![]() |
What is the Santa Fe Institute?
In the spring of 1997 journalist Kenneth
Baake spent time at the Institute talking with researchers about the "rhetorical
challenges" inherent in trying to explain the nature and work of the
Institute. In the following description, he summarizes his impressions
of SFI.
The Santa Fe Institute (SFI)
draws scientists from universities and research institutions throughout
the world to pursue broad research problems. Much of the work focuses on
the science of complexity, which examines underlying patterns and regularities
behind a wide assortment of real-world phenomena. Researchers collaborate
at the Institute on projects ranging from the communication patterns of
ants to the way information spreads across economic markets. The aim of
the Institute's exploration of these phenomena is to help define new research
directions within the scientific community at large and to shed light on
problems that challenge our global society. |
Founders of the SFI in 1984 set it up as an independent, nonprofit research
center committed to the idea that the most exciting problems in science
today require insights from many disciplines. SFI discourages the traditional
academic barriers that often keep scientists of different backgrounds from
working together. Here you can find physicists, biologists, psychologists,
mathematicians, economists, immunologists and others nurturing various
ideas and techniques. Visitors ranging from undergraduates to senior scientists
share ideas at the Institute's campus in northern New Mexico. Long-term
research and discussion continue by means of electronic mail, off- site
collaborations, fax, telephone and return visits.
Over the course of a year, the Institute hosts some 150 scientists (about
40% of whom are first-time visitors), for varying stays with about 35 in
residence at any time. There are a few
faculty members with multi-year appointments and several postdoctoral
fellows and
graduate interns, with the balance comprised of scientific visitors
predominately from universities in the United States and Europe. Lunch
time on the patio is multilingual; along with English one might hear French,
German, Japanese, and Norwegian. Researchers like political scientist John
Padgett from the University of Chicago and biochemist Walter Fontana from
the University of Vienna may be found discussing the emergence of social
and political organization in Renaissance Italy. Out in an alcove in the
hall and impromptu gathering and scientists and students huddle over a
computer debating whether one species depends on another to compel evolutionary
changes.
It is difficult to sort the work into traditional academic categories in
an environment as fluid as the SFI. Because some of the questions being
asked here are new and cross many academic disciplines, scientists often
find it challenging even to define the concepts they are studying. But
it is safe to say that the SFI studies tend to follow living and non-living
agents and groups of agents as they emerge, as they organize themselves
into complex communities and networks, and as they adapt, evolve and learn.
The processes of emerging, organizing and evolving often are inseparable;
in a way the three are merely different filters through which we view the
dynamics of complex systems.
As an example of current work, a group of researchers at the Institute
is asking how the first replicating biological life forms emerged on earth.
Experimentalists and theorists at SFI and elsewhere--including Harold Morowitz,
Stuart Kauffman, Reza Ghaderi, Peter Wills, Philip Anderson--are embarking
on a new, joint approach to some of the questions that have plagued the
study of the origin of life for decades. What were the first replicating
biological molecules on earth? What are the thermodynamic conditions that
have to be satisfied for systems to become progressively more ordered and
specialized? What sort of chemical reaction network is needed to produce
anything as complex as cellular biochemistry? Is there some special principle
underlying the coordinated processes that maintain the integrity of organisms,
even the simplest cells? The SFI-based team intends to combine laboratory
and theoretical work in its attempt to find fresh answers.
A second example of SFI research is the study of stock market dynamics.
Traditional economic theory suggests that markets remain in equilibrium
as economic agents display rational behavior in anticipating changes in
supply and demand. Yet the stock market generates a type of volatility
that does not match the rational expectations model. How does this behavior
emerge? W. Brian Arthur and a team of SFI researchers have developed an
artificial model of the stock market. Analysis of the model over the past
year indicates that the stock market converges to one of two attractors--one
corresponding to the equilibria posited in conventional theory and the
other exhibiting the more volatile qualities observed in the real stock
market. The researchers determined that conventional behavior is observed
when individual agents have slow rates of learning about conditions affecting
the market. But when learning is rapid, the artificial stock market behaves
in the more volatile manner.
The stock market model project typifies the multidisciplinary, multigenerational
approach that the Institute strives to encourage. In addition to economist
Arthur, the team comprised Blake LeBaron, economist from the University
of Wisconsin; John Holland, computer scientist from the University of Michigan;
Richard Palmer, physicist from Duke University; and Brandon Weber, an undergraduate
intern from Bard College.
Another example of multidisciplinary work is the collaboration of Los Alamos
National Laboratory physicist Geoffrey West and University of New Mexico
biologists Jim Brown and Brian Enquist. They began research at SFI a couple
of years ago to study universal scaling laws in biology. Scientists have
long recognized that nature preserves proportions--that is, the relationship
of metabolic rate, heart rate and lifespan scale to size--from species
to species. Yet it has not been understood why this occurs as some multiple
of the one- fourth power rather than the one-third power as expected from
geometric scaling. Working together from their different perspectives,
Brown, Enquist and West have presented for the first time a general model
that explains this quarter-power law.
Many of the complex systems under study at SFI and elsewhere share a common
self-organizing architecture made up of a distributed (noncentralized)
collection of autonomous "agents" interacting in the context
of a dynamic environment. Examples range over the self-assembly of viruses,
behavior of social insects, and large-scale ecosystems. What makes these
collective groups scientifically interesting is the coupling between individual
and global behaviors. Although the individuals may be relatively simple,
their collective behavior can self-organize to become quite complex. The
self-organization of the group as a whole emerges in a nonlinear manner
from the behavior of the individuals: this involves a critical feedback
loop between the behavior of the individuals and the behavior of the whole
collection.
