| Santa Fe Institute - Annual Business Network Meeting
October 16-17, 1998
Notes Courtesy of Paul Plsek
John Holland, Professor of Psychology, Engineering, and Computer Science University of Michigan
Program Note: Professor Holland talks about
a theory of "emergence" which can predict many complex behaviors,
and can teach us much about life, the mind, and organizations. He
did not actually talk about the topic above; instead a talk about economics
and modeling.
"Inter temporal trade-off"
— Getting the long horizon into the present valuation of a firm.
A quick way to get more $ to the bottom
line is to shut off new investment; of course, this creates a future
difficulty.
How do I get the value of investment reflected into my current valuation?
Standard accounting procedure tries to do this, but not well.
Transition: Cell development in biology
Every cell in your body is constantly competing for resources.
Induction: the ability of one cell to force another cell to switch into another path that
it would not otherwise choose.
Competence: the ability of cells to resist induction by another.
So, less competent cells reproduce less. Further, competence goes down over time. This is
why your liver and other organs stay about the same size.
Cells and molecules are changing over all the time; you are a pattern that is preserved
rather than a stable mass.
Cancer is when a cell becomes over-competent and reproduces at very high rate.
It takes about 40 generations of cell division to go from a single cell to a mature human
body of about 1012.
Some cells stop dividing (by induction) at about the 12th generation; these are
the germs; later these germs become central figures in now mature body.
How do germs do this "long range planning?" (Answer came later: They don’t
do long range planning they simple travel up gradients; take the best next step.)
Complex adaptive systems
Lots of agents interacting, learning and evolving all the time.
CAS is non-linear. Cannot study the parts and then infer the behavior of the whole.
Non-linear systems are not covered by traditional mathematics. In a CAS innovation is a
regular feature, equilibrium is rare and temporary. CAS have leverage points; small action
produces big effect.
Immune system is example of CAS. Constantly adapting and compensating for new invaders.
Has a memory. A vaccine is a non-linear leverage point; a cheap fast shot makes the system
immune to the disease for the future
In a CAS, anticipation changes the course of the system, even when what is anticipated
doesn’t actually happen. KEY: Anticipation is, therefore, very important.
Models (and metaphors) are therefore key because they are the way by which we anticipate.
Can we build organizations that better anticipate?
Back to Economics: Discussion of Options
Example: You have the opportunity to build a plant. You have the option,
with risk, to build the plant or you can build a pilot for less cost
and less risk. By building the pilot you have taken an option just like
a financial option. You have paid now to have flexibility in the future.
Ref: Tirgeorges, Real Options, MIT Press, 1996(?)
Options are a non-linear leverage point. If the stock goes down a dollar, option goes down
a dollar; but you never pay as much for an option as for the full asset.
An idea: Suppose a company sets up a lottery on its future product. You pay for a lottery
ticket as a bet on the future success of that product. If product hits, you’re rich;
if fails, you lose the value of the option. This would be a way of getting future value
into your current cash flow.
Application to Organizations
odels are key. Organizations have lots of models. Every spreadsheet
is a model. Unfortunately, these are linear models, only valid over
the short term.
Another type of model is a flight simulator. This can take in complexity.
How do you validate a flight simulator? Can’t do it by examining the code; too
complex. The only way is to let an experienced pilot try it out. Same is true for models
in corporations; let experienced managers try out the simulation.
If model is valid for usual conditions, next step is to push the envelope, try outrageous
things and see if it plays out per you intuition.
Having a good organizational flight simulator is an option strategy. You invest a small
amount into it with the hope that it will help you pick good strategies and avoid bad
strategies in the future.
This is an "exploratory model." Metaphor is also an example. It helps you
explore something.
David Stark, Professor of Sociology Columbia
University
Program Note: An examination of institutional learning and evolutionary
economics through a discussion of "heterarchies" — an
emergent organizational form with distinctive network properties, asset
ambiguity, minimal hierarchy, and multiple organizing principles.
Topic: Organizational innovation under extreme conditions of uncertainty; example is
former Soviet Union.
