1 Principles of Complex Systems How to think like nature: Part
II Russ Abbott Does nature really think?
Slide 2
2 Complex systems overview Part 1. Introduction and motivation.
Overview unintended consequences, mechanism, function, and purpose;
levels of abstraction, emergence, introduction to NetLogo.
Emergence, levels of abstraction, and the reductionist blind spot.
Modeling; thought externalization; how engineers and computer
scientists think. Part 2. Evolution and evolutionary computing.
Innovation exploration and exploitation. Platforms distributed
control and systems of systems. Groups how nature builds systems;
the wisdom of crowds. Summary/conclusions remember this if nothing
else. Lots of echoes and repeated themes from one section to
another.
Slide 3
3 Are there autonomous higher level laws of nature? Fodor cites
Greshams law. The fundamental dilemma of science How can that be if
everything can be reduced to the fundamental laws of physics? The
functionalist claim The reductionist position It can all be
explained in terms of levels of abstraction. My answer
Emergence
Slide 4
4 Game of Life Gliders A 2-dimensional cellular automaton. The
Game of Life rules determine everything that happens on the grid. A
dead cell with exactly three live neighbors becomes alive. A live
cell with either two or three live neighbors stays alive. In all
other cases, a cell dies or remains dead. The glider pattern
Nothing really moves. Just cells going on and off.
Slide 5
5 The Game of Life Click Open File > Models Library >
Computer Science > Cellular Automata > LifeLife
Slide 6
6 Gliders are causally powerless. A glider does not change how
the rules operate or which cells will be switched on and off. A
glider doesnt go to a cell and turn it on. A Game of Life run will
proceed in exactly the same way whether one notices the gliders or
not. A very reductionist stance. But One can write down equations
that characterize glider motion and predict whetherand if so whena
glider will turn on a particular cell. What is the status of those
equations? Are they higher level laws? Gliders Like shadows, they
dont do anything. The rules are the only forces! Good GoL
website
Slide 7
7 Amazing as they are, gliders are also trivial. Once we know
how to build a glider, its simple to make as many of them as we
want. Can build a library of Game of Life patterns and their
interaction APIs. By suitably arranging these patterns, one can
simulate a Turing Machine. Paul Rendell.
http://rendell.server.org.uk/gol/tmdetails.htm Game of Life as a
Programming Platform A second level of emergence. Emergence is not
particularly mysterious. What does it mean to compute with
shadows?
Slide 8
8 Downward causation The unsolvability of the TM halting
problem entails the unsolvability of the GoL halting problem. How
strange! We can conclude something about the GoL because we know
something about Turing Machines. Yet Turing Machines are just
shadows in the GoL world. And the theory of computation is not
derivable from GoL rules. Downward causation entailment Reduce GoL
unsolvability to TM unsolvability by constructing a TM within the
GoL. Paul Davies, The physics of downward causation in Philip
Clayton (Claremont Graduate University), Paul Davies
(Macquarie/NSW/Arizona State University), The re-emergence of
emergence, 2006
Slide 9
9 A GoL Turing machine is an entity. Like a glider, it is
recognizable; it has reduced entropy; it persists and has
coherenceeven though it is nothing but patterns created by cells
going on and off. obeys laws from the theory of computability. is a
GoL phenomenon that obeys laws that are independent of the GoL
rules while at the same time being completely determined by the GoL
rules. Reductionism holds. Everything that happens on a GoL grid is
a result of the application of the GoL rules and nothing else.
Computability theory is independent of the GoL rules. Just as
Schrdinger said.said Living matter, while not eluding the laws of
physics is likely to involve other laws, [which] will form just as
integral a part of [its] science. Schrdinger
Slide 10
10 Level of abstraction: causally reducible yet ontologically
real A collection of entities and relationships that can be
described independently of their implementation. A Turing machine;
biological entities; every computer application, e.g., PowerPoint.
When implemented, a level of abstraction is causally reducible to
its implementation. You can look at the implementation to see how
it works. Its independent description makes it ontologically real.
How it behaves depends on its description at its level of
abstraction, which is independent of its implementation. The
description cant be reduced away to the implementation without
losing information. If the level of abstraction is about nature,
reducing it away is bad science.
Slide 11
11 Supervenience A set of predicates H (for Higher-level) about
a world supervenes on a set of predicates L (for Lower-level) if it
is never the case that two states of affairs of that world will
assign the same configuration of truth values to the elements of L
but different configurations of truth values to the elements of H.
In other words, L fixes H. Or, no change in H without a change in
L. Think of L as statements in physics and H as statements in a
Higher-level (special) science. Or of L as statements in a computer
program and H as the specification of the programs functionality.
Or of L as a description of cells on the GoL grid (which are on and
which are off) and H as a description of the patterns (like
gliders) on the grid. Developed originally in philosophy of mind in
an attempt to link mind and brain.
Slide 12
12 Supervenience example H: {An odd number of bits is on:.
True, False. The bits that are on are the start of the Fibonacci
sequence: False, False. The bits that are on represent the value
27: False, True. } H supervenes over L1. The truth value of a
statement in H depends on the truth values of the statements in L1.
But not over L2. The H statement An odd number of bits is on. can
be either true or false (by varying bit 2) without changing the
truth values in L2since L2 ignores bit 2. The world in two
different states L1: {Bit 0 is on., Bit 1 is on., Bit 2 is on. Bit
3 is on., Bit 4 is on.} t, t, t, t, t t, t, f, t, t L2: {Bit 0 is
on., Bit 1 is on., Bit 3 is on., Bit 4 is on.} t, t, t, t Models
Library > Biology > Evolution > Peppered Moths Click
Open">
32 Try it out File > Models Library > Biology >
Evolution > Peppered Moths Click Open
Slide 33
33 The evolutionary process There is a population of elements.
The elements are capable of making copies of themselves perhaps
with variants (mutations) and perhaps by combining with other
elements. The environment affects the likelihood of an element
surviving and reproducing. This results in evolution by natural
(i.e., environmental) selection. Darwin likened it to breeding. The
environment plays the role of the breeder.
