an introduction to system dynamics · an introduction to system dynamics george p. richardson...

1

Rockefeller College of Public Affairs and Policy University at Albany

An Introduction to System Dynamics

George P. Richardson Rockefeller College of Public Affairs and Policy

University at Albany, State University of New York

2

Rockefeller College of Public Affairs and Policy University at Albany

Events and Decisions

Patterns of Behavior

System Structure

Reactive

Adaptive

Generative

A Systems Perspective

3

Rockefeller College of Public Affairs and Policy University at Albany

A systems view stands back just far enough to... • Deliberately blur discrete events into patterns of

behavior • Deliberately move from a focus on individual

decisions to a focus on policy structure

“Distancing...”

4

Rockefeller College of Public Affairs and Policy University at Albany

Four Key Patterns of Thought

• Dynamic thinking (graphs over time) • Causal thinking (feedback loops) • Stock-and-flow thinking (accumulations) • Thinking endogenously (system as cause)

5

Rockefeller College of Public Affairs and Policy University at Albany

New York, Chicago & Philadelphia, 1800-2000

6

Rockefeller College of Public Affairs and Policy University at Albany

Prejudice and Minority Achievement (Myrdal, Merton)

Prejudice

Discrimination

Opportunities for the minority

Achievements of the minority

– +

– +

(R) Prejudice

Aspirations of the minority

Minority efforts to achieve

Minority perceptions of the gap

+

+

–

+

(B) Striving

+ (R)

Hope or despair

7

Rockefeller College of Public Affairs and Policy University at Albany

Stocks, Flows, and Feedback Loops Here: a gasoline shortage crisis

8

Rockefeller College of Public Affairs and Policy University at Albany

Dynamics

• Define problems in terms of graphs over time. • Graph important variables • Graph historical data • Graph anticipated dynamics • Graph preferred dynamics

• Use these to focus systems thinking and modeling

9

Rockefeller College of Public Affairs and Policy University at Albany

U.S. Unemployment 1948-2012 (annual average %)

0.0

2.0

4.0

6.0

8.0

10.0

12.0

1948 1958 1968 1978 1988 1998 2008

10

Rockefeller College of Public Affairs and Policy University at Albany

Carbon Emissions 1800-2000

11

Rockefeller College of Public Affairs and Policy University at Albany

• Accumulations (populations, resources…) • Causal structure: “feedback” loops • Delays • Perceptions (a kind of accumulation) • Pressures • Affects, emotions, (ir)rationalities • Policies governing decisions

Systems Structure

12

Rockefeller College of Public Affairs and Policy University at Albany

Causal Diagrams

• Causal mapping is a powerful tool for representing structure in complex systems.

• Arrows indicate causal influence.

13

Rockefeller College of Public Affairs and Policy University at Albany

Polarities of Causal Links

• Positive and negative signs show the direction of causality:

– +

+ –

+ –

+ ... “direct” relation – …“inverse” relation

14

Rockefeller College of Public Affairs and Policy University at Albany

Feedback Loops

A feedback loop exists when decisions change the state of the system, changing the conditions and information that influence future decisions.

15

Rockefeller College of Public Affairs and Policy University at Albany

Two kinds of feedback loops

• Reinforcing loops • Growth producing • Destabilizing • Accelerating • Positive: an even number of –’s

• Symbolized by

• Balancing loops • Counteracting • Goal seeking • Stabilizing • Negative: an odd number of –’s

• Symbolized by

16

Rockefeller College of Public Affairs and Policy University at Albany

Examples of Reinforcing Loops

+

+

+ +

+

+ +

– –

17

Rockefeller College of Public Affairs and Policy University at Albany

Typical Reinforcing Loop Behaviors 20,000

15,000

10,000

5,000

0 0 25 50 75 100

10,000

9,000

8,000

7,000

6,000 0 25 50 75 100

Population and Births Loop

Businesses and Taxes Loop

18

Rockefeller College of Public Affairs and Policy University at Albany

Examples of Balancing or Counteracting Loops

+

–

+

+ –

-

+

+

–

+

19

Rockefeller College of Public Affairs and Policy University at Albany

Typical Counteracting Loop Behaviors 20

15

10

5

0 0 10 20 30 40

Filling a Glass

10,000

7,500

5,000

2,500

0 0 25 50 75 100

Population and emigration

•0 •7.5 •15 •22.5 •30

Predator-prey interactions

20

Rockefeller College of Public Affairs and Policy University at Albany

But There are Subtleties: Not all Word-and-Arrow Diagrams are Alike!

21

Rockefeller College of Public Affairs and Policy University at Albany

These arrows mean ‘and then’

• We start with some understandings of the problem and its systemic context, and then we conceptualize (map) the system.