Outline: (1) Introduce concepts, (2) Indicate findings on enterprise restructuring in
former Soviet Union, and (3) Translate that into our kinds of organizations.
How do we foster innovation in our organizations?
Lessons learned from study of former Soviet economies
- We need to dispense with the notions of clean slate recipes
for how social systems should work. Societies are not a single entity, they are
collections of systems. So, the goal cannot be to toggle from one system to another, but a
rearrangement of systems.
- Too quick lock-in to "what works" might actually
impede an organization’s ability to adapt. In Soviet, westerners tended to say,
"do what we have done and do it quickly." But this could impede long term
ability to adapt. We lose the skills of search because we think we know the
"best" ways. Exploitation versus exploration; need a balance.
- Institutional friction sustains diversity by preserving
forms that might be useful in later recombinations. Resistance to change may actually be
helpful dissonance. Dissonance helps learning. Perfectly functioning may be bad.
- Adaptability is a function of diversity. More diversity
increases the chance that appropriate solutions will already be know when conditions
change. Adaptability is enhanced when different organizational principles exist in active
rivalry. This rivalry provides more opportunity for recombinations.
Eastern Europe key buzzword: Privatization
Study of firms in Hungary show that they are embedded in dense networks with other firms.
Lots of connection and interconnection. This helps in risk spreading and, therefore, risk
taking. The firms that had most connections were more likely to introduce new products.
Three "blurrings": blurring of government versus private ownership, blurring of
organizational boundaries, blurring of legitimzing principles.
Examples of blurring of legitimizing principles: a shop that sells swimwear and
umbrellas; a road side stand selling rabbits with the sign "pets and meat."
Translation to Our Organizations
New organizational form: Heterarchy (rather than hierarchy)
Four characteristics:
- Network properties. Unit of economic action is not the firm
but a network of firms. We need to know more about networks. What properties work well in
various situations?
- Marked by "asset ambiguity." Don’t worry
about who owns what. Lots of relations of interdependence. Open boundaries to other firms
to get deep proprietary information. Projects are launched before all property rights and
design issues are fully settled. Don’t wait for the lawyers; you’ll miss the
first mover advantage. Also ambiguity about what constitutes an asset. We don’t know
what may be an asset in the next moment. Waste may be later an asset.
- Have minimal hierarchy. Don’t displace hierarchy all
together, but have virtual hierarchy. Note that with all the ambiguity in the point above,
the planning horizon is not far out. Strategic planning by senior people makes little
sense. More inclusive. Every department is searching for new products and markets.
Can’t be managed centrally. Focus on coordination and information flow. Authority
emerges laterally. Disconcerting to old managers, "I don’t know who my boss
is." May still report to boss, but now accountable to other departments laterally in
the organization.
- Has multiple organizing principles. Pragmatic reflexivity.
Key question: What sorts of processes can help us construct organizations that learn?
Erica Jen, VP for Academic Affairs SFI
Future research directions for SFI…
- Evolutionary dynamics. Biological,
and social systems via metaphor. How do systems evolve over time?
How do algorithms evolve? Exploration of stepwise evolution versus
gradual evolution.
- Interplay between form and behavior.
We have very few tools or models to explain network models in organizations.
What works? Will start with chemical self-organizing network. Then
work on biodiversity. Then, social insect behavior. Then, human social,
political, and economic systems.
Note to myself: talk to Curt about how healthcare might be involved
in this research.
- Resilience in social and natural systems.
Start with biology then move on to social systems. Will look at redundancy
and robustness.
- Complex network dynamics. Messy dynamics,
information processing, etc.
Doug Engelbart, Inventor of the computer mouse and now
President of the Bootstrap Institute
Program Note: Dr. Engelbart will discuss Improvement
Communities — part of society’s diffuse, organic Improvement
Infrastructure. The escalating complexity and urgency of humanity’s
problems require a new degree of effectiveness from this Improvement
Infrastructure.
Topic: Boosting Collective IQ
Society’s problems are getting increasingly complex.
We need better tools to deal with this.
