A System Dynamics (SD) Approach to
Modeling and Understanding Terrorist Networks
BAA-07-01-IFKA Proactive Intelligence (PAINT):
Model Development
Massachusetts Institute of Technology (MIT)
–
–
–
Sloan School of Management
Political Science Department
Engineering Systems Division
and National Security Innovations, Inc. (NSI)
V11 2007-02-22
MIT
1
Agenda
• Team
• What is System Dynamics (SD) Modeling
• Why is SD Modeling important
• Challenge Problem to be addressed
• Example of SD Modeling
• Collaboration with other PAINT areas
• Metrics & Validation
• Management of Model Complexity
–
Key Sub-systems
• Tasks, Deliverables & Timetable
• Conclusion
MIT
2
Key Personnel
Massachusetts Institute of Technology (MIT)
–
Stuart Madnick, Sloan School of Management, Information
Technologies & School of Engineering, Engineering Systems Division
–
Nazli Choucri, School of Humanities and Social Sciences, Political
Science Department
–
Michael Siegel, Sloan School of Management, Information
Technologies
National Security Innovations, Inc. (NSI)
–
–
Robert Popp, Founder and Chairman
Greg Ingram, Vice President for Operational Technology
All Key Personnel have considerable experience with the organization
and management of large-scale projects that combine modeling and
diverse data with application requirements in related areas – such
as DARPA’s Pre-Conflict Analysis and Shaping (PCAS) effort
3
MIT
Philosophy of System Dynamics
• Every action has consequences
• Often through complex non-linear feedback loops
• Human are good at understanding individual pieces,
• but difficult at comprehending the full impact
Do you feel crowded in – and frustrated?
MIT
4
See if you can get a bit more space by pushing on that wall …
MIT
5
Oops …
MIT
6
History of System Dynamics Modeling (SDM)
SDM used as modeling & simulation method over 30 years
•
•
Eliminate limitations of linear logics and over-simplicity
 Typical human assumptions and behaviors
Better understanding system structure, behavior patterns,
interconnections of positive & negative feedback loops, and
intended & unintended consequences of action
SDM has been applied to numerous domains, e.g.,
•
•
•
•
•
Software development projects
Process Improvement projects
Crisis and threat in the world oil market
Stability and instability of countries
… many many others …
SDM helps to uncover ‘hidden’ dynamics in system
•
•
•
MIT
Helps understand ‘unfolding’ of situations
Helps anticipate & predict new modes
Explore range of unintended consequences
7
Appropriateness of Modeling Methodologies
(adapted from Axelrod, 2004: “Modeling Security Issues of Central Asia”)
Game Theory
Agent-Based
Modeling with
Light Agents
Agent Based
Modeling with
Heavy Agents
Dynamical
Systems
System
Dynamics
Data-driven
Models
•
•
Construction
time
User
Pre-requisites
Learning
Time
Flexibility
Repertory
size
Transparency
L
M
L-H
L
L-H
L
M
M
H
M
L-H
H
H
M
M
H
L
M
M
H
H
M
M
M
L
L
L
H
L
H
M*
M
L
M*
M
H
Ideally, first three criteria should be Low, and the last three criteria should be High.
The Criteria
–
–
–
–
–
MIT
–
Construction Time. Time and effort needed for a modeler skilled in this methodology to build a useful model with
input from users.
User Prerequisites. Amount of technical background needed by the user to understand as well as use the model.
Learning Time. Time and effort for a typical user with the necessary prerequisites needs to learn a specific model.
Flexibility. Ease with which the modeler can modify the model to incorporate a new variable.
Repertory size. The number of published models of this type with features that could be adapted for use as part of a
model on issues relevant to security in central Asia.
Transparency. The ease with which the user can discover anything in the model that might bias the results.
8
Unique Capabilities of System Dynamics Modeling
•
•
•
•
•
•
•
•
•
MIT
Objective input: Utilize data to determine parameters affecting the
causality of individual cause-and-effect relationships.
Subjective (expert) judgment: Represent and model cause-and-effect
relationships, based on expert judgment.
Intentions Analysis: Identify the long-term unintended
consequences of policy choices or actions taken in the short term
Tipping point analysis: Identify and analyze “tipping points” – where
incremental changes lead to significant impacts.
Transparency: Explain the reasoning behind predictions and outputs
of the SD model.
