Wicked Complex Systems of Systems

Wicked Complex Systems of Systems

Sandia National LabsJanuary 31, 2007

1January 2007, [email protected], attributed copies permitted

Rick Dove, Industry ProfessorStevens Institute of Technology


SDOE 790 D i f A il S t d E t iSDOE 790 -- Design of Agile Systems and Enterprises: Design Quality and Self Organizing Systems

Esthetic quality in systems and enterprises makes the difference between enforced compliance and embraced experience; and determines the positive or negative vectors of p p ; p gself organization and emergence. This module explores the value and nature of esthetic

design quality, principles and architectures for harnessing self organized extended enterprise, agility as risk management and reality confrontation, and similar issues at the

edge of agile system and enterprise knowledge.

775Thinking

780Engineering

785Architecting

790Designing

g g y p g

Thinking Engineering Architecting Designing

rationale

strategy

fundamentals

architecture

infrastructure

architecting

application

elegance

strategy

science

rationale

vision

agile systems and enterprises

architecture

shape

art

architecting

process

elegance

performance


Why Now?

Nuclear physicsNuclear physicsPersonal computerSemiconductors in everythingSpace travel

Knowledgebuilds on Decisions

must be madef tSpace travel

InternetGlobalizationGenetic engineering

knowledge

The more

faster…

…andimplementedi di t lGenetic engineering

CloningNano-technology

you havethe moreyou get

immediately

Hydrogen economy?Extended lifespan? Quantum computing?

you get

The kneef th

KnowledgeExplosionQuantum computing?

Machine intelligence?

Cold fusion?

of the curveis passed


Cold fusion?Anti-gravity?

What Goes Around Comes Around...Fasthttp://www.cs.bell-labs.com/~ches/map/

A t B Ch i k & H B hInterconnected Art: B.Cheswick & H.Burch

Complexity

Companies

Emergence andUnintended

ConsequencesCompaniesMachinesPeoplePartsB t

Consequences

AOLBBNac.jp

att.net

BotsEtc

UUNetdla.mil

Netcomsprint.net

cw.net (+MCI)bellglobal combellglobal.com

10 Networks61,000 Routers

12/98 Wired MagazineData mid September ‘99

Speed:KnowledgeAnd Response


Data mid-September 99Color based on IP address

(very old news)

pAre Mismatched

Sailing in Uncharted Waters

We don't control the fpace of new

knowledge

Unpredictable, Uncertain,ContinuousContinuous...- New rules- New decisions - New values- New values - New strategies - New priorities- New projects

The World Is Flat, Thomas L. Friedman,First edition cover art


- New projects- New technologies

2005, Farrar, Straus and Giroux

Agility Is ...

The ability toity)

Agile The ability torespond effectively

at all times,ve (V

iabi

l Agile

reactively and proactively...within missionR

eact

iv

Fragile

Proactive (Leadership)

... the ability to survive and thrive in an unpredictable and uncertain environment


Agile-Systems Research

Solution Search:

Examined 100s of systems of various types (products, processes, procedures, peopled)

Looked for systems that responded effectively

Looked for metrics that defined effectively

Looked for categories of response types

Looked for principles that enabled response p p p

(Facilitated by the Agility Forum, Lehigh University, 1991 – 1997


Over 1,000 participants from over 250 organizations)

What Analysis Found (design)

O l t tOne general strategy:Reusable modules Reconfigurable in a Scalable framework

Ten general design principles:1. Evolving Framework Standards2 Encapsulated Modules2. Encapsulated Modules3. Plug Compatibility4. Facilitated Module Reuse5. Module Redundancy/Diversity6. Elastic Capacity7 Di t ib t d C t l & I f ti7. Distributed Control & Information8. Deferred Commitment / Late binding9. Loosely Coupled Flat Interaction


y p10. Self Organization

Class 1 Agile Systems are Reconfigurable

Useful Metaphor: Plug and Play Drag and DropUseful Metaphor: Plug-and-Play – Drag-and-Drop

Class 2 Agile Systems are Reconfiguring

Useful Metaphor: Complex Adaptive Systems


A Framework withRequisite Variety, Parsimony, and Harmony

BANTAM BOOKS

The Three Lawsof Robotics

(Isaac Azimov)

1) A robot may not injure a human being or, through inaction, allow a human being to come to harm.