The limitations of current paradigms are the biggest barrier and the biggest leverage
for improving collective IQ.
Focus on capability infrastructure (technology) and on whole, human systems (social).
Part of the general capability infrastructure is the improvement infrastructure. Need
rapid improvement cycle.
The goal is to boost collective IQ by concurrently developing, integrating, applying,
and re-using knowledge. How do we do a better job of scanning for, ingesting, and
interacting with knowledge?
As a tool, have developed "Open Hyper-document System" (OHS) to give to
Improvement Communities to aid their work. (He did not explain this nor give an example,
but it seems to be an advance beyond hypertext.)
We are not going to get fundamental improvement just by automating current ways of
doing things.
Categorization of knowledge that we need to capture and manage: (1) Recorded knowledge
(memos, meeting minutes, change reports, etc.); (2) Intelligence collection (papers,
books, news, proceedings, brochures, surveys); (3) Product knowledge (plans, proposals,
reports, specs, source code).
The whole augmentation system: (1) Human systems: paradigms, organizations, procedures,
customs, methods, language, skills, knowledge, learning, and attitudes; (2) Tools system:
facilities, media, tools, machinery, vehicles, etc.
We need new human systems to take advantage of the explosion of technology in tools
systems.
Jointly developed scenarios might be a first step on this boosting. Having people think
about the future under assumptions of dramatically different technology.
Doug has formed a "Bootstrap Alliance" of organizations that want to work
collaboratively over time to become an Improvement Community. This is a network based in
California and one in Japan. Dream is of a constantly evolving, constantly giving birth to
new forms, type of network.
Brian Arthur, Professor of Economics Santa Fe Institute
Program Notes: A dialogue and exchange
of ideas regarding the impact of increasing returns in today’s
competitive environment.
Topic: Increasing returns and the new economy
Increasing returns: success breeds success, you get
further ahead as you get further ahead, positive feedback loops.
We tend to groove into high-tech products; if I get comfortable with Microsoft Word it is
harder to get me to change.
If there are increasing returns, don’t be passive. Try to enhance your increasing
returns. America Online is a good example of actively pursuing this strategy.
People have suggested that the era of increasing returns is over in these days of the
Internet when anyone can start a new business. He doesn’t believe so.
New economy is dominated by: (1) Digitization and (2) Interconnectedness.
Service businesses of old were local deliverers, now they are networked businesses.
Examples: banking, book retailing, insurance.
He believes that in networking businesses there are mild increasing returns tied to
network size/membership.
Arthur’s prediction: Of networks, there will be few.
Metcalf’s Law: Advantage of belonging to a network is proportional to the square of
the number of members on the network.
Arthur doesn’t believe it. He feels that the value has a plateau. If Amazon.com goes
up in members by 10x, what good is it to me?
Kevin Kelly has new book: 10 Rules for the New Economy.
Arthur’s feeling, there is a new economy based on digitization and interconnection.
But it is yes and no about whether there is a new economics. No, there is no new economics
because human behavior is not changed. Yes, there is a new economics in the sense that we
may need new assumptions. It is like relativity (needed at high speeds) and Newtonian
physics (fine under normal conditions).
Question: Aren’t there 2 kinds of networks; hub and
spoke (like Amazon.com) and highly-interconnected (like AOL chat rooms). Yes. Most of his
thinking has been about hub and spoke. But he would still say that there is a point of
plateau. The telephone network is highly interconnected. But if there were suddenly 10
times the number of telephones, it wouldn’t mean much to me because all the people I
want to talk to already have a telephone.
Point: Most examples of increasing returns lock-in a
technology, not a single company. VHS is locked in, but many companies compete to produce
to that standard. What may be new now is that increasing returns are locking in individual
companies, like Microsoft. He is not so worried about this lock-in regarding price. What
is worrisome is whether innovation is harmed by lock-in by a single company. A key
question then is: increasing returns for whom?
Murray Gell-Mann, Distinguished Fellow Santa Fe
Institute and Professor of Theoretical Physics CalTech
Program Notes: "Ways in which scaling
laws can arise." A comparison of different explanations that are
offered for scaling laws in a variety of fields.