Modularity: Can organize SD models into collections of
communicating sub-models (e.g., terrorism recruitment, economic
development, religious intensity, regime stability)
Scalability: Use the modularity to increase complexity without
becoming unmanageable.
Portability: Utilize the same basic SD model in different regions of
the world without requiring re-formulation.
Focusability: Increase details in specific areas of the SD model to
address specific (and possibly new) issues.
9
PAINT Challenge Problem
How should the Government analyze
terrorist networks in the context of
the political, religious, social and
economic networks that intersect
with, influence, and are influenced
by, the terrorist network; predict the
formation, evolution, vulnerabilities,
and dissolution of the network; and
identify strategies to shape or
influence the network through
selective action?
MIT
10
Example of System Dynamics Modeling:
Dissident and Terrorist Network Escalation
(very simplified)
Factors that affect
rate of Flow
Flows
Avg Time as
Dissident
Stocks
Appeasement
Rate
Population
Births
Dissidents
Becoming
Dissident
Recruits Through
Social Network
MIT
Appeasement
Fraction
Terrorist
Recruitment
Terrorists
Desired Time to
Remove
Terrorists
Removing
Terrorists
Removed
Terrorists
Regime
Opponents
11
Dissident and Terrorist Network Development
(slightly more detailed)
Avg Time as
Dissident
Appeasement
Rate
Population
Births
Dissidents
Insurgent
Recruitment
Becoming
Dissident
Recruits Through
Social Network
Regime
Opponents
Normal Propensity
to be Recruited
Propensity to be
Recruited
MIT
Removing
Terrorists
Violent Incident
Intensity
Protest
Intensity
Removed
Terrorists
Propensity to
Commit Violence
Relative Strength
of Violent Incidents
Incident
Intensity
Effect of Incidents on
Anti-Regime
Messages
Regime
Resilience
Social
Capacity
Political
Regime
Capacity
Legitimacy
Terrorists
Propensity to
Protest
Effect of Regime
Resilience on
Recruitment
Economic
Performance
Desired Time to
Remove
Terrorists
Appeasement
Fraction
Effect of Anti-Regime
Messages on
Recruitment
Perceived Intensity
of Anti-Regime
Messages
Message
Effect
Strength
Fifth-order system of non-linear differential equations:
> 140 equations & > 100 feedback loops
12
Sample of Structure to Equations:
Recruitment Section
Population
Becoming
Dissident
Pop Growth
Fr Gr Rate
Total
Contacts
Incitability
Dissidents
Terrorist
Recruitment
Terrorists
Regime
Recruits Through
Opponents
Social Network
Cond Prob of
Fract of Contacts
Recruit
with Regime
Opponents
Contacts Between
Opponents and
Population
<Total Pop>
Stocks
Parameters
Variables
P = INTG(PG-BD)dt+Po
D = INTG(TR-BD)dt+Do
T = INTG(TR)dt+To
FGR = 0.001706
I = 0.4
CPR = 0.1
RO = D+T
TP = P+D+T
FCWRO = RO/TP
TC = P*I
CBOP = TC*FCWRO
RTSN = CBOP*CPR
Flows
MIT
PG = P*FGR
BD = RTSN
13
Example Intervention Policies: Removing Terrorists
vs. Preventing Recruitment
Increased
Removal
Effectiveness
Avg Time as
Dissident
Appeasement
Rate
Population
Births
Dissidents
Insurgent
Recruitment
Becoming
Dissident
Recruits Through
Social Network
Regime
Opponents
Normal Propensity
to be Recruited
Propensity to be
Recruited
Social
Capacity
MIT
Political
Regime
Legitimacy Capacity
Terrorists
Removed
Terrorists
Removing
Terrorists
Propensity to
Commit Violence
Violent Incident
Intensity
Protest
Intensity
Relative Strength
of Violent Incidents
Incident
Intensity
Propensity to
Protest
Effect of Incidents on
Anti-Regime
Messages
Effect of Regime
Resilience on
Recruitment
Regime
Resilience
Desired Time to
Remove
Terrorists
Appeasement
Fraction
Effect of Anti-Regime
Messages on
Recruitment
Perceived Intensity
of Anti-Regime
Messages
Message
Effect
Strength
Preventing
Recruitment
14
Example Intervention Policies: Removing Terrorists
vs. Preventing Recruitment
Terrorists
27,000
25,250
23,500
Better terrorist removal
21,750
20,000
2005
Preventing recruitment
2006
2007
2008
2009
2010
Time (Year)
MIT
Better terrorist removal
Preventing recruitment
intelligence sharing
moderate rhetoric
Removing terrorists has a limited effect
Preventing recruitment effects a sustained reduction
15
Collaboration with other PAINT areas:
Architecture and Integration, Key Indicators,
& Dynamic Gaming and Strategies
•
•
•
Worked with other potential PAINT researchers, such as in PCAS.
