves that as K increases, the
fitness of the highest peak falls to the average fitness; i.e., evolution
becomes a random walk in a random landscape.

How do we get out of this "Red Queen" universe? By limiting the amount of
interaction. Kaufmann shows that K=2 systems can find significant fitness
peaks. So can higher K systems if the interaction functions are "canalyzing".
That is, if the interactions tend to turn things off instead of on, which is
exactly how enzyme systems and brains are structured - more inhibitory
connections than excitory connections.

The briefest thing I've read by Kaufmann is his 1991 Scientific American
article "Antichaos and Adaptation". You can get the flavor for most of
his work from this ten page article. Truly, he is a windy son of a bitch.

1.2 The New Paradigm: Connectionist Democracy

My attempt to state the new paradigm goes like this:

Research in brain science has revolutionized our view of how we think.
The picture is by no means complete, but the idea of rationality has
taken a serious beating. Antonio D'amasio's "Descartes Error" is a
thorough debunking of the "cold rationality - good; hot emotion - bad"
simplification of how we think. He demonstrates that emotions pre-dispose
our rational minds to make judgements. In other words, emotional feelings
represent the "value system" used by our rational minds to make decisions.
Furthermore, emotions are laid down in the first five years of life. That
is why the old saying "As the twig is bent, so grows the tree." is more
than a cliche.

Neurophysiologists are busy tracing the connections between the limbic
system (seat of emotions) and the cortex, especially the pre-frontal
cortex (the seat of reason and planning).

In the last 30 years, we have learned that what we would call thinking,
or intelligent behavior, or adaptive behavior in complex living systems
depends on calculational systems (or algorithms) based on N-dimensional
representations. Neural net explanations of the Vestibulo-ocular reflex
and the cerebellum are based on training an N-dimensional content-
addressable memory. Kanerva (Sparse Distributed Memory) has derived
properties of N-dimensional space that explain how a distributed system
can remember, learn, and behave intelligently. Some audio and video
compression systems are built on "vector quantizers", which are yet another
flavor of N-dimensional learning system.

In short, anywhere that technologists are trying to create artificial
learning behavior, we find N-dimensional representational systems. (
The GOFAI (good old-fashioned AI) crowd has met its Waterloo; nobody
takes rule based _low-level_ vision systems seriously anymore. Rules
are a high-level phenomenon.) The geometric properties of N-dimensional
space are not intuitive - intuition would lead us to expect nothing
interesting, but it is wrong. There seems to be an emergent property
of learning in N-dimensions.

The conventional definition of rationality, on the other hand, is one-
dimensional: we analyze the situation (no explanation of how we know
what is important and what to ignore - the "framing problem" which
defeated GOFAI); we give a weight (in the K=1 sense) to each possible
outcome (no explanation of how we arrive at the weights), and we decide,
i.e., turn the crank (sum the N independent weights) on the rational
machinery. The explanation ducks the hard part: in a unique or unprece-
dented situation, how do we decide what to attend to and how much weight
to give to various events? If there is not enough time or not enough
information available to evaluate the situation, what shortcuts
do we take?

The conventional explanation is full of holes at the level of the
individual, but it is even more obviously ludicrous at the level
of the society. The current form of representative government
was proposed about the time of Descartes, in a Newtonian, determi-
nistic worldview. At that time, only the tiniest fraction of the
male, land-owning or propertied population was allowed to vote,
and the role of government was very limited compared to today.

Here is the conventional rationality explanation of how society is
governed: government is in the hands of the people. Our elected
representatives merely reflect our desires. The citizens analyze
the situation, decide what is important, and tell their elected
represenatives how to vote. In an era of 500-page Congressional
omnibus bills and 10,000 page tax codes with 40,000 lobbyists
dispensing hundreds of millions of dollars of PAC money, it is
a complete fantasy.