Topic: Scaling laws
We often build models with lots of assumptions that make
them easy to build but potentially unrealistic.
What can you claim if you have an oversimplified model? Maybe you can identify some
phenomenon that does appear in the simple model but also plays out in the more complicated
model. This is great and fortunate.
What should we do when phenomena have complex causation and are not fully understood? Ref:
Naven, by Gregory Bateson and Ravens in Winter, by Bernd Heinrich. Vague explanations can
be helpful even if they don’t explain everything. You feel more comfortable with it.
Survey of Power Laws
A scaling law is usually a power law… some quantity is distributed
as power of something.
A power law is the only kind of relationship that is insensitive to the scale; that is,
the power law is smooth at all scales (of course, excluding end-effects at the top and
bottom limits).
These power laws are found in many different fields. They apply to adaptive systems (human
systems) and non-adaptive systems (galaxies).
Example… Zipf’s Law: rank metropolitan areas by populations; population is
proportional to 1 over the rank. That is, the population of the 200th area is
about half that of the 100th area. This is another power law. A plausible
explanation comes from Gibart’s law that states that the distribution of growth rates
is statistically arranged around a common mean. This makes Zipf’s law more
understandable.
He gave several examples of the universality of power laws (including Per Bak’s
self-organized criticality). (Note: Gell-Mann points out that power laws do not actually
apply to sand piles, but close enough.)
Note to myself: Point seems to be that it is not enough to show that something follows
a power law; because lots of things follow power laws. Need a deeper explanation. It is OK
if you don’t have a complete explanation, a vague explanation is still helpful.
Universality classes: Many phenomena have the same power function. We don’t fully
understand this but it is a potentially important observation — if we can explain it.
Per Bak points out that big events make history and small ones are just footnotes; but
they are all on the same power law curve. Bak suggests that, therefore, the underlying
phenomena are the same for big and small events, Gell-Mann says maybe, maybe not. Without
deeper explanation, we don’t know. "It is not true that the question of whether
there are power laws is the same as the question of whether they arise from the same
explanation."
Gell-Mann’s summary: So, there are lots of connections among these power laws. But we
don’t know what that means really. We don’t know if there are some universal,
underlying phenomenon; but there might be and that would be very exciting.
Geoffrey West, Physicist Los Alamos National Lab
Program Note: "Fractals and the tree of
life: A unifying theme for creatures great and small" Although
life is the most complex physical system in the universe, many of its
general physiological features obey remarkably simple scaling laws.
Dr. West’s quantitative, unifying model can be used as a paradigm
for many other complex systems ranging from river systems to corporate
structures.
Theme: Scaling Laws
Things exist on many scales ranging from sub atomic
particles, through microorganisms, through us, through the earth, through the universe.
Many orders of magnitude.
Metabolic rate: How much energy you need to sustain life.
Plot metabolic rate versus mass of creature. Follows a power law; that is, a straight line
on a log plot. Slope is 3/4; that is, metabolic rate varies as the 3/4th power
of creature-mass. Holds true for creatures ranging from mouse to an elephant. Further
study showed that it extends 27 orders of magnitude down to cell components. Further study
showed that it was true also for plants and ecosystems!
Meaning of this power law: A gram of mouse requires 9 times the energy as compared to a
gram of elephant. Small may be beautiful, but big is more efficient. Does this also apply
to organizations?
Finding: The number of heartbeats in a lifetime is relatively invariant across all
mammals (~1.5 billion). Guess what, the total number of engine revolutions over the
lifetime of a car is roughly 1.5 billion!
West feels there must be some universal principle here.
What might underlie these universal principles?
Postulate three general principles:
- All life is sustained by hierarchical, space filling (that
is, has to go everywhere) branching systems (like the cardiovascular system).
- The terminal branch of the system is the same for all (for
example: cells are the same size and so the capillary system has to be the same size for
all mammals).
- The networks that evolve are the ones that minimize the use
of energy.
A key idea here is hierarchy and branching. It
is fractal in nature.
|