2) A robot must obey orders given it by human beings except where such orders would conflict with the First Law.

3) A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.


with the First or Second Law.

This is a generative framework, the basis of emergence

THE MANDELBROT SET THE MANDELBROT SET

( 1) ( )2INFINITE VARIETY FROM A SIMPLE EQUATION INFINITE VARIETY FROM A SIMPLE EQUATION

z(n+1) = z(n)2 + c


Complex Adaptive Systems Have Simple Generative FunctionsComplex Adaptive Systems Have Simple Generative Functions

Another Generative Framework“Companies seeking an “empowered” or decentralized p g pwork environment should first and foremost impose a tight ideology, screen and indoctrinate people into that ideology eject viruses and give those who remain theideology, eject viruses, and give those who remain the tremendous sense of responsibility that comes with membership in an elite organization. It means getting g g gthe right actors on the stage, putting them in the right frame of mind, and then giving them the freedom to ad lib as they see fit It means in short that cult likelib as they see fit. It means, in short, that cult-like tightness around an ideology actually enables a company to turn people loose to experiment, change, adapt, and – above all – to act.”

Built to Last, pg. 139, Collins and Porras


Fast, Cheap and Out of Control"Ghengis was a robot that could walk over anything in its path as it followed a person. Ghengis had six legs, bumper antennas, and infrared sensors for following the heatGhengis had six legs, bumper antennas, and infrared sensors for following the heat signature of the person it is following. The software for Ghenghis was not organized as a single program but fifty-one parallel programs Brooks called Augmented finite-state machines, which can send numbers to components on fixed wires. The first forty eight AFSMs allow Ghenghis to scramble around rough terrain. The walk machine uses six

t t th t th i l t t k t Th I f d S hi ioutputs that sequence the six legs to take steps. The Infrared Sensors machine receives input from six pyroelectronic sensors and each one has an on or off state that feeds into the Prowl or Steer machine. The Prowl machine is connected to inhibit the outputs of the Walk finite-state machine. If the robot detected some infrared activity, it walked toward it. If the sensors had been rotated to the back of the robot, it would walk away. ... The insect-likesensors had been rotated to the back of the robot, it would walk away. ... The insect like Ghengis was a turning point for robotics. Ghengis followed an emergent trajectory that was a product of both of its actions and its situation in the terrain of the world. (Amazon Reviewer: D. Nishimoto) [Rodney Brooks heads the CSAI Lab at MIT]

www.amazon.com/Flesh-Machines-Robots-Will-Change/dp/037572527X www.britannica.com/eb/article-219109/artificial-intelligence

© MIT, Artificial Intelligence Laboratory

14January 2007, [email protected], attributed copies permittedAttila, MIT (1989–91) Genghis

Amazingly brilliant, AFSM, Allen, Shakey, Genghis, Attila, Hannibal, Gog, April 6, 2006Reviewer: D. Nishimoto

www.amazon.com/Flesh-Machines-Robots-Will-Change/dp/037572527X

Ghengis was the name of a robot that could walk over anything in its path as it followed a person. Genghis had six legs, bumper antennas, and infrared sensors for following the heat of the person signature it is following. The software for Genghis was not organized as a single program but fifty-one parallel programs and Brooks g p g g g g g p g y p p gcalled these programs Augmented finite-state machines (AFSMs). AFSMs can send numbers to components on fixed wires. The first forty eight AFSMs allow Genghis to scramble around rough terrain. The walk machine uses six outputs that sequence the six legs to take steps. Genghis allow the robot too be out of balance as it walked and the beta balance machine corrected The Infrared Sensors machine receives input from six pyroelectronic sensors and each one has an on or off state that feed into the prowl or steer machine. These sensors are motion detectors and are tuned to the radiation band emitted by mammals. The prowl machine is connected to inhibit the outputs of the walk finite-state machine. If the robot detected some infrared activity, it walked toward it. If the sensors had been rotated to the back of the robot, it would walk away. Genghis had no sense of the directions backward, or forward, or away; it only had interaction embedded in it according to sensor input. Steer machine had left and right IR sensors and if the left sensor fired then the machine sent out a message to the left legs to take smaller steps, and if the right sensor fired then make the right legs take smaller steps. The insect-like Genghis was a turning point for robotics. The Genghis followed an p , g g g p g g p gemergent trajectory that was a product of both of its actions and its situation in the terrain of the world.