Here is what is really going on. The mass media decide what
issues will be given coverage. The mass media, by the amount of
time or column inches, decide what weight will be given to each
issue. (Of course, the concentrated mass media are owned by
the richest segment of our society and reflect their opinions
faithfully, and often manipulatively and deceptively. Once every
couple of years, the voter gets to pick a representative on the
basis of summing (i.e., turning the rationalist crank) all
the media-determined issues that have occurred during the intervening
years. Throughout, the voter is bombarded by a bunch of partisan-
sponsored polls and focus group-based advertisements which are
crafted to appeal to the emotions. Sound-bite and factoid media
coverage further obscure the true issues and prevent anything
resembling true rational analysis.

The conventional explanation posits an everything-to-everything
connected neural net. Each citizen is supposedly capable of
informing himself on all important issues. But, everyone knows
this is nonsense, so we say that politicians "represent" us.
However, they cannot represent us if we have no opinion. In
reality, they are a "homunculus" - the little man inside our
brain that does the thinking and tells us what the answer
is. Once representative democracy has been reduced to homuncular
behavior by the drastic increase in complexity required to manage
a mixed, global economy, it is only a matter of time before
that homunculus will be captured by the most powerful
elements of the society. This is exactly what has happened.

The situation is not directly a matter of corruption, although
there is certainly an immense amount of that present. The capture
of homuncular representatives arises from an inadequate organization
of governmental information processing. Since the structure of
checks and balances, committees, party discipline, etc was codified
over two hundred years ago, it is not surprising that it is
inadequate to the 20th century situation.

Nevertheless, we must get political advertising back under voter
control through campaign finance reform. Otherwise, the constant
brainwash of consumerism will determine peoples' values insidiously
in their early childhood - witness Joe Camel, Mighty Morphin Power
Rangers, sex stereotyping, etc. Then they will acquiese to the
Social Darwinist crowd without a whimper because it is mother's
milk to them.

1.3 Connectionist politics

Somewhere I read that people rarely vote against evil, but they
can be made to vote against stupidity. Liberals have failed with
an appeal to justice. Progressives can succeed by pointing out
the utterly unscientific structure of politics and supplying
a workable alternative. Nobody wants to be perceived as championing
a discredited policy. This is why the "Republican revolution"
has tried to paint a return to the plantation economy as a step
into the future.

Here is an attempt to map current politics onto the unscientific
and unworkable label, and an alternative. Be warned, it may
infringe on some things you think are sacred:

1.3.1 How the current situation works for the moneyed interests

Corporate interests have already organized themselves along NK-
model lines. Each industry pays attention only to those matters
which directly bear on it; i.e., they keep K small. They further
canalize the situation by vetoing (i.e., inhibiting) anything
not in their interest. Each industry has its captive Congress-
critters, so that Congress-critters become the locus of the
algorithmic calculation of NK-fitness.

Senator Porkbarrel does whatever balancing act keeps enough of
his PAC funders happy to fund him and keeps enough of his
constituents mollified to assure re-election.

This is a very non-rational, situation-dependent process. It
works just fine for the corporations, because it is their
fitness function that is being calculated. It doesn't work
at all for the citizen because their issues never get on the
table. Voters are constantly being asked to choose between
being hung or being shot when they really want to vote for
more cops on the street.

1.3.2 Why the current situation works against the common man

This decline of democracy has come about due to technology.
It is a historical truism that democracy rises and falls
with the need for mass armies. If the military technology
becomes dominated by small elites of weapons, like armored
knights, then order can be maintained by a moneyed elite.
If, however, mass armies are the dominant military factor,
then democracy is needed to pursuade the masses to fight.
As the Gulf War demonstrated, the military has moved to an
elite of jet pilots, special forces, and expensive naval
weapons platforms. Masses of first world ground soldiers
cannot be used for fear of politcally disastrous casualties.
Third world countries can afford to trade their ground
soldiers three or five to one for first world soldiers; so
the first world takes the high-tech approach, ala the British
vs the Zulus.