Slide 34
34 The nature of evolution Moth coloring confers survival value
(fitness)which depends on the environment. Hence Darwins natural
selection, i.e., environmental selection. The environment selects
the winners. There may be multiple winners. All one needs is a
niche, not domination. Moth coloring confers survival value
(fitness)which depends on the environment. Hence Darwins natural
selection, i.e., environmental selection. The environment selects
the winners. There may be multiple winners. All one needs is a
niche, not domination. Nature is not necessarily red in tooth and
claw. The dark and light moths dont compete directly with each
other. Survival of the fittest doesnt mean survival of the
strongest. It means survival of those that best fit the
environment. There are no moth-on-moth battles. Nor do the dark
moths attempt to convince the light moths that its better to be
dark or vice versa. Moths (and their colors) are rivals, not
adversaries. Its more like a race than a boxing match. They are
rivals with respect to their ability to survive and acquire
resources from the environment. Nature is not necessarily red in
tooth and claw. The dark and light moths dont compete directly with
each other. Survival of the fittest doesnt mean survival of the
strongest. It means survival of those that best fit the
environment. There are no moth-on-moth battles. Nor do the dark
moths attempt to convince the light moths that its better to be
dark or vice versa. Moths (and their colors) are rivals, not
adversaries. Its more like a race than a boxing match. They are
rivals with respect to their ability to survive and acquire
resources from the environment.
Slide 35
35
Slide 36
36 Six time scales of evolution Social/economic/cultural
systems evolve at medium speeds. As rivals: a social system that
does well for its members thrives and expands. As adversaries:
social systems sometimes compete for resourcesland in the past; now
other resources. Social/economic/cultural systems evolve at medium
speeds. As rivals: a social system that does well for its members
thrives and expands. As adversaries: social systems sometimes
compete for resourcesland in the past; now other resources. Markets
are evolution speeded-up. Coke and Pepsi are rivals for consumer
dollars, not adversaries. They dont attempt to kill each others
CEOs or to sabotage each others delivery trucks. Markets are
evolution speeded-up. Coke and Pepsi are rivals for consumer
dollars, not adversaries. They dont attempt to kill each others
CEOs or to sabotage each others delivery trucks. Warfare: often
super fast evolution. IED tactics and counter tactics. Warfare:
often super fast evolution. IED tactics and counter tactics.
Biological evolution is generally slow. Thought: thinking through
options is even faster. Let ones hypotheses die in ones stead. Karl
Popper Thought: thinking through options is even faster. Let ones
hypotheses die in ones stead. Karl Popper Simulation: computer
modeling of evolutionary processes is faster yet.
Slide 37
37 Application to engineering problems The Traveling Salesman
Problem (TSP). Connect the cities with the shortest tour that is a
permutation of the cities. Starts and ends at the same city.
Includes each city exactly once. In this case the problem is easy
to solve by inspection. In general, its computationally explosive
since there are (n-1)! possible tours. B B A A D D E E C C 20 13 12
14 12 7 9 4 24 The obvious tour will include the sequence ACED-54
(or its reverse). No diagonals: A-E or C-D. The question is where
to put B: ABCED- 55, ACBED-57, or ACEBD-56? Why not n!
Slide 38
38 An exchange (or reverse or mutation) solves this problem in
one step. ACBED-57 ABCED-55 Genetic algorithm approach Create a
population of random tours. AEBCD-59, ACBED-57, ADCBE-59, ACDEB-71,
In this case there are only 4! = 24 possible tours. Could examine
them all. Usually thats not possible. Repeat until good enough or
no improvement. But beware local optima. Select one or two tours as
parents, e.g., AEBCD and ACBED. Ensure that better tours are more
likely to be selected. Generate offspring using genetic operators
to replace poorer elements. Exchange two cities: ACDEB-71 ACBED-57
Reverse a subtour: ACBED-57 AEBCD-59 (Re)combine two tours:
AEBCD-59 & ACBED-57 AEDCB-71. Possibly mutate the result:
ADCBE-59 ACBDE-70 B B A A D D E E C C 20 13 12 14 12 7 9 4 24
Slide 39
39 Try it out: TSP.jar After starting a run, double click in
the display area to add a city or on a city to remove it. New
cities are added to the tour next to their nearest neighbor. Stop
and restart for new random cities. The number of new cities will be
the same as the number of old cities. The differences between the
current best and its immediate predecessor are shown by link color.
New links are shown in green. Removed links are in dashed magenta.
No geographical heuristics are used. Just the structural ones shown
on the previous slide.
Slide 40
40 Genetic algorithms: parameter setting/tuning The number of
variables is constant. Both the TSP and the peppered moths examples
illustrate genetic algorithms. Peppered moths: one parameter
(color) to set. TSP: N variables. As a parameter setting problem
think of each tour as consisting of N variables, each of which may
contain any city number. The additional constraint is that no city
may repeat. Often there are hundreds of variables (or more) or the
search space is large and difficult to search for some other
reason. There is no algorithmic way to find values that optimize
(maximize/minimize) an objective function. Terrile et. al. (JPL),
Evolutionary Computation applied to the Tuning of MEMS gyroscopes,
GECCO, 2005. Abstract: We propose a tuning method for MEMS
gyroscopes based on evolutionary computation to efficiently
increase the sensitivity of MEMS gyroscopes through tuning and,
furthermore, to find the optimally tuned configuration for this
state of increased sensitivity. The tuning method was tested for
the second generation JPL/Boeing Post-resonator MEMS gyroscope
using the measurement of the frequency response of the MEMS device
in open-loop operation.