Brooks robots response to situation with conditional reactions and Cynthia Breazeal set out too write AFSMs in a higher-level language called the Behavior language. Colin Angle and Cynthia Breazeal built twin robots Attila and Hannibal each with 19 motors, 11 onboard computers, and hundreds of sensors. Eventually Breazeal produced over 1,500 AFSMs with her Behavior language code and through a model of pain through inconsistent sensor readings, they were able to ignore bad sensors and reintegrate them once they started to operate again. The legs of the robot were able to cooperate when the robots encountered rough terrain, lifting the body together, holding things up while a leg search for a difficult foot holding, and backing up and going around obstacles when needed. These robots were built from layered control systems without a central cognition box and coupled sensors to actuators.built from layered control systems without a central cognition box and coupled sensors to actuators.

The philosophers George Lakoff and Mark John argued that higher-level representation of language and thought are based on metaphors for our bodily interactions with the world. Metaphors develop from childhood from physical and social experiences, for example affection uses warmth because the child is exposed to the warmeth of the parents body. High level concepts are built on metaphors and rely on bodily experience in the world. Our language reflects these metaphors.

Metaphors make it worth exploring the building of a robot with a human form and seeing what metaphors can be derived from the experience. Robots are not people. However, people will know how to interact with robots in human form by making eye contact, nods, and other sublinguistic murmurs and other social clues. The robot will know when to talk and when to listen dependant on the social clues and Cog would pave the way in this research.The robot will know when to talk and when to listen dependant on the social clues and Cog would pave the way in this research.

One way to build a robot that can interact with people is a natural way is to build it with a vision system and with eyes that saccade and verge, and that look like human eyes. Each of Cog's eyes has two cameras. One has a wide angle lens so Cog can see peripheral view and the other has a narrow-angle lens to give Cog a fovea. Each of Cog's camera eyes are mounted on gimbals that can pan and tilt and its head and neck give it more freedom of motion of exploring. When Cog looks off in a direction, its head also turns in the direction. Cog vestibular-occular reflex allows its eye motions to successfully saccade and Cog is able to smooth flow someone walking in front of it. Cog's head has a gyroscope too play the role of an inner ear.

Cynthia Breazeal's robot Kismet paid attention to three sorts of things: moving things things with saturated colors and things with skin colors Kismet has internalCynthia Breazeal s robot Kismet paid attention to three sorts of things: moving things, things with saturated colors, and things with skin colors. Kismet has internal drives that get larger and larger unless they are satiated. As these drives get larger they release certain behaviors. If Kismet's bored drive gets large, it might start deliberately looking around, saccading from place to place, looking for something. The weighting on its attention system on saturated colors will direct the eyes while saccading to bright colors in the periphery view. The overall behaviors emerge from the interactions of the simpler behaviors.

Kismet has an auditory system and analyzes four pitch types known as prosody. Human infants recognize approval, prohibition, attention-getting, and soothing through prosodic patterns. Kismet has three emotional states: its valence, it arousal, and its stance. Valence is a measure of its happiness, and its arousal is how tired versus how stimulated Kismet is, and stance is how open it is too stimuli. It displays its emotional states with a set of eyebrows, its lips, and its ears and can put prosody in its voice


put prosody in its voice.