Similarly, it used to be that mobilizing voters was an
important part of politics. Labor supplied "armies" of campaign
workers. With the rise of TV advertising, we have moved to a
small elite of pollsters, pundits, and advertising agencies.

Further, the increasing complexity of interest groups in the
society makes it hard to create broad consensus on any policy.
There are very few genuine problems that all can agree on,
but there are immense numbers of manufactured, phony,
emotionally provocative issues that we can be divided by.

1.3.3 Reorganizing into an NK electorate

The beginnings of an NK electorate are already in place:
the genuine citizen special interest groups: Planned
Parenthood, Zero Population Growth in the area of
population; Sierra Club, Wilderness Society, Greenpeace,
etc. in the area of environment; etc. The problem is,
that citizens still vote by the conventional rationality.

Corporations maximize their effectiveness by concentrating
their vote. The electorate must specialize in order to
obtain similar effectiveness.

1.3.3.1 Voter specialization

We need to define a couple of hundred governmental
specializations. These specializations are continually
adapted on a multi-year basis, sort of like the census.
(A good place to start is the Congressional sub-committee
specializations.) Voters get to sign up to vote in say
five or ten specializations only. Then, we need to
reorganize the elections so that Congressmen specialize
in some of the specializations. (Which is effectively
what Congressional staffers do today; only we don't get
to vote for staffers.) Specialized citizens vote for
specialized Congressmen, just like the corporations
do today. Citizens could change their specializations
whenever their is an election.


There are some mechanics necessary to prevent everyone
from specializing in the currently hot topic. Overall,
it makes sense if a trucker specializes in transpor-
tation legislation, and a telecommunications engineer
specializes in telecommunications legislation. I think
that the hot topics problem may work itself out as people
realize they have more relative influence in less-popular
groups. Some may opt for enfranchisement above relevance,
just as some scientists work on obscure but safe topics
rather than get involved in the high-risk, high-reward
hot topics like cancer research.

Needless to say, the representative assembly of these
specializations would bevoted for on the national level.
This kind of voting, with a 5% or 10% cutoff to get
a representative, is quite common in all advanced country
democracies except the US. It is called proportional
representation. It works something like the brain, which
suppresses low-amplitude signals, but allows medium amplitude
signals to co-exist and cross-stimulate.

"Wait", you may say. "I am giving up my right to vote on
all these other issues! I am being disenfranchised." But we
are already disenfranchised, so what are we losing?
Besides, many voters today are already single-issue voters.
Very few track as many as five issues seriously. Why not
formalize the information-overload situation that is the
de facto reality? Why not acknowledge that everything-to-
everything connectivity doesn't work? Information overload
wasn't even a concept four hundred years ago.

A good feature of national, proportional representation is
that it is much harder to intimidate a distributed, nationwide
constituency either by moving factories out of the
state or cutting off government pork.

What we need to make this work is that these various
specializations can be connected together into a coherent
governmental information processing algorithm. That is,
we need a new theory of government based on 20th century
knowledge of self-organizing learning systems, just as the
17th century system of checks and balances was based on
the best mechanical knowledge of that day.

The recent discovery that specialized brain regions (like
the ones that track motion, the ones that determine color,
and the ones that recognize objects) are pulled together
on a millisecond by millisecond basis via a 20-40Hz broadcast
network offers some ideas about how to organize the government.
Each specialization gets its own internet bulletin board.
Legislation must be displayed for all to see for a certain amount
of time prior to Congressional debate and voting. This period
allows specialized citizens to organize and have their say
via the same network, perhaps even actually voting
.

But notice, this is not the simple-minded electronic
townhall that has been proposed. It is not a mindless,
unfocussed feels-good-to-get-that-off-my-chest gripe
session. The participants are all specialists. There
is a formalism or algorithm. Information is displayed
and acted upon in some kind of N-dimensional set of
bulletin boards, where it may also be cross-posted
to K related bulletin boards.