Slide 41
41 Genetic programming: design The number of variables (and the
structure of the possible solution) is not fixed. Original goal was
to generate software automatically. Not very successful, but hence
the name. Applied successfully to other design and analysis
problems. Circuit design Lens design Bongard and Lipson (Cornel),
Automated reverse engineering of nonlinear dynamical systems, PNAS,
2007. Abstract: Complex nonlinear dynamics arise in many fields of
science and engineering, but uncovering the underlying differential
equations directly from observations poses a challenging task. The
ability to symbolically model complex networked systems is key to
understanding them, an open problem in many disciplines. Here we
introduce for the first time a method that can automatically
generate symbolic equations for a nonlinear coupled dynamical
system directly from time series data. This method is applicable to
any system that can be described using sets of ordinary nonlinear
differential equations, and assumes that the (possibly noisy) time
series of all variables are observable. Symbolic regression
Slide 42
42 The Human-competitive awards: Humies Each year at the
Genetic and Evolutionary Computing Conference (GECCO), prizes are
awarded to systems that perform at human-competitive
levelsincluding the previous two slides. See
http://www.genetic-programming.org/hc2005/main.html An
automatically created result is considered human-competitive if it
satisfies at least one of the eight criteria below. A.The result
was patented as an invention in the past, is an improvement over a
patented invention, or would qualify today as a patentable new
invention. B.The result is equal to or better than a result that
was accepted as a new scientific result at the time when it was
published in a peer-reviewed scientific journal. C.The result is
equal to or better than a result that was placed into a database or
archive of results maintained by an internationally recognized
panel of scientific experts. D.The result is publishable in its own
right as a new scientific result independent of the fact that the
result was mechanically created. E.The result is equal to or better
than the most recent human-created solution to a long-standing
problem for which there has been a succession of increasingly
better human-created solutions. F.The result is equal to or better
than a result that was considered an achievement in its field at
the time it was first discovered. G.The result solves a problem of
indisputable difficulty in its field. H.The result holds its own or
wins a regulated competition involving human contestants (in the
form of either live human players or human-written computer
programs). John Koza
Slide 43
43 Genetic Algorithm for Constellation Optimization (GACO)
Finds optimal constellation orbits using a genetic algorithm under
multiple design constraints and with multiple sensor types. For low
number of sats, GA arrangement is significantly better than
Walker
Slide 44
44 Principles of Complex Systems: How to think like nature
Organizational innovation Russ Abbott
Slide 45
45 Innovative environments Net-centricity and the GIG Inspired
by the web and the internet Goal: to bring the creativity of the
web and the internet to the DoD What do innovative environments
have in common? How can organizations become innovative? What do
innovative environments have in common? How can organizations
become innovative? Other innovative environments Market economies
Biological evolution The scientific and technological research
process
Slide 46
46 The innovative process: exploration and exploitation
Innovation, including human creativity, is always the result of an
evolutionary process. If I were to give an award for the single
best idea anyone has ever had, I'd give it to Darwin, ahead of
Newton and Einstein and everyone else. In a single stroke, the idea
of evolution by natural selection unifies the realm of life,
meaning, and purpose with the realm of space and time, cause and
effect, mechanism and physical law. Daniel Dennett, Darwin's
Dangerous Idea Generate new variants (e.g., ideas)typically by
combining and modifying existing ones. This is a random process in
nature. But random or not isnt the point. The point is to generate
lots of possibilities, to explore the landscape. (Select and)
exploit the good ones Allow/enable the good ones to flourish. The
hard part! The easy part!
Slide 47
47 Exploration and exploitation in nature Evolution. E. Coli
navigation. The immune system. Ant and bee foraging. Termite nest
building (to come). Building out the circulatory and nervous
systems.
Slide 48
48 Exploration and exploitation: like water finding a way down
hill Quite a challenge! We are very well defended. But we still get
sick! If there is a way, some will inevitably find it. (Murphy's
law?) The trick is to make the inevitability work for you, not
against you. Microbes attempting to get into your body must first
get past your skin and mucous membranes, which not only pose a
physical barrier but are rich in scavenger cells and IgA
antibodies. Next, they must elude a series of nonspecific
defensesand substances that attack all invaders regardless of the
epitopes they carry. These include patrolling phagocytes,
granulocytes, NK cells, and complement. Infectious agents that get
past these nonspecific barriers must finally confront specific
weapons tailored just for them. These include both antibodies and
cytotoxic T cells. From a tutorial on the immune system from the
National Cancer InstituteFrom a tutorial on the immune system from
the National Cancer Institute.
Slide 49
49 Exploration, exploitation, and asymmetric warfare It is the
nature of complex systems and evolutionary processes that conflicts
become asymmetric. No matter how well armored one is there will
always be chinks in the armor, and something will inevitably find
those chinks. The something that finds those chinks will by
definition be asymmetric since it attacks the chinks and not the
armor.
Slide 50
50 Exploration and exploitation: groups and individuals
Successful group exploration typically requires multiple, loosely
coordinated, i.e., autonomous, individuals. Thats because nature is
not regular; one cant fully plan an exploration. If one knew in
advance what the landscape looked like, it wouldnt be an
exploration. Much exploration is wasted effort. One may hit the
jackpot while the others find nothing.
Slide 51
51 Exploration and exploitation: groups and individuals For a
group to benefit from the discoveries of individuals, there must be
mechanisms to bring the discoveries back and allow the group them
to use them. Mechanisms to internalize successful/promising
discoveries must be built into a groups process. This frequently
requires creative destruction, which may be more difficult to
acceptespecially if its your job that is being destroyed. Markets
are how we integrate creative destruction into society. Recall ant
foraging and pheromone following. Joseph Schumpeter, Capitalism,
Socialism, and Democracy Its amazing how well we have tamed
destruction. Its now an accepted part of our normal processes.
Slide 52
52 How does this apply to organizations? To ensure innovation:
Sounds simple doesnt it? Exploration: creation and trial Encourage
the prolific generation and trial of new ideas. Exploitation:
establish the successful variants Allow new ideas to flourish or
wither based on how well they dorather than for political
reasons.
Slide 53
53 Initial funding Prospect of failure ApprovalsEstablishment
Biological evolution Capitalism in the small. Nature always
experiments. Most are failures, which means death. (But no choice
given.) None. Bottom-up resource allocation defines success.
Entrepreneur Little needed for an Internet experiment. Perhaps some
embarrassment, time, money; not much more. Few. Entrepreneur wants
rewards. Bottom-up resource allocation. Bureaucracy Proposals,
competition, forms, etc. When 100% Mission Success is the group
goal, who wants a failure in his/her personnel file? Far too many.