Expertise that we can contribute to the overall PAINT effort.
Architecture and Integration
Innovative IT Architectures for Integration are major research foci for our
MIT group at MIT.
•
•
“Context Interchange: Using Knowledge about Data to Integrate Disparate
Sources,” was projects under DARPA’s Intelligent Integration of Information (I3)
research program - further improved and tested in various environments,
including a recent project to facilitate the integration of intelligence data.
Key Indicators
Key Indicators are important part of our proposed work on the PAINT
effort. We have experience with identifying and understanding Key
indicators in other projects.
Dynamic Gaming and Strategies
System Dynamics extensively used by MIT in dynamic gaming, called
“management flight simulators” to demonstrate how managerial
“instincts” often lead to counter-intuitive and erroneous results.
MIT
16
‘What if’ “Virtual / Gaming mode”
- Parameter Inputs with Sliders
MIT
17
•
Metrics & Validation
Many ways to validate a System Dynamics model
–
–
•
Behavioral Reproduction
–
–
–
–
•
MIT
12 ways on p. 6 of proposal
we will use all of them; two are below:
Use past data (as well as other sources) to help determine
parameters up to, say, two years ago.
Each “stock” (e.g., number of terrorists) is a metric.
Measure how well SD model projections match the following years
 planned changes, known 2 years ago, to policy are included.
In PCAS effort, our SD model predictions were very accurate.
System Improvement
–
Does the model generate useful insights that are appreciated by
decision makers?
–
In PCAS effort, our results were presented to PACOM, etc.
18
Managing Model Complexity
•
•
“A model should be as simple as possible and only as
complex as needed.” Unneeded complexity will be avoided
in this project.
The primary method to manage SD model complexity is the
use of subsystems (which can be further decomposed into sub-subsystems,
if needed.)
–
•
(a) regime resilience
(b) terrorist network activities and growth.
(c) government capacity & interactions with terrorist networks
Each of these subsystems have internal dynamics as well as
dynamic interactions with the other subsystems.
–
–
MIT
Our current plan is divide our High Level Model (HLM) into at least
three major subsystems:
Multi-level layer approach simplifies the complexity both in model
development and refinements as well as model usage and
understanding.
Used very effectively in many SD modeling projects.
19
Proposed Tasks & Timetable
(timetable on p. 16, details of 36 tasks on pp. 23-26 of proposal)
Working Integrated SD model delivered each year and improved
each year.
Phase 1 (18 months) – Component Predictive Models Integrated into a Virtual
World/Dynamic Gaming Collaborative
Key task is to design, develop, and complete the High Level Model (HLM) including all subsystems: (a) regime resilience, (b) terrorist network activities and growth, and (c)
government capacity and interactions with terrorists.
Basic data for the HLM compiled to provide an empirical view of the overall model.
Phase 2 (12 months) – Prediction Using Specific Challenge Problem with Historical
or Synthetic Data
All subsystems enhanced; focus on improving the regime resilience sub-system.
Phase 3 (12 months) – Prediction using Real World Data Instrumentation,
Feedback and Fine tuning
All subsystems enhanced; focus on the terrorist network activities and growth sub-system
Phase 4 (12 months) – Grand Challenge Problem: Influence Strategies for
Alternative Futures
All subsystems enhanced; focus on the government capacity sub-system and interactions with
terrorists; development and analysis of strategies leading to better improved alternative
futures.
MIT
20
Conclusions
• System Dynamics methodology important
•
•
•
•
MIT
and critical method for addressing the broad
scope of PAINT.
SD has been shown effective is related
efforts (e.g., PCAS).
We have assembled superb multidisciplinary team
We are committed to the success of PAINT.
Thank you.
21
Backup Slides – For Q&A
MIT
22
Quick Primer: What (and Why) of System Dynamics
Consider the domain of Software Development
1. “Knee jerk” reaction to a project behind schedule is to add people.
2. “Brooks Law” noted that “Adding people to a late project, just makes
it later”
•
Because the new people must be trained, this takes productive
people off the project – which was not obvious before.