The Swarm simulation system, developed by Christopher Langton, probes this
mechanism by which ants, bees, birds and other organisms exhibit collective
behavior, such as in group foraging for food or flying in pattern. Swarm
is a general purpose simulation package for the investigation of such concurrent,
distributed systems. It provides a wide spectrum of generic artificial
worlds populated with generic agents, a large library of design and analysis
tools, and a kernel to drive the simulation. The package is being used
to model systems within ecology, anthropology, chemistry, economics and
political science among others. Swarm 1.0 is now in release.
Researchers James Crutchfield, Melanie Mitchell, Eric van Nimwegen and
colleagues are using genetic algorithms to study how individual entities
that are limited to local interactions are able to produce sophisticated
global outcomes. They are using genetic algorithms to design cellular automata
to perform computations. A cellular automaton consists of a lattice of
identical entities, known as "cells," with each cell existing
in one of a number of possible states. A cell can change states only through
interaction with its close neighbors. As each cell changes it in turn affects
its neighbors, causing the entire system to behave in hard-to-predict ways.
The goals of this research are to understand how computation might take
place in complex systems, and how evolution might design such systems.
Understanding how idealized complex systems such as cellular automata can
compute may someday contribute to our understanding of information processing
in complex systems in nature, for example, how neurons in the brain process
sensory information. In addition, understanding how genetic algorithms
work and how best to use them is an important goal for computer science.
Such models may someday help shed light on aspects of natural evolutionary
processes.
Even as they are emerging and organizing themselves, agents and groups
are constantly evolving and learning--which is another focus of SFI research.
For example, in work that has a strong theoretical and computational as
well as empirical component, Andreas Wagner (SFI postdoctoral fellow) and
others are looking at the evolution of redundant gene functions. Biological
systems provide abundant evidence of this type of redundancy. For example,
study of the gene mutation rate of one chromosome on one particular yeast
cell (Saccharomyces cerevisiae) suggests that a loss-of-function mutation
rate of 40% has little or no effect on its function. Gene duplication is
clearly a prominent mechanism generating redundant genes. Because redundancy
provides protection against harmful mutations, natural selection is likely
to be involved in generating and maintaining partial redundancy. Current
research at SFI focuses on the large body of experimental data on redundant
genes, and proposes conceptually simple mathematical models for the evolution
of redundancy.
In recent years a new view of learning and learning-based systems has emerged:
namely, that the understanding (or design) of systems capable of complex,
robust, open-ended learning and cognition requires a framework in which
intelligence is shared among multiple, possibly heterogeneous, agents interacting
with each other and often with their environment. This approach shifts
the emphasis from static structures and discrete operations to continuous
change, puts cognition in the same dynamical domain as the brain, body
and environment, and makes contact with the principles of self- organization.
The critical question for distributed systems is to understand the relation
between local mechanisms and the learning process of the whole. How does
a network of agents learn to behave differently than do the agents individually?
How do we detect and analyze the emergence of cooperative learning? SFI
researchers and off-site collaborators are mapping out a new initiative
to address these issues. The work will coordinate a range of theoretical,
computational, and experimental research that focus on distributed dynamical
systems.
Human populations of course represent some of the most challenging of distributed
systems. Several years ago SFI turned its attention to the evolution of
human culture through several different projects headed by Marcus Feldman,
George Gumerman, Joshua Epstein and others. These initiatives will continue
to look at topics as diverse as the dynamics of community aggregation and
combat, the evolution of the social contract, and the role played by cultural
transmission, that is, the spread of human cultural values in affecting
human behavior.
Finally, a topic of growing interest at SFI relates to the dynamics of
extinction. Some scientists have suggested that the interaction of different
species can lead to a co-dependency, such that if one dies out it causes
an avalanche of extinctions among the others--a sort of domino effect.
Recent studies suggest that the distribution of the sizes of mass extinction
events in the earth's fossil record may follow a power law, which is a
mathematical distribution frequently indicative of critical behavior.
In the past year, however, researchers Mark Newman and Kim Sneppen have
developed a new model for mass extinction that examines the effect of differing
environmental stresses on species. The most well-known example of an environmental
stress is found in the theory of a large space rock colliding with the
earth some 60 million years ago, causing the extinction of dinosaurs. Newman
and Sneppen have developed a computer model to generate predictions about
species extinction. The predictions are compared with fossil records. This
model matches the power-law distribution of the data found in the records,
suggesting that mass extinctions may be caused by forces external to the
ecosystem.
Just as agents emerge, organize and evolve, so do research programs at
the SFI. Collaborations grow, mutate, join other groups or eventually die
out as interest among the scientific community changes. Yet some enduring
threads tie the organization together. The first is excellence: SFI applies
rigorous standards of excellence to its programs and will not undertake
new research unless it can attract outstanding, creative and dedicated
people. Researchers must be able to contribute not at the margins, but
in setting new directions for science. Such high-risk research can take
years to complete, or can even fail. Another thing that unites members
of the SFI community is their conviction that the most fragile and valuable
aspect of the Institute is its commitment to supporting fresh catalytic
research not likely to occur elsewhere. This conviction demands that our
researchers and administrators continually challenge ourselves. Are we
defining and evaluating scientific excellence in the best ways? Are we
engendering continuous broadly based discussions of fundamental intellectual
themes? Are we encouraging the processes of renewal, follow-up, and evolution
at SFI?
As a recipient of funding
from a variety of federal and private sources, the Institute is subject
to regular site visits and evaluations. That SFI continues to successfully
extend its funding from these sources within a highly competitive environment
is testament to the fact that--at least for the time being--it is asking
the right questions. | More about SFI's Research Focus Areas | 【工事中です。】 | ブライアン・アーサー | 9/26/97 webmaster |