Managers have other priorities. Top-down resource allocation. New
ideas arent the problem. Trying them out Innovation in various
environments Getting good ideas established We save ourselves by
spin-doctoring and benign neglect Viagra: making friends in
Afghanistan
Slide 54
54 Garages and laboratories, workbenches, and scribbled napkins
are filled with brilliant ideas unmatched with determination,
resources, and market sensibilities. Jack Russo, Silicon Valley
intellectual-property lawyer. In 1999, when Nathan Myhrvold left
Microsoft (formerly CTO; brilliant, but he missed the importance of
the web) he set himself an unusual goal. He wanted to see whether
the kind of insight that leads to invention could be engineered. He
formed a company called Intellectual Ventures. He raised hundreds
of millions of dollars. He hired the smartest people he knew. It
was not a venture-capital firm. VVenture capitalists fund existing
insights. They let the magical process that generates new ideas
take its course, and then they jump in. Myhrvold wanted to make
insightsto come up with ideas, patent them, and then license them
to interested companies. Malcolm Gladwell (May 12, 2008) In the
Air, The New Yorker,
http://www.newyorker.com/reporting/2008/05/12/080512fa_fact_gladwell
http://www.newyorker.com/reporting/2008/05/12/080512fa_fact_gladwell
Matt Richtel (March 30, 2008) Edison...Wasnt He the Guy Who
Invented Everything?, New York Times,
http://www.nytimes.com/2008/03/30/weekinreview/30richtel.html
http://www.nytimes.com/2008/03/30/weekinreview/30richtel.html
Slide 55
55 Planned invention? When Myhrvold started out, his
expectations were modest. Although he wanted insights like
Alexander Graham Bells, Bell was clearly one in a million, a genius
who went on to have ideas in an extraordinary number of areassound
recording, flight, lasers, tetrahedral construction, and hydrofoil
boats, to name a few. Invention has its own algorithmsome
combination of genius, obsession, serendipity, and epiphany. How
can you plan for that? The original expectation was that I.V. would
file a hundred patents a year. Its filing five hundred a year and
has a backlog of three thousand ideas. It just licensed off a
cluster of patents for $80,000,000. Its ideas are not trivial.
IImproved jet engines NNew techniques for making microchips AA way
to custom-tailor the mesh sleeve used to repair aneurysms
AAutomatic, battery-powered glasses, with a tiny video camera that
reads the image off the retina and adjusts the fluid-filled lenses
accordingly, up to ten times a second. Why doesnt every technical
organization do this?
Slide 56
56 VC arithmetic Its hard work to harness the power of
innovation.
Slide 57
57 Practical organizational innovation Hamel and Skarzynski: an
innovation architecture. An innovation pipeline for managing and
opportunities A core set of people trained in the processes of
innovation A systematic process for generating and managing
strategic insights The right evaluative criteria at every stage of
the development process to prevent potentially valuable ideas from
being killed off prematurely Ideas that are sufficiently radical to
deliver breakthroughs Mechanisms for rapidly reallocating resources
behind new opportunities Mechanisms to manage growth opportunities
with different timescales and risk profiles Metrics to measure
innovation performance Linkages between innovation and management
compensation A self-sustaining enterprise capability and a tangible
core value Prediction (20 years). To survive outside a protected
environment an organization will need a successfully functioning
innovation architecture. Corollary. Some organizations will focus
on preserving their environments. Do we have an innovation
architecture? Most companies want to make money. What metrics would
we use?
Slide 58
58 Successful innovative organizations: W.L. Gore, Best Buys,
Whole Foods, GE, Whirlpool, P&G, CEMEX, Google Lower levels
discover opportunities through exploration. New initiatives often
grow from the edges, where perception occurs. Constrained by rules
of engagement, which protect them from harm. Must be possible for
initiatives to originate at all levelseven the top. Higher/broader
levels provide perspective, impose constraints, shape direction,
and add or withhold resources as events develop. They do not
primarily issue commands. This is primarily a bottom-up model of
resource allocation. Decisions about increasingly significant
commitments are made at increasingly higher/broader levels. If the
entire organization commits, becomes an organization/organism-level
goal. Top-level strategy: stay healthy and build skills, resources,
and capabilities that can be recruited/applied/committed when
needed. Lower levels discover opportunities through exploration.
New initiatives often grow from the edges, where perception occurs.
Constrained by rules of engagement, which protect them from harm.
Must be possible for initiatives to originate at all levelseven the
top. Higher/broader levels provide perspective, impose constraints,
shape direction, and add or withhold resources as events develop.
They do not primarily issue commands. This is primarily a bottom-up
model of resource allocation. Decisions about increasingly
significant commitments are made at increasingly higher/broader
levels. If the entire organization commits, becomes an
organization/organism-level goal. Top-level strategy: stay healthy
and build skills, resources, and capabilities that can be
recruited/applied/committed when needed. Just what your mother
always told you: eat right, exercise, get plenty of sleep, study
hard, practice, and save money.
Slide 59
59 Innovation in the military Our military is deliberately
mission drivenwhere the missions are determined by civilian
authority. We dont want our military to take the initiative to find
new missions for itself. What kinds of innovation does it make
sense for the military to attempt? Innovation that makes it more
effective at doing what it is charged to do. How can success be
made self-validating in the way making money or reproducing are?
Innovations that save lives (e.g., anti-IED techniques) are
self-validating and are adopted relatively quickly. Need a way to
aggregate resources/success bottom-up. Establish a
military-specific innovation architecture. Our military is
deliberately mission drivenwhere the missions are determined by
civilian authority. We dont want our military to take the
initiative to find new missions for itself. What kinds of
innovation does it make sense for the military to attempt?
Innovation that makes it more effective at doing what it is charged
to do. How can success be made self-validating in the way making
money or reproducing are? Innovations that save lives (e.g.,
anti-IED techniques) are self-validating and are adopted relatively
quickly. Need a way to aggregate resources/success bottom-up.
Establish a military-specific innovation architecture.
Slide 60
60 Principles of Complex Systems: How to think like nature
Design: platforms, services, cycles, and energy flows Russ
Abbott
Slide 61
61 How would you gather wood chips into a pile? Probably not
like this. File > Models Library > Biology > Termites
Click Open
Slide 62
62 Termite rules Wander about aimlessly (randomly) until you
bump into a wood chip. If you are not holding a wood chip Pick up
the new chip. Move away from your current location. Go back to
wandering about aimlessly. If you are holding a wood chip Put down
your chip in a nearby empty space. Move away from your current
location. Go back to wandering about aimlessly. Net effect: wood
chips are deposited near other wood chips, eventually forming a
single pile. Wikipedia commons Run the program and watch what
happens. Exercise: prove that this will always happen
Slide 63
63 No horizontal communication. No dashed lines. (Is that
good?) Its not accurate as a communication or operational
structure. It may represent how authority is delegated, and it may
represent how responsibility is assigned, but it doesnt represent
how communication occurs or how organizations really work. Downward
pointing arrows: commands. Upward pointing arrows: results/reports.
Can be implemented with point-to-point communication links.
Organizational/system structure: Whats wrong with this
picture?