1.3.3.1.1 Issues of general interest

Here, I can't quite see how to proceed. If a K-related
legislature decides something cross-posted is important,
how does it act? Is a supra-majority needed to add it
to the legislative calender? (Otherwise, if there could
be such a thing as a majority party with 200 almost
independent legislatures, the whole NK model could be
collapsed back to today's system by voting for the adding
of all crossposts to all calenders.)

Suppose that an issue is genuinely relevant to many
constituencies, how do they all get to vote? Does the
K-related crossposting ripple outward, or does it make more
sense to have a special mechanism for identifying general
issues? It sounds like I'm backing myself into a position
where there is some "limbic system" legislature that
decides which of the issues of the 200 simultaneous
legislatures gets into "the general public's consciousness"
as an issue of general importance.

The minute we have a limbic legislature, we are right
back to today's committee system. But isn't that system
much like the brain? The limbic system uses the cortex
as its committees. I guess the issue is that today's
committee system is too hierarchical - the committees
are bottlenecked by the necessity of having all
committee work re-validated by the assembly at large.
The choke point gives control to the gatekeepers who
sell out to the moneyed interests.

1.3.3.2 Genuine decentralization

Notice that I said the specialized representatives are
voted for nationally. This makes perfect sense. The
federal government is concerned with abstract matters
of regulation. The issues of how to arrange the details
of your locality should be handled in local legislatures,
just as with the states today.

The next change that needs to happen is decentralization
of as much as possible; but, destoying federal oversight
without destroying multinational corporate power will
result in local corporate feudalism. So, can we decentralize
without disarming? If the specialized federal legislature
can keep the corporations and mass media from suppressing
local citizen activists, then local government will cease
to be dominated by absentee landlords.

Today, nationwide mass media destroy the intermediate
layers of calculation necessary for citizens not to be cut
out of the local governmental process. Local news has
been reduced to "if it bleeds, it leads". We must restore
a genuine participation in local governments, which have
long been run by construction companies, real estate agents,
the local gentry, and the big employers in the area.

The first thing that has to go are the current state boundaries.
As Michael Lind points out in "The Next American Nation",
currently, 16% of the population (in sparsely populated rural
states like Montana or New Hampshire) elects 50% of the Senators.
This gives the conservative elements in those rural states an
effective veto over the other 84% of the citizenry. Other
countries, even England, redraw state boundaries based on
census data. In England, they are called New Counties.

What should the new boundaries look like? Jane Jacobs has
addressed this in her eye-opening "Cities and the Wealth
of Nations". Essentially, the city is the basic unit of
the economy, not the country. The country was the basic
unit of international politics for the past four hundred
years, but a single national currency lumps together
prosperous cities like San Jose, CA with dead cities like
Camden, NJ. The powerful economic feedback of currency
value adjustment is denied within countries. If corporations
are free to shift their assets all over the world to seek
better financial conditions, the same technology could be
used by new states to use local revaluation to spur
economic health. It could be tracked by the same means the
corporations are currently using.

I think that state boundaries should be some kind of Voronoi
tesselation of the terrain, with the 50 largest cities as the
nodes of the tesselation. This was proposed in the early 80s
by someone from U. Michigan. By preventing suburbanites from
fleeing to bedroom-community neighboring states, states
centered geographically on their city will promote regional
solidarity and more effective economic integration. If the
borders are biased to land on mountain ranges and other
impassable and economically useless terrain, then inter-state
rivalries and frictions going back hundreds of years could
be mitigated. The current state boundaries were determined
by farming community settlement patterns and ancient political
events like the Missouri compromise and the Civil War. Isn't
about time they were rationalized to account for the urban
areas that drive today's economies and policies?

Sensibly-bordered, economically-integrated states will
increase localization of production and consumption and
citizen participation. Both these factors are needed
to counter the deliberate destruction of local competitors
by global megacorps. These corporations are applying the
Stalinist development solution of binding disparate provinces
into unwilling economic unity by destroying their self-
sufficiency and forcing them into an oppressive empire
dominated by a planning elite.