• These points are now fairly well-known by most software developers
– but still naïve.
•
•
Many other factors: length of project, type of project, expertise of
staff available, approach to and time needed to do training, stage of
project, etc.
Over the years, all of these individual factors have been well-studied
individually – but how do they interact ?
• System dynamics helps model & study the dynamics of the
interdependencies. Non-obvious outcomes frequently found.
(e.g., sometimes Brooks is wrong! When and Why?)
Source: Software Project Dynamics: An Integrated Approach, by T.K. Abdel-Hamid and S. Madnick, Prentice-Hall, 1991,
and Fred Brooks, The Mythical Man-Month, 1975.
MIT
23
Validation of System Dynamics Models
•
•
•
•
•
•
•
•
•
•
•
•
MIT
Boundary Adequacy: Does the selection of what is endogenous, exogenous,
and excluded make sense?
Structure Assessment: Is the level of aggregation correct, and does the
structure conform to reality?
Dimensional Consistency: Do the units of the model make sense, and does the
model avoid the use of arbitrary scaling factors?
Parameter Assessment: Do the parameters have real life meanings, and are
their values properly estimated?
Extreme Conditions: Do extreme parameter values lead to irrational behavior?
Integration Error: Does the behavior change when the integration method or
time step are changed?
Behavioral Reproduction: How well does the model behavior match the
historical behavior of the real system?
Behavior Anomaly: Does changing the loop structure lead to anomalous
behavior consistent with the changes?
Family Member: How well does the model “scale” to other members within the
same class of systems?
Surprise Behavior: What is revealed when model behavior does not match
expectations?
Sensitivity Analysis: Do conclusions change in important ways when
assumptions are varied over their plausible range? Changes in conclusions
could be numerical changes, behavior mode changes, or policy changes.
System Improvement: Does the model generate insights that actually lead to
the hoped for improvements?
24
‘What if’ “Virtual / Gaming mode”
- Parameter Inputs with Sliders
MIT
25
Example End-User (Non-Technical) Interface Design
MIT
26
Resumes of Key Personnel - MIT
Dr. Stuart Madnick is the John Norris Maguire Professor of Information Technology, Sloan School of
Management and Professor of Engineering Systems, School of Engineering at the Massachusetts
Institute of Technology. He has been a faculty member at MIT since 1972. He has served as the head
of MIT's Information Technologies Group for more than twenty years. He has also been a member of
MIT's Laboratory for Computer Science, International Financial Services Research Center, and
Center for Information Systems Research. Dr. Madnick is the author or co-author of over 250 books,
articles, or reports including the classic textbook, Operating Systems, and the book, The Dynamics of
Software Development, which received the Jay Wright Forrester Award for "Best Contribution to the
field of System Dynamics in the preceding five years" awarded by the System Dynamics Society. His
current research interests include connectivity among disparate distributed information systems,
database technology, software project management, and the strategic use of information technology.
He is presently co-Director of the PROductivity From Information Technology Initiative and co-Heads
the Total Data Quality Management research program. He has been active in industry, as a key
designer and developer of projects such as IBM's VM/370 operating system and Lockheed's DIALOG
information retrieval system. He has served as a consultant to corporations, such as IBM, AT&T, and
Citicorp. He has also been the founder or co-founder of high-tech firms, including Intercomp, Mitrol,
and Cambridge Institute for Information Systems, iAggregate.com and currently operates a hotel in
the 14th century Langley Castle in England. Dr. Madnick has degrees in Electrical Engineering (B.S.
and M.S.), Management (M.S.), and Computer Science (Ph.D.) from MIT. He has been a Visiting
Professor at Harvard University, Nanyang Technological University (Singapore), University of
Newcastle (England), Technion (Israel), and Victoria University (New Zealand).