Slide 64
64 A somewhat more realistic picture The focus is on
interaction among participants in the organization. David Sloan
Wilson, Evolution for Everyone Everything is both an entity and a
group.
Slide 65
65 From point-to-point links to platforms Need more than fixed
point-to- point communication channels The communication system
(even if just a telephone system) is the start of net-centricity
Must distinguish between communication structure and command
hierarchy. Becomes reified as an additional componentnot just a
collection of interfaces. Platform But a network/platform may do
nothing on its own. The fundamental question How will the
organization use the network/platform? Enabling communication
neither eliminates responsibility nor undermines command intent. As
a common resource, how does it fit into the hierarchy? How is it
governed?
Slide 66
66 Layered architectures not functional decomposition
Presentation Session Transport Network Physical WWW (HTML) browsers
+ servers Applications, e.g., email, IM, Wikipedia Asymmetric
warfare Each layer is a platform that a)is built on the layers
below it b)enables higher level layers to be built on top of it
c)is vulnerable to disruption.
Slide 67
67 How does Aerospace send mail to Aerospace? El Segundo mail
routes (carts) Chantilly mail routes (carts) LAX IAD (commercial
aircraft) Aerospace USPS (trucks) Sort & route Many interlinked
processes. The various systems provide platforms for each other.
Infrastructure A system of (many!) systems. Aerospace, USPS,
commercial airlines, airports, traffic, road maintenance, A system
of (many!) systems. Aerospace, USPS, commercial airlines, airports,
traffic, road maintenance,
Slide 68
68 Multi-sided software platforms Evans, Hagiu, and Schmalensee
(2006) Invisible Engines: How Software Platforms Drive Innovation
and Transform Industries, MIT Press. (freely downloadable)
Operating systems, the web browser. Markets/mechanisms that connect
disparate groups. A stock exchange matches buyers and sellers. A
credit card system matches merchants and cardholders. Shopping
centers, dating websites, TV channels, TV talk shows, Amazon
resellers, telephone & telegraph systems. Large retail stores
(Wal-Mart, supermarkets) rent shelf space. Not your usual business
model: buy; add value; sell. The value to each group increases as
the size of the other group(s) grow. (Also known as network
effect.)
Slide 69
69 Platforms as refactorings A multi-sided platform may be
understood as the standardization and factoring out (refactoring)
of a hard part of an interaction and providing it as a service. The
hard part is done by the platform. USPS: sending & receiving
materials. Credit card: paying and being paid. Dating service:
finding the other party and making an initial contact. Roberts
Rules of Order: the interaction protocol
Slide 70
70 Standards as (ephemeral) platforms Since a platform is a
level of abstraction, it can be characterized by a specification.
The specification can then serve as the definition of the platform,
e.g., HTML, SQL, . Multiple vendors can be encouraged to compete to
implement it. Defangs platform owners. Empowers platform users.
Some platforms are single-sided: programming languages, automobile
& public transportation system, (woodworking, etc.) tools. Have
similar value.
Slide 71
71 Platforms as infrastructure and environments Sometimes
platforms define an environment. The free market economic system is
defined primarily by two platforms. The monetary and banking
system. Factors out the economic notion of value. Allows value to
be abstracted, stored, exchanged with minimal overhead. The legal
and judicial system. Factors out agreements (contracts) and
enforcement mechanisms. Overhead not so minimal (lawyers) but
better than hiring your own enforcers. Used to rely more on
reputation. Still do in eBay. Much too important to be controlled
privately. In general, the set of platforms available in an
environment is the environments infrastructure.
Slide 72
72 Governance and change Once a platform has been established,
what mechanisms are available so that it can evolve as needed?
Since its use is embedded in the workings of many users, its
difficult to change. Since its use is central to the survival of
many users it must be able to change as needed. Platforms Open at
the bottom. Its the interface that matters, not the implementation.
Open at the top. New uses are encouraged. Stable but slowly
changing. Must be stable enough to be relied on but flexible enough
to change as needed.
Slide 73
73 An unusual platform-based design E. coli can produce lactase
which digests lactose. But for efficiency sake it should produce
lactase only when lactose is present. Imagine that you were asked
to design a system that would produce a product only under certain
conditions. How would you do it?
Slide 74
74 A (quasi-top-down) functional analysis solution Lactose
sensor How does one know one can build these pieces? What enables
the interfaces? What holds it all together? The unasked questions
Lactase production system Switch Off On Is it really top-down?
Assumes platforms of components and framework. Engineers can always
design a lower level if needed.
Slide 75
75 E. coli lets lactose flip its own switch Lactose Repressor
Lactose itself binds to the repressor, pulling it out of the way.
lacYlacAlacX RNAP Repressor Three lac genes RNA polymerase cant
bind to DNA. Transcription blocked. lacYlacAlacX RNAP lac operon
Its often said that a first step in systems engineering is to agree
on the system boundaries. What are the system boundaries in this
case? Its often said that a first step in systems engineering is to
agree on the system boundaries. What are the system boundaries in
this case? Lactose Repressor lacYlacAlacX RNA polymerase can now
bind to DNA. Transcription enabled. The genes are expressed. RNAP
Wheres the platform? The DNA protein processing system. Once that
processing cycle was built, nature found out how to turn it on and
off with gene switches. It then became possible to use that
mechanism to allow lactose to turn on generations of its own
digestive enzyme.
Slide 76
76 Principles of Complex Systems: How to think like nature
Energy, platforms, cycles, and complex systems New section.
Preliminary thoughts. Russ Abbott
Slide 77
77 No general framework for describing the organization or
functioning of a complex system. Lets make a list of complex
systems. Distinguish complex from complicated. An economy. A
biological organism. An ecological system. Most large
organizations. Systems that give rise to (dynamic) emergent
phenomena(?). What is it about such systems that allows that to
happen? -(Initially) unplanned interactionsthat (generally) arent
one-shot affairs. Multiple autonomous agents interacting within an
environment. Thats why agent-based simulations tend to be useful.
Is that why theyre called complex? Is that why theyre called
complex? No. Cycles, Design, and Requirements Salt doesnt count.
Will this really provide one? Ambitious. Audacious.