1.3.4 What I want to do

I need to do two computer simulations to test fly these ideas.
Both of them depend on serious databases. I need to create
some kind of N-dimensional representation of political issues.
This is akin to the net-browser/filter problem that people
are just realizing is necessary. (How do I filter all the
crap on the net down to the stuff that is really important
to me? Can I define a clear category for that filter? Could
a vector-quantizer assistant learn to sort on the basis of
my selections?)

Second, I need to tesselate the US, based on population
density, natural features like rivers, lakes, mountains,
and economic features, like industrial, agricultural,
clerical, etc. This is going to take more serious analysis
of what is important, and may again require a self-learning
algorithm.

Also, I need to play with some numbers about the appropriate
size of political units. Sociologists have determined that
about 100-150 people are about all that a normal person can
deal with as a primary community. This leads me to try to
organize two or three levels of local government, with each
containing 256 people - a nice computer number. Three layers
(ward - 256; town - 64k, and city - up to 16 million) of local
government are easy for most people to relate to and historically
this is what has been done. Nothing really new here except the
binary numbers.

Actually, what is new is that the size of the unit is fixed;
and, as the population grows, new units are created. What
currently happens is that the number of units are fixed,
so that their size becomes an increasing impediment to
democracy as the population grows.

But, I am modeling on the brain. We seem to build about three
layers of hierarchical processing before we start using the
20-40 Hz broadcast system. That is, we have some hierarchy
in order to calculate something more interesting than what a
single neuron can do; but we tend not to go too hierarchical -
"grandmother cells" don't work, but the broadcast system does.

I'm still not sure if the local government algorithm is organized
differently than today. I mean, its supposed to be dealing with
local infrastructure issues, such as should we build another
road, float bonds to expand the port, open a new community
college, etc. Since each citizen's money is being spent, they
all seem to need a non-specialized vote in local government.
I think that the three-layer structure and the immediacy of
local issues can work within the existing policial framework
as long as the federal government keeps the big money from skewing
things.

(Idea of multi-threaded school buildings: sharing expensive facilities,
but broken into smaller units whose students don't really interact.
Smaller units are more socially cohesive. Why wreck cohesiveness
in the name of economic efficiency when you can have both?)

Will local governments interact only with their geographic
neighbors (low K) or will they form into geographically-based
groups (all on the Mississippi) or industrial groups (all
heavy manufacturers, all farmers)? (The latter case falls into
the national-specialist paradigm.) I hope that good state boundaries
will create mixed and somewhat self-sufficient groupings of farmers
and city workers. But without database work, I can't predict if
the cities won't go to war with each other instead of minding
their own business.

Regarding the federal, proportional representation scheme, how
many voters vote for each representative? Do I use three levels
in this scheme as well? Should the number of Congress-critters
stay at the current 500, or should we be honest and recognize
that the 20,000 or so staffers are actively make government policy?

If 200 million voters cast 5 votes each, there are 1 billion
votes for 20,000 staffers. Each staffer needs, on average,
50,000 votes for election. That's about right for two
layers of voter hierarchy. Only this hierarchy would have
to be mediated via the internet due to geography. Also,
20,000 staffers divided by 200 specialties results in
assemblies with 100 members - which is also about the right
size for effective debate and legislation with alliances
forming and reforming to represent the actual situation.
One of the problems with the H. of Reps is that 435 is just
too big.

Of course, these numbers should be adjusted for the size
of the country. If this were tried in a country of 10 million
people, you might only be able to support 2,000 staffers.
Thus, the number of legislatures would have to go down while
the jurisdiction of each legislature went up. I'm not sure
if this idea scales well. (On the other hand, small countries
don't have the resources to do everything. They need to
specialize. This system enforces specialization.)

Also, should voters be asked to choose whether to cast
federal or local votes? Maybe this should be a further
area of specialization.