Ritchie says to Kismet, "I want to show you this watch my girlfriend gave me." Kismet dutifully looks at the watch. Kismet was picking up on the social clues and the directions of attention. When Ritchie brought the watch into Kismets center of view, a few inches below his face where Kismet was focused and when he brought his index finger up and tapped the watch the motion activated Kismets attention system and Kismet maintained eye contact with the watch. Eventually, Kismets attention system decided Ritchie's face was more interesting and looked back at the eyes of Ritchie. There is nothing qualitatively different from the mechanism in Genghis.

Detail for later reading

Paraphrasing a comment Brook's madein a personal communication...

I don't think humans can write predictable emergent programspredictable emergent programs.

The answer may lie in e a s e ay ehumans providing the initial seeds,

which are then improved by genetic algorithms.


Will They Attack Boston By Mistake?www.military-aerospace-technology.com/article.cfm?DocID=686

C t t t ll d F t ill b t


Current swarms are remote controlled. Future swarms will be autonomous.Patrolling the eastern seaboard. Evolving toward willfulness.

Scientific American, March 2000, pps 72-79

What can SoS learn from swarms?

Are they willful?

Does Nature care if they succeed?


Peopled Systems – SoS By Nature

Conflict (misaligned objectives)Conflict (misaligned objectives)Irrational behavior (by whose definition of normal?)

Willful agents in an SoSdecide every moment

what belonging means,g g ,and

will beinconsistent.

Ethics (how it ought to be ... by somebody's local rules)Insanity (abnormal behavior)Insanity (abnormal behavior)Selfishness: Prisoners dilemma (rat on your buddy and you go free)Willfulness (aggressive insistence, control freak)Drunk ill scared angry vengeful (temporarily out of sorts)


Drunk, ill, scared, angry, vengeful (temporarily out of sorts)Perception (interpretation of input filtered by accumulated knowledge)

Ni L f G dNine Laws of GodDistribute being.

Control from the bottom up.

Cultivate increasing returns.

Grow by chunking.

Maximize the fringes.

Honor your error.

Pursue no optima; have multiple goals.

Seek persistent disequilibrium.

Change changes itself.


Kevin Kelly, Wired Exec Editor

SoS Intelligence

Examines three kinds of group problems:- Cognition

C- Coordination- Cooperation

Groups work well under certain conditions...Groups work well under certain conditions... - Need rules to maintain order and coherence- Individuals think and act independently- Individuals talk-to/learn-from each other

B t t h i ti i b d- But, too much communication is bad

Conditions necessary:- Diversity

Doubleday, 2004

y- Independence- Decentralization - with aggregation

Lesson:


Lesson:- Even the author can't accept the truth

An Open Opt-In System-of-Systems

Cooperative EnterpriseSoS Concepts

- Such as VISA –

Wi dWisdom:• aggregation not centralization

• minimal rules of order/coherence

• non-interference (independence)

• self-selection (decentralization)

Lessons:

• explicit synergy (collective value)

• High growth to large group but then


Barrett-Koehler, 1999, www.bkconnection.com

High growth to large group, but then...• Backslid to conventional structure

Control-System Model of Human Goal Pursuit...Studies in Human and Organization Interaction

Amity and Enmity - I, Rudolf Starkermann, 2003Amity and Enmity I, Rudolf Starkermann, 2003

Will


2-Partner Unconscious Attitude LinkageAmity and Enmity - I, Rudolf Starkermann, 2003Amity and Enmity I, Rudolf Starkermann, 2003


Is Selfless Interest Super Natural?Amity and Enmity - II, Rudolf Starkermann, 2003Amity and Enmity II, Rudolf Starkermann, 2003


Amity and Enmity, Rudolf Starkermann, 2003Editions a la Carte, Zurich, Switzerland


Editions a la Carte, Zurich, Switzerlandwww.editions.ch, [email protected]

Aberrant behavior arising in a stable social systemis detected and opposed

Example: Female penguin attempting to steal a replacement egg for the one she lost is prevented from doing so by othersfor the one she lost is prevented from doing so by others.


Complex Adaptive Systems

Systemic Stabilization Emerges As Needed


The Biology of Business, John Henry Clippinger III, Editor, Jossey-Bass, 1999, p 9.