MIT
27
Resumes of Key Personnel (continued) - MIT
Dr. Nazli Choucri is Professor of Political Science at the Massachusetts Institute of Technology,
and Director of the Global System for Sustainable Development (GSSD), a distributed multilingual knowledge networking system to facilitate uses of knowledge for the management of
dynamic strategic challenges. To date, GSSD is mirrored (i.e. synchronized and replicated) in
China, Europe, and the Middle East in Chinese, Arabic, French and English. As a member of
the MIT faculty for over thirty years, Professor Choucri’s area of expertise is on modalities of
conflict and violence in international relations. She served as General Editor of the
International Political Science Review and is the founding Editor of the MIT Press Series on
Global Environmental Accord. The author of nine books and over 120 articles Professor
Choucri’s core research is on conflict and collaboration in international relations. Her present
research focus is on ‘connectivity for sustainability’, including e-learning, e-commerce, and edevelopment strategies. Dr. Choucri is Associate Director of MIT’s Technology and
Development Program, and Head of the Middle East Program. She has been involved in
research, consulting, or advisory work for national and international agencies, and in many
countries, including: Abu Dhabi, Algeria, Canada, Colombia, Egypt, France, Germany, Greece,
Honduras, Japan, Kuwait, Mexico, North Yemen, Pakistan, Qatar, Sudan, Switzerland, Syria,
Tunisia, Turkey
Dr. Michael Siegel is a Principal Research Scientist at the MIT Sloan School of Management.
He is currently the Director of the Financial Services Special Interest Group at the MIT Center
For eBusiness. Dr. Siegel’s research interests include the use of information technology in
financial risk management and global financial systems, eBusiness and financial services,
global ebusiness opportunities, financial account aggregation, ROI analysis for online financial
applications, heterogeneous database systems, managing data semantics, query optimization,
intelligent database systems, and learning in database systems. He has taught a range of
courses including Database Systems and Information Technology for Financial Services. He
currently leads a research team looking at issues in strategy, technology and application for
eBusiness in Financial Services.
MIT
28
Resumes of Key Personnel (continued) – NSI
Dr. Robert Popp is cofounder of National Security Innovations (NSI), Inc., presently serving as its
Chairman and CEO. Prior to NSI, Dr. Popp served as Executive Vice President of Aptima, Inc. Prior
to Aptima, Dr. Popp served for five years as a senior government executive within the Defense
Department: one year at the Office of the Secretary of Defense as Assistant Deputy Undersecretary
of Defense for Advanced Systems and Concepts, and four years at the Defense Advanced
Research Projects Agency (DARPA). At DARPA, Dr. Popp served as Deputy of the Information
Awareness Office (IAO) where he oversaw a portfolio of over 25 programs exceeding $170M
focused on novel IT-based tools for counter-terrorism, foreign intelligence and national security. Dr.
Popp was also Deputy PM to Dr. Poindexter on the Total Information Awareness (TIA) program, a
program that integrated and experimented with analytical tools in text processing, collaboration,
decision support, foreign languages, predictive modeling, pattern analysis, and privacy. Dr. Popp
also served as Deputy of the Information Exploitation Office (IXO), where he established a novel
research thrust in stability operations and quantitative/computational social science modeling for
nation state instability and conflict analysis. Prior to government service, Dr. Popp held senior
positions with ALPHATECH, Inc. (now BAE Systems) and BBN. He has served on the Defense
Science Board (DSB), is a Senior Associate for the Center for Strategic and International Studies
(CSIS), and is a founding Fellow of the Academy of Distinguished Engineers at the University of
Connecticut. Dr. Popp also served in the military from 1982 – 1988 as a Staff Sergeant in the US Air
Force as an Aircraft Maintenance Technician of F106 fighters and B52 bombers. Dr. Popp holds a
Ph.D in Electrical Engineering from the University of Connecticut, and a BA/MA in Computer
Science (summa cum laude, Phi Beta Kappa) from Boston University.
Gregory J. Ingram is the Vice President for Operational Technology for National Security Innovations
MIT
(NSI), Inc. He has twenty-four years of experience in the Army in the fields of Special Forces,
Infantry, Civil Affairs, and Psychological Operations (PSYOP). Fifteen of his twenty-four years have
been on active duty and the remainder in the reserves. He has deployed in various capacities to
Lebanon, Italy, Chile, Korea, Haiti, Afghanistan, and Iraq. For the last five years, Greg has been
heavily involved in developing, integrating, and operationalizing leading-edge technologies in the
areas of knowledge discovery, planning and analysis, human language technologies, and
quantitative social science methodologies. Greg served as the lead PSYOP/IO Planner in the
Special Operations Joint Interagency Collaboration Center (SOJICC) and as an Operational
Manager for the development of the PSYOP Planning and Analysis System (POPAS) as part of the
PSYOP Global Reach (PGR) Advanced Concept Technology Demonstration (ACTD) at the United 29
States Special Operations Command (USSOCOM).
Descargar

Very high-level overview (motivation from our proposal