Slide 78
78 Complex systems are often said to be on the boundary between
chaos and order, i.e., on the edge of chaos. Cosma Shilizi thinks
its a terrible term.thinks In other words, they are neither
(easily) predictable nor chaotic where easily predictable means
that it can be predicted more easily than simulating it, what Bedau
called weak emergence. Are there other alternatives? Shalizi doesnt
offer one. Something that is growing at a predictable (even
exponential) rate is still (easily) predictable. A Universal Turing
machine is neither chaotic nor easily predictable since it must be
run to determine what it will do. Most people wouldnt consider a
Turing machine a complex system. Most people wouldnt consider the
computation of a Turing machine (even weakly) emergent. But recall
levels of abstraction! What about stability?
Slide 79
79 How is stability defined? Stability implies some relatively
constant features. What about a pot of steadily boiling wateror
most of the other dissipative structures of Prigogine? What do
dissipative structures dissipate? Energy. Energy flows through a
dissipative structure. Will return to this in a few slides. At
maximum entropy and hence stable with respect to its internal
energy contenteven though it is experiencing a constant flow of
energy? E.g., Rayleigh-Bnard cells?
Slide 80
80 Basins of attraction Two kinds of attractorswith respect to
energy considerations Homeostatic mechanisms, e.g., a
economy/biological organism/ecology. Require continual supply of
energy. But also at max entropy? Energy wells, e.g., a lake/ocean.
At energy equilibrium. The Lorenz attractor is neither. Energy isnt
a consideration. This is an important (and often overlooked)
feature of computation. Lorenz attractor
Slide 81
81 The agent as a Persistent Turing Machine If you ignore
quantum/random/creative phenomena, each agent can be characterized
by a computer programa computation. Or agent-based simulations
wouldnt work. Each computation is an endless loopor the agent dies.
The agents program is not an algorithmsince (by definition) to be
an algorithm a computation must terminate. See Goldin and Wegner,
The Interactive Nature of Computing: Refuting the Strong
Church-Turing Thesis (Minds and Machines, 18/1, March 2008, 1738)
on Persistent Turing Machines.The Interactive Nature of Computing:
Refuting the Strong Church-Turing Thesis A TMs state transition
rules and execution cycle are both finite. The Turing machine
itself may do arbitrary computation if unlimited memory is
available. For agents that represent biological organisms internal
memory is limited. Any unlimited memory must be externalin the
environment. The code, a finite description of the program. What it
does when it runs.
Slide 82
82 Cycles: the underlying mechanism Based on the previous
considerations, agents with internal cyclese.g., the TM state
transition cycleare the underlying mechanism for complex systems.
Cycles are also central to platforms. A platform most generally is
a servicein the sense of Service Oriented Architecture. It can be
modeled as an agent. If a platform/service could not be described
finitely, we couldnt build it. So a platform/service is a PTM.
Every complex system is built on a collection of basic PTMs.
Somethe widely used servicesare more central than others. Each PTM
is defined by its underlying state transition cycle. So each
complex system is built on top of a set of basic cycles.
Slide 83
83 But not the usual way we think of cycles. The Nitrogen
cycle. What are the real cycles? Each processing step is a service
powered by an internal cycle.
Slide 84
84 Impossible cycle: cant always go downhill
Slide 85
85 What powers the cycles? The invisible element: energy Each
processing step is a service powered by an internal cycle.
Slide 86
86 Far from equilibrium Generally applied to systems that have
energy flowing through them. Not at equilibrium. But often stable.
Importantand often overlookedfeatures of complex systems are the
energy flows through them. Follow the flows, especially the energy
flows. Money is a way to store valueincluding energy. Grasslands
Conversation Council of British Columbia, Canada But I dont see
these three as equivalent. Energy powers the mechanisms that
perform the other cycles. Is water different from nutrients or are
they all resources? Not a cycle! Need source and sink.
Slide 87
87 Infrastructure So the key to many systems is to understand
the systems primary services/platforms, what flows they enable, and
how they are powered. These make up the systems infrastructure. Not
all flows are cyclesespecially energy. Resource flows increasingly
areand generally require energy for recycling.
Slide 88
88 Morowitz theorem: A system with an energy flow must have a
cycle. The flux [produced by the flow of energy through a system]
is the organizing factor in a dissipative system. When energy flows
in a system from a higher kinetic temperature, the upper energy
levels of the system become occupied and take a finite time to
decay into thermal modes. During this period energy is stored at a
higher free energy than at equilibrium state. Systems of complex
structures can store large amounts of energy and achieve a high
amount of internal order. Therefore, a dissipative system develops
an internal order with a stored free energy that is stable, has a
lower internal entropy and resides some distance from thermostatic
equilibrium. Furthermore, a dissipative system selects stable
states with the largest possible stored energy. The cyclic nature
of dissipative systems can be seen in the periodic attractors.
Their cyclic nature allows them to develop stability and structure
within themselves. As paraphrased by Piero Scaruffi.Piero
Scaruffi
Slide 89
89 The fundamental cycles Most systems have fundamental cycles
(services) that all the other processes ride on top of. In a
computer, its the instruction execution cycle. Ride on top of means
that higher level processes are built by running the basic cycles
to implement the higher processes. (Recall emergence and levels of
abstraction.) A good way to understand a complex system is to
identify those fundamental services and to understand how they are
powered. In the ecology/nitrogen example, one of the fundamental
cycles is the DNA Protein generation cycle. Requires both energy
(ATP) and resources (amino acids). Bacteria and all other living
things depend onand ride on top ofit as well as other basic
cycles.
Slide 90
90 Services/platforms/cycles and system acquisition Traditional
system design starts with a needs statement, from which a
requirements document is generated, from which the system
specification and design documents are created. From requirements
on down these documents provide a static view of the systemwhat
will be sold and delivered. The system is often described in a
top-down manner. What else is possible? But whats it all for? The
original need (generally a capability, i.e., a service) is
frequently lost. The useConcept of Operations (CONOPS)is not
ignored, but it often takes second place to other
requirements.
Slide 91
91 A CONOPS version of acquisition A capability-based
approachnot a new idea! starts with the system as a service and
maintains that perspective. Terrible as he was as a foreign policy
advisor, Rumsfeld had some good ideas with respect to defence
acquisition. A capability-based, i.e., service-based view of a
system to be acquired starts at the operational level and stays as
close to that level as possible.
Slide 92
92 How would it work? How will it fit into the context of all
other existing and expected systems? How will it be run day-to-day?
