🚀 go-pugleaf

---------------------------------------------------------------------
To: steve@next-s.LANL.GOV
Subject: Re: Memes, Genes and Neural Networks
Status: RO

> 
> Harry -
> 
> Can you fill me in on more context.  I get the drift of your line
> of research but I don't understand if you are striving for accurate
> theoretical modeling, predictive capability or even manipulative
> capability.

I'm striving for an understanding of why urban systems are fundamentally
unstable. I've also done a lot of modelling work in social systems, and
I'm trying to understand the mechanism that lies behind the non-linear
structures I've identified.
> 
> You seem to discount Dawkins' memes.  I understand that they may
> not be a particularly effective way to model sweeping cultural
> change but I think they are extremely relevant in modeling very
> small, critical changes as well as, perhaps, resistance to certain
> changes.  The real value, of course, is in modeling specific
> self-propogating, self replicating, mutating, evolving "ideas, 
> ideological elements, or idea complexes" - to wit, memes,
> in a culture, a community, a socio-politico-economic unit.

The problem with a meme is that it evolves by Lamarckian rather than
Darwinian evolution. The two have distinctly different dynamics. In
Darwinian evolution, variation is -->random<--, followed by selection. In
a Lamarckian system, variation is not random. This leads to distinctly
different dynamics. (I've demonstrated this in simulation experiments.) In
particular, the optimal rate of evolution in a Darwinian system is very
low if not zero. In a Lamarckian system, chaos is not unexpected, leading
to a need for a non-zero rate of evolution. The evolutionary theory that
has been developed for Darwinian evolution needs to be recast for
Lamarckian evolution. In particular, we need to understand innovation in a
cultural context. I -->think<-- an understanding of cognitive networks is
needed here.

> 
> Your ideas regarding cognitive models of social dynamics are 
> appropriate... the collective behaviour of a socio-politico-economic
> unit does seem to be analogous to the behaviour of an organism
> and the perceptual, recognition, strategic and response behaviours
> of organisms can be used in analogy to predict the collective behaviour
> of such systems.
> 
> I don't know that this invalidates the usefulness of darwinian models
> of evolution of ideas - memes, or of holographic mind, or of mental
> models in studying social dynamics.  It seems to be an issue of granularity
> and perspective.
> 
> Your slides lay out on the table a lot of loosely related ideas.  Without
> the content of your talk, I cannot guess how much and where you filled
> in the connections but I can offer the following general observations.
> 
> 1) Cognitive models, as developed to understand human understanding may
> offer useful models for studying social organisms, looking for similar
> processes.  Unfortunately, the human perceptual and cognitive system is
> based on top of a much greater degree of complexity than the most complex
> society.  Social organisms complexity are much more related by complexity
> measures to something between an amoeba the simplest of insect life.  That
> is where this study might be best focused.

Yes. 

> 
> 2) Social dynamics are often driven and constrained by one or a few individuals
> in political, military, religious or economic power of the social unit.  That
> means that much of the behaviour of a social unit is simply an amplified and
> usually distorted version of one or a few individual's responses.  This leads
> to a social unit "appearing" to act according to very familiar principals.
> 
Again, yes. I haven't gotten my arms around this issue, yet. Working out
the separation between those aspects of social group behavior that reflect
individual behavior and those aspects that are emergent is my current
goal.

> 3) If we are to choose to model social units as organisms, we must remember
> that they are incredibly "different" than those we are familiar with.

Totally agreed, but the positive side of this is that it gives us yet
another source of cognitive models.

> 	A) Referring back to point 1)... they are very simple, and built up
> 	of complex units.

It's a whole different type of connectionist model<--not unwelcome,
either.

> 	B) Individual units (people) often directly direct the "organs" and
> 	"limbs" of a social organism. 

An issue that I'm wrestling with.

> 
> 	C) Much of the behaviour, "language" and "values" of the social
> 	organism are inherited and integrated from multiple sources.

Again that issue of how "memes" evolve. I did some of the early work
(~1982-7) on genetic algorithms wrestling with this issue. I didn't expect
to encounter chaos when I did.

> 	The US of A, for example, "inherits" it's values from it's dominant
> 	subcultures, of which it has many.  The integration of the values
> 	is often superior or inferior to it's source, usually directly
> 	recognizeable, and subject to mutation and change on very short
> 	time-scales.
> 	
> As an arguement for memetic modeling... take the case of the
> pro-death/anti-choice battles of the last three decades. 
> The meme complex surrounding the protection of life, the species
> and the individual all got mixed together to develop a
> rhetoric of "pro-life".  The meme complex surrounding individual
> freedoms, the right to choose a personal destiny, the sanctity
> of one's own body, and the like, mixed together to develop a
> rhetoric of "pro-choice".  The two camps, completely addicted to
> the rhetoric of these meme complexes continue to argue entirely
> distinct points as if they are simply the good and evil counterparts
> of the other.  These meme complexes have absorbed much of the nations
> energy and attention and have influenced other meme complexes 
> such as occur in the area of gun-legislation, law enforcement,
> radical feminism, and the medical system.
> 
> The overall dynamics of the US as a social organism, or western
> culture for that matter, are not trivially describable in terms of a few,
> a few hundred or even a few thousand meme complexes but a few of them ARE
> significant enough to be traceable, just as the genes for skin and eye color
> can be directly linked to survival traits in hot, dark jungles and in bright,
> cold forests, plains and shores.
> 
> In closing, I think an "artificial life" model of social dynamics is a
> useful one and can include cognitive behaviours but may very well be
> dominated by more mundane mechanisms exhibited in organisms, and organs
> of lesser complexity than neurological structures.
> 
> 
> Steve Smith
> 
I think we're in agreement. It was the discovery that memes evolve
differently from genes that has started me looking at cognitive networks
in a more general sense.