A SE t l i

Waterfall MethodActual DesignerGather data

Are SE tools in conflict with

reality?Problem

Solution

Analyze data

Formulate solution

Implement solution

"..the MCC Elevator Study is significant because, for the first time, we have a model of the process that people actually follow when they tackle hard problems. And it is not the orderly, linear process we have assumed is proper. "The non-linear pattern of activity that expert designers follow gives us fresh insight... It reveals that in normal problem-solving behavior, we may seem to wander about, making only halting progress towards the solution. "This non-linear process is not a defect not a sign of stupidity or lack of trainingThis non linear process is not a defect, not a sign of stupidity or lack of training, but rather the mark of a natural learning process. It suggests that humans are oriented more toward learning (a process that leaves us changed) than toward problem solving (a process focused on changing our surroundings).


Wicked Problems: Naming the Pain in Organizations, E. Jeffrey Conklin & William Weil,http://www.touchstone.com/tr/wp/wicked.html

SDOE 790HarmoniousHarmoniousArchitectureArchitecture

Systems-Art Appreciation

Embracement vs Enforcement

Complex Adaptive Systems

Self Organizing SystemsSelf Organizing Systems

Systems of Systems

S S tSwarm Systems

Willful Systems

Open Systems

Resilience


...and more

[email protected], Stevens Institute of Technology, 505-586-1536

Panel Questions


1st of 3 Panel Questions

1) What is the difference between Complex Systems, System of Systems, andWicked Systems? Define characteristics of each.

CS WS SoSBehavior

Fixed Changing ChangingFixed Changing ChangingCause Generating Generating Generating

Function Function Function

Effect Bounded Unbounded Bounded

Intriguing thought at this point – needs some work, and maybe a paper.


2nd of 3 Panel Questions

2) What are some methods used to address Complex Systems2) What are some methods used to address Complex Systems, provide examples? Can any standard Systems Engineering methods be applied to developing these systems?

a) - Design generating functions.a) Design generating functions.- Design behavior attractors.- Design behavior boundary functions.- Design emergent stabilizationDesign emergent stabilization.- Design relationship-evolution managers rather than interfaces. - Employ principles we must now go discover (for the above design activity).- Employ autonomous agent and swarm conceptsEmploy autonomous agent and swarm concepts

... AFSMs (Subsumption architecture, Rodney Brooks).- Employ Agile Systems principles and concepts.- Get loose – you can't control anything – think influence.y y g

b) - Beyond systems thinking and the ConOps ... nothing comes to mind.- At this point in time we apply traditional SE methods to traditional systems,and explore new SE methods for SoS/CS/WS systems In time when these


and explore new SE methods for SoS/CS/WS systems. In time, when thesenew methods are mastered, they will replace the traditional SE methods forso-called traditional systems as well.

3rd of 3 Panel Questions

3) What roles do organizations like the INCOSE and National Laboratories have in helping the world/SE community with these types of complex orwicked problems that exist?

a) Doing what they do ... their role has no choice but to play out. These kinds of systems are the systems being increasingly dealt with.

b) Discover employable fundamental principles. Natural laws are at work. Science learns natural laws through observation hypothesis and testScience learns natural laws through observation, hypothesis, and test.


Rick DoveIndustry Professor, Stevens Inst. of Tech.y ,

- Agile Systems & Enterprisesgraduate certificate co-director

35+ years of start-ups and turnarounds35 years of start ups and turnaroundsand interim executive management

Carnegie Mellon: BSEE UC Berkeley: graduate work in CompSciUC Berkeley: graduate work in CompSci

Early Life: Systems Programmer/Analyst

Co-founder of Agile Enterprise conceptin '91 at Lehigh University

Director of Strategy/Research, Sr. Fellow, Agility Forum, Lehigh University

Author: "Response Ability: The Language, Structure, and Culture Of Agile Enterprise"


Photo: Brian ZickLives in Taos, New Mexico, at 8200 feet ......Land of Enchantment (and thin air).

Documents

Wicked Complex Systems of Systems