Who are its intended users? Can that user-set be enlarged? Will it
have a Community of Interest? What will the COI role be in defining
requirements? Envisage the use of the intended service in the
context of all other services. What are its fundamental cycles? How
are they powered? How is higher level functionality built on top of
them? What are its operational resources and how will they be
supplied? How will its operation be paid for? Do the people who
will pay for operating it really want it? How will it be governed?
Who are the stakeholders? Who are its intended owner/operators? How
will decisions be made about enhancements?
Slide 93
93 Principles of Complex Systems: How to think like nature
Organizations: how nature builds systems; the wisdom of crowds
Innovation required individual autonomy. What do groups add? Russ
Abbott
Slide 94
94 Self-organizing groups Craig Reynolds wrote the first
flocking program two decades ago: http://www.red3d.com/cwr/boids.
http://www.red3d.com/cwr/boids Heres a good current interactive
version: http://www.lalena.com/AI/Flock/
http://www.lalena.com/AI/Flock/ Separation: Steer to avoid crowding
birds of the same color. Alignment: Steer towards the average
heading of birds of the same color. Cohesion: Steer to move toward
the average position of birds of the same color.
Slide 95
95 Self-organizing groups: how nature builds systems Debora
Gordon on ant colonies Debora Gordon on ant coloniesThe bird,
termite, and ant models illustrate emergence (and multi-scalarity).
(See video Debora Gordon on ant colonies.)Debora Gordon on ant
colonies More recent talk (1 hour):
http://www.youtube.com/watch?v=R07_JFfnFnY In both cases,
individual, local, low-level rules enabled the group to achieve
emergent higher level results. The birds flocked. The wood chips
were gathered into a single pile. The food was brought to the nest.
These systems are the product of the evolution of individual
actions that resulted in coordinated benefits. Emergence is
successful group design. Group exploration extends the perceptual
reach of any individual. Group behavior extends the functional
capability of any one individual. Group exploration extends the
perceptual reach of any individual. Group behavior extends the
functional capability of any one individual. Virtually everything
is both an entity and a group.
Slide 96
96 Breeding groups Chickens are fiercely competitive for food
and water. Commercial birds are beak-trimmed to reduce
cannibalization. Breeding individual chickens to yield more eggs
compounds the problem. Chickens that produce more eggs are more
competitive. Instead Muir bred chickens by groups. At the end of
the experiment Muir's birds' mortality rate was 1/20 that of the
control group. His chickens produced three percent more eggs per
chicken and (because of the reduced mortality) 45% more eggs per
group. Group (and more generally multi-level) selection is now
accepted as valid. Traditional evolutionary theory says there is no
such thing as group selection, only individual selection. Bill Muir
(Purdue) demonstrated that was wrong. Wikipedia commons
http://www.ansc.purdue.edu/faculty/muir_r.htm Groups are entities.
You and I are both entities and cell colonies.
Slide 97
97 But then groups found that coordination, specialization, and
coordinated specialization enabled emergence. Consider any
multi-cellular organism, or any organism with multiple organs, or
any society with any sort of specialization, or any social grouping
with coordinated and/or specialized roles. These groups exemplify
real emergence. Entirely new capabilities appear. Wind instruments
can play melodies. Piano and guitar can play chords as well. Why
groups? Perhaps groups formed initially because they increased
survival value. A team will generally beat an individual of
approximately the same skill level. This is not so much emergence
as power in numbers. Why groups? Two steps.
Slide 98
98 David Sloan Wilson on social groups What holds for chickens
holds for other groups as well: teams, military units,
corporations, religious communities, cultures, tribes, countries.
Successful groups are those that minimize within-group conflict and
organize to succeed at between-group conflict. Groups with
mechanisms for working together can often accomplish far more
(emergence) than the sum of the individuals working separately.
Corporations, military organizations; reproduction; mitochondria
and us. But if a group good is also an individual good (e.g.,
money, security), the group must have mechanisms to limit cheating
(free-ridership). Group traits (although they are carried as rules
by individuals) evolve because they benefit the group. (E.g.,
insect behavior.) These traits may be transmitted genetically (by
DNA). They may also be transmitted culturally (by
training/parenting/indoctrination/mentoring/). Human groups can be
more complex because its not all built-in. Moral systems are
interlocking sets of values, practices, institutions, and evolved
psychological mechanisms that work together to suppress or regulate
selfishness and make social life possible. Jonathan Haidt We
evolved to be pro-social within groups but xenophobic between
groups. Michael Shermer
Slide 99
99 Experimental games Prisoners Dilemma. One shot. Defect is
the only rational strategy. Iterated. Tit-for-tat: Cooperate
initially and then copy the other guy. Pavlov: repeat on success;
change on failure. (More robust.) Ultimatum Game. Proposer must
offer to divide $100. Responder either accepts the proposed
division or rejects itin which case neither gets anything. Only
rational strategy: proposer offers as little as possible; responder
always accepts. Real experiments (world-wide). Responder rejects
unless offer ~1/3. Some societies are different, e.g., where giving
a gift means power. What would you offer/accept? Try it. (Played
anonymously. Write offer.) Try it table against table. Each table
prepares an offer. CD C3/30/5 D5/01/1 A far-from-equilibrium
system. New energy is supplied for free.
Slide 100
100 Homo economicus vs. strong reciprocity Homo economicus:
individual selection Agents care only about the outcome of an
economic interaction and not about the process through which this
outcome is attained (e.g., bargaining, coercion, chance, voluntary
transfer). Agents care only about what they personally gain and
lose through an interaction and not what other agents gain or lose
(or the nature of these other agents intentions). Except for
sacrifice on behalf of kin, what appears to be altruism (personal
sacrifice on behalf of others) is really just long-run material
self-interest. Ethics, morality, human conduct, and the human
psyche are to be understood only if societies are seen as
collections of individuals seeking their own self-interest. Moral
Sentiments and Material Interests: The Foundations of Cooperation
in Economic Life Herbert Gintis, Samuel Bowles, Robert T. Boyd, and
Ernst Fehr (eds), MIT Press, 2005.
Slide 101
101 Homo economicus vs. strong reciprocity Strong reciprocity:
group selection A predisposition to cooperate with others, and to
punish (at personal cost, if necessary) those who violate the norms
of cooperation even when it is implausible to expect that these
costs will be recovered at a later date. Strong reciprocators are
both conditional cooperators They behave altruistically as long as
others are doing so as well. and altruistic punishers They apply
sanctions to those who behave unfairly even at a cost to
themselves. Socialization: norm internalization. There's no such
thing in biology, economics, political science, or anthropology.