BTW, would you mind if I forwarded this to Professor Minsky. He has an
interest in these issues.

Cheers,
 Harry Erwin    erwin@trwacs.fp.trw.com

------------------------------------------------------------------------
Date: Thu, 21 Jan 93 14:35:24 GMT-0700
From: steve@next-s.LANL.GOV (Steve Smith)
Message-Id: <9301212135.AA25047@next-s.lanl.gov>
Received: by NeXT Mailer (1.62)
To: erwin@trwacs.fp.trw.com (Harry Erwin)
Subject: Re: Memes, Genes and Neural Networks - Quick turnaround responses!
Cc: steve@next-s.LANL.GOV
Status: RO

Harry,

 Thanks for the fast turnaround...
 


> I'm striving for an understanding of why urban systems are fundamentally
> unstable. I've also done a lot of modelling work in social systems, and
> I'm trying to understand the mechanism that lies behind the non-linear
> structures I've identified.

I think I get that now... The nature of copies of overheads is that
they don't always provide a good projection of the emphasis which is
added by the speaker.  Your handwritten notes DID help, by the way.


> The problem with a meme is that it evolves by Lamarckian rather than
> Darwinian evolution.

Either and both.  I'm probably missing something important here.
Maybe a specific reference on this would help me sort this out better.
--The key characteristic of Darwinian evolution is that variation is
--random. I've seen no evidence that cultural variation is random.
--Harry

> The two have distinctly different dynamics. In Darwinian evolution,
> variation is -->random<--, followed by selection. In
> a Lamarckian system, variation is not random.

>  This leads to distinctly different dynamics. (I've demonstrated this in
>  simulation experiments.)

The results of these simulations would be interesting and have implications
that transcend the issues of the dynamics of social systems.  Are these
published or readily available.
--In press. I haven't seen the journal issue, so I can't provide
--a reference. Alluded to in my paper that was incorporated into
--the proceedings of the Cambridge Conference.
--Harry

> In particular, the optimal rate of evolution in a Darwinian system is very
> low if not zero.

>  In a Lamarckian system, chaos is not unexpected, leading
> to a need for a non-zero rate of evolution.
Same questions... I'm not making the leap here.   The words are compelling
but I can't fill in the details.

--Lamarckian systems can be chaotic, since new variation attempts to
--respond to the demands of the situation. Especially for intra-species
--competition, this may result in a shifting environment. Since 
--evolution is not continuous for a number of reasons, this leads us
--to the theory of non-linear discrete processes, and often to chaos.
--Harry

> The evolutionary theory that
> has been developed for Darwinian evolution needs to be recast for
> Lamarckian evolution. In particular, we need to understand innovation in a
> cultural context.

You are saying that the driver of memetic evolution is the creative efforts
of individuals and institutions within a social system and therefore, the
dynamics implied by this "self-selecting" mechanism is more appropriate?
Sounds promising.

> I -->think<-- an understanding of cognitive networks is needed here.

I may be misunderstanding your use of the term "cognitive network" or
at least it's referent.  Studying the network of information gathering,
flow and processing within a culture as a cognitive network?  Or 
something else?

--Yes.
--I've seen strong evidence that selection at the group level keys
--off the effectiveness of communications, and it's not just
--communications in a hazy, general sense, but the effectiveness of
--the communications networks that the culture has created to respond
--to opportunities and threats. Sometimes it involves a formal
--organization, but more frequently it involves an evolved network,
--and it appears to be closely connected with the industries within
--the culture that we sometimes are able to perceive in the
--archaeological record as culture traits or (invalidly) as separate
--cultures.
--Harry

>> (simpler organisms . . . ) That
>> is where this study might be best focused.

>Yep. 


>> 

>>2) Social dynamics are often driven and constrained by one or a few individuals
>>in political, military, religious or economic power of the social unit.  That
>>means that much of the behaviour of a social unit is simply an amplified and
>>usually distorted version of one or a few individual's responses.  This leads
>>to a social unit "appearing" to act according to very familiar principals.
>> 

> Again, yes. I haven't gotten my arms around this issue, yet. Working out
> the separation between those aspects of social group behavior that reflect
> individual behavior and those aspects that are emergent is my current
> goal.

A social "organism" seems to be a symbiotic mass of host infrastructure and
a multitude of suborganisms (institutions, parties, individuals) that combine
to produce a net behaviour which exhibits both the best and the worst of each
of it's parts.  As in the case of Nazi Germany or any other unit where a particular
symbiotic element dominates the others and virtually becomes a parasitic element,
any co-organism (my term, coined just now) in this mass potentially can take
over the resources of the others and express it's own behaviour through the
collective organism.

--Sounds like a generalized connectionism
--Harry

>> 3) If we are to choose to model social units as organisms, we must remember
>> that they are incredibly "different" than those we are familiar with.

> Totally agreed, but the positive side of this is that it gives us yet
> another source of cognitive models.

Yes, and in fact, I think that it is important to look at (almost) closed
ecological systems for paradigms.  Though they don't seem to provide much
in the way of cognitive models, their symbiotic interrelationships are 

more similar to a social organism, I think, than any single organism might
be.

--Yes, and I suspect there will be strong connections here if we can
--every get a good understanding of a functioning ecology.
--Harry


I clearly need to do more homework... references?

--I'll owe them to you for a day or so. I have some files offline...
--Harry

> I think we're in agreement. It was the discovery that memes evolve
> differently from genes that has started me looking at cognitive networks
> in a more general sense.

As I keep coming back, it is clear that I am missing some important background
and connections here.  More references or detail might help me.

--Gordon Hines's paper on "Learning the ESS" (JTB back in the '80s) is
--a starting point. Other than that, only my two papers in press. (My son
--and I did this work back in the late '80s, but I was too diffident
--to publish until Tom Bell got on my case. It's not publish or perish for
--me, unlike some.)
--Harry

1) Memes evolve differently from genes in the sense that they are selected
for at two levels... a meme is transcribed and mutated dominantly by 

"intelligent agents" which evaluate the meme in a given context and 

"accelerate" it's mutation and therefore, usually, its evolution.  Is this
what you are refering to?  Is there more to it that I am missing?  It
is certainly enough.

--That's a starting point, but the resulting dynamics are dramatically
--different. Gordon Hines had some insight into this when he demonstrated
--that an ESS inherited genetically, but also influenced by learned
--behavior would be perturbed, but my work then showed that the learned
--component could dramatically change the evolution of the ESS, not
--just to a limit cycle, but to true chaos as well. It surprised me.
--I presented preliminary results at the 1987 ESS Conference at
--Sheffield, but then went back to explore things further. I was only
--willing to claim chaos at the 1990 Conference after a lot of restudy.
--Harry

2) I am still not clear on how you are applying cognitive networks here...
it seems very promising but I must not have the context set right.

--No cognitive networks yet, but it appears social evolution occurs in
--that context.
--Harry

> BTW, would you mind if I forwarded this to Professor Minsky. He has an
> interest in these issues.

Certainly, I always assume my e-mail to be like a hallway-conversation,
likely to be overheard and sure to be repeated if it was noteworthy.

Is it possible that more of this conversation should happen in a more
public forum?  I don't remember where I "met" you, I am guessing it
may have been through either the A-life Digest, the Australian 
mail list on GA's or in my occasional meanderings through internet
news groups.  Is one of the above a better forum than private mail
for you?

--Probably the best forum would be sci.cognitive or comp.ai.
--Please choose one.
--Harry

A related "one-liner" that fits right in with this conversation...

"An idea cannot be held responsible for it's popularity"

My own interest in your work is multifold... 


1) Design memes... I am studying the application of memes and of
GA's to the evolution of design.  I am interested in "memes" in this
context because of the "crossover" between "design disciplines" such
as the influence of art-deco furniture, illustration and sculpture
on automobile design or architecture.  A central, basic meme, in
this context might be described as "streamlinedness".  Everything in
that era seemed to be "moving fast".  Much of the aerodynamic affectation
was not, in fact, aerodynamic and this became quite evident in automobile
design where the aerodynamics are often actually put to test.  Seating
and dashboard design borrowed directly from interior and furniture design
as logo design borrowed directly from graphic design, typography and illustration.
Hood Ornaments were designed by sculptures of the day or "borrowed" from
real sculpture.  Other aspects are not translated so directly, they are
translated through an underlying "value system" perhaps.  One that values
"speed" or the illusion thereof, one that values symmetry, et cetera.

Consider the affectation of off-center hood-bumps on turbo-charged sports
sedans today... not all of them have anything protruding into the space
but the off-center bump makes a "statement" about performance.

2) Artificial life... the application of paradigms, starting out at
the "bottom" with self-organizing systems all the way up through
cognitive models, to areas not traditionally thought of as "life-like"
but emerging as ripe for modeling using life-based paradigms.

3) Dynamical systems theory - The wonderful mathematics of dynamical
systems and the implications all over the map of science and technology.


Regards,
 Steve

-- 
Harry Erwin
Internet: erwin@trwacs.fp.trw.com