Humans can want things even when they are costly to ourselves
because we were socialized to want them: to be fair, to share, to
help your group, to be patriotic, to be honest, to be trustworthy,
to be cheerful.
Slide 102
102 Wise crowds: more than the sum of their parts Web wise
crowd platforms Wikis Mailing lists Chat rooms Prediction markets
Condorcet Jury Theorem (18 th century) example Five people (a small
crowd). Each person has a 75% chance of being right. Probability
that the majority will be right: ~90% With 10 people: ~98%. Simple
if you think about it. Traditional wise crowds Teams Juries
Democratic voting
Slide 103
103 Wise crowd criteria Diverse: different skills and
information brought to the table. Decentralized and with
independent participants: No one at the top dictates the crowd's
answer. Each person is free to speak his/her own mind and make own
decision. Distillation mechanism: to extract the essence of the
crowd's wisdom. Participant autonomy. James Surowiecki, The Wisdom
of Crowds
Slide 104
104 Example from The Difference Which person from the following
list was not a member of the Monkees (a 1960s pop band)? (A) Peter
Tork (B) Davy Jones (C) Roger Noll (D) Michael Nesmith Imagine a
crowd of 100 people with knowledge distributed as follows: 7 know
all 3 of the Monkees 10 know 2 of the Monkees 15 know 1 of the
Monkees 68 have no clue In other words, less than 10 percent of the
crowd knows the answer, and over two- thirds are culturally
deprived of any Monkees knowledge. We assume individuals without
the answer vote randomly. The Condorcet Jury Theorem, then, doesnt
apply because only a small minority knows the answer. Still, the
crowd will have no problem getting the right answer. The 7 who know
all the Monkees vote for Noll; 5 of the 10 who know 2 of the
Monkees will vote for Noll; 5 of the 15 who know 1 of the Monkees
will vote for Noll; and 17 of the 68 clueless will vote for Noll.
So Noll will garner 34 votes, versus 22 votes for each of the other
choices. Diverse groups of problem solvers outperformed the groups
of the best individuals at solving problems. The diverse groups got
stuck less often than the smart individuals, who tended to think
similarly.
Slide 105
105 A wise crowd as assistant and companion
Slide 106
106 Distillation mechanism: prediction markets Statement
Statement Statement issued by 25 world-famous academics. May 2007.
Including: Kenneth Arrow, Daniel Kahneman, Thomas Schelling, Robert
Shiller, Cass Sunstein. Abstract: Prediction markets are markets
for contracts that yield payments based on the outcome of an
uncertain future event, such as a presidential election. Using
these markets as forecasting tools could substantially improve
decision making in the private and public sectors. We argue that
U.S. regulators should lower barriers to the creation and design of
prediction markets by creating a safe harbor for certain types of
small stakes markets. We believe our proposed change has the
potential to stimulate innovation in the design and use of
prediction markets throughout the economy, and in the process to
provide information that will benefit the private sector and
government alike.
Slide 107
107 Often Beats Alternatives Vs. Public Opinion I.E.M. beat
presidential election polls 451/596 (Berg et al 01) Re NFL, beat
ave., rank 7 vs. 39 of 1947 (Pennock et al 04) Vs. Public Experts
Racetrack odds beat weighed track experts (Figlewski 79) If
anything, track odds weigh experts too much! OJ futures improve
weather forecast (Roll 84) Stocks beat Challenger panel (Maloney
& Mulherin 03) Gas demand markets beat experts (Spencer 04)
Econ stat markets beat experts 2/3 (Wolfers & Zitzewitz 04) Vs.
Private Experts HP market beat official forecast 6/8 (Plott 00) Eli
Lily markets beat official 6/9 (Servan-Schreiber 05) Microsoft
project markets beat managers (Proebsting 05) from Robin
HansonRobin Hanson
Slide 108
108 Prediction markets Contracts: Intrade (Ireland-based): real
money or play money.real moneyplay money Panos Ipeirotis But, there
is evidence that prediction markets are not efficient.prediction
markets are not efficient Slates Election Market Page Split off
from TradeSports
Slide 109
109 Concerns and Myths Self-defeating prophecies Decision
selection bias Price manipulation Rich more votes Inform enemies
Share less info Combinatorics Risk distortion Moral hazard Alarm
public Embezzle Bubbles Bozos Lies Crowds dont always beat experts.
People will not work for trinkets. High accuracy is not assured.
from Robin HansonRobin Hanson The prediction markets got both the
New Hampshire and California primaries wrong.
Slide 110
110 Other distillation mechanisms: making the crowds wisdom
actionable Elections, polls, etc. Traditional. Many possible
processes, e.g., transferrable ballots, etc. Expression of
preferences. Many online options (and more options).options
Collaboration: wikis and other collaboration tools (shared spaces),
mailing lists, chat rooms, etc. Explicit: Generation of new work
products. Heres a (long!) list of collaborative work
environments.collaborative work environments Implicit: Googles page
rank, reputations (e.g., eBay), recommendation engines (e.g.,
Amazon) A hard problem. Yet evolution and markets do it
automatically.
Slide 111
111 Principles of Complex Systems: How to think like nature
Remember this Russ Abbott
Slide 112
112 Complex systems Emergence: the creation of a new entity,
one which has new properties (often a group or a system), through
interaction among multiple autonomous elements. Multiscalarity:
everything is both an entity and a group. A level of abstraction
has both a specification (requirements) and an implementation.
Throwing away the specification once an implementation exists
produces a reductionist blind spot. Its the specification (of the
interface) that ensures loose coupling. Interactioneven (or
especially) intra-systemoccurs through an environment. An
environment that provides functionality that facilitates
interaction is a platform. Architectures: agents and platforms vs.
stovepipes and functional decomposition. Platform governance
becomes a fundamental issue. Who owns it, runs it, controls it?
Evolutionary processes are unavoidableleading to unexpected
consequences. They are also the source of all creativity. Their
essence combines exploration with exploitation of discoveries.
Organizations can plan to be innovative. Groups are natures way to
build systems. We can build powerful groups because we evolved to
live in groups and we can learn. How can a groups wisdom be
distilled as action? Bottom-up resource allocation. Nature and
markets have self-validating criteria: reproductive success and
profits. By looking carefully you can see the world in a grain of
sand.