• Then we build the beginnings of a model,

which we then test to understand it. • Then we reformulate, or reconceptualize,

or revise our understandings, or do some of all three, and then continue…

22

Rockefeller College of Public Affairs and Policy University at Albany

Arrows here are flows of material The words here represent accumulations of carbon –stocks -- and the arrows represent flows. This is not a causal diagram. This is a view of the “carbon cycle.”

23

Rockefeller College of Public Affairs and Policy University at Albany

Only this one is a causal loop

This causal loop tells a very compelling and important self-reinforcing story.

24

Rockefeller College of Public Affairs and Policy University at Albany

Stocks and Flows

Stocks are accumulations. • Stocks are increased by inflows and decreased by outflows. • When a link means “add” or “subtract” we have a stock-and-

flow structure. • Example: Inventory

25

Rockefeller College of Public Affairs and Policy University at Albany

Human Activity, CO2, and Global Temperature

Thought experiment:

26

Rockefeller College of Public Affairs and Policy University at Albany

The system dynamics modeling process

Adapted from Saeed 1992

27

Rockefeller College of Public Affairs and Policy University at Albany

Processes focusing on system structure

28

Rockefeller College of Public Affairs and Policy University at Albany

Processes focusing on system behavior

29

Rockefeller College of Public Affairs and Policy University at Albany

Two kinds of validating processes

30

Rockefeller College of Public Affairs and Policy University at Albany

Pictures Can Get Really Complicated!

31

Rockefeller College of Public Affairs and Policy University at Albany

The Endogenous Point of View The “X/N” Matrix

Striving for understanding and

leverage, but failing

Achieving understanding and

leverage

Accepting fate, Predicting, Preparing

Confused, Misguided, Misguiding

Exogenous Endogenous True (Predominant) State of Affairs

Exog

enou

s En

doge

nous

Pred

omin

ant M

ode

of A

naly

sis

32

Rockefeller College of Public Affairs and Policy University at Albany

A Lightening-Fast Example

• Security on off-shore oil platforms during a technology transition

• Mixed consulting / theory building intervention • Two group model building workshops, May and

September with various high-level management people from Norsk Hydro and related professional groups

33

Rockefeller College of Public Affairs and Policy University at Albany

Hopes and fears

Find a common case for all

(AUC,SINTEF, HYDRO)

HYDRO is new to group modelling

We do not catch Hydro’s interest

Hope Hydro becomes very

involved

focus on security

to much focus on safety

Establish a Platform for

Communicating SD for Hydro.

Clear Picture of Workshops down

the road

Too little time to be successful

How to simulate Risk/Stress

Establishing a SD Model that gives no

new insights

Get a firm understanding

system dynamic mapping

Centrifugal Forces blow AMBASEC, IRMA, HYDRO

apart

Scenario discussions

Getting lost in detail

This SD-Modell does not improve security (& safety)

in e-Operation

Consensus about 3 yr agenda

come up with a useful case/ model

Establish a SD Model giving new

insight

Will get a few really dynamically interesting cases

come up with some artificial case

Identify valuable insights for all

Understanding of integrated operation

More safe & secure e-operations

Hopes Fears

Process will help Hydro understand what may happen

Someone on-shore will inadvertently

intervene off-shore

Norsk Hydro may not be the right

case

We don’t understand the

situations

Some managers think we should never open the

platforms

34

Rockefeller College of Public Affairs and Policy University at Albany

Stakeholders

INTEREST

INFL

UE

NC

E

CRO Control Room

Operator

Management Crises Team

Interval Intruder Attacker

CIO

Control Room Manager

Organized Crime

PLATFORM CHIEF

Chief Executive Officer

Government Incident

Response Team Member

Incident Response

Team Manager

Terrorists

Operator

Supplier (maintenance of

PCs/ESD)

Service Supplier (handling within

their organization)

MALICIOUS OUTSIDER

SYSTEM ADMIN

CISO (Chief InfoSeq.

Officer)

PTIL

Local Communities/

Generel, Public

SOFTWARE PROVIDERS

Developers of COTS

CSIRT in a virtual

organisation

HSE responsible

STOCK OWNERS

DEVELOPERS OF ATTACK

TOOLS

COMPETING OIL COMPANY

MEDIA OLF

IT EXPERT (Support)

OIL COMPANY (System Owner)

Worker (uses IT systems)

The one detecting the

incident

ECOLOGICAL ACTIVISTS

NSM

Managers of unsecured

external networks

National Politicians

Work planner / administrator

NPD PSA DSB

Local Politicians

Contractors doing drilling &

modification fixed platforms

Prioritization by group members

Added day two

Nature

People responsible for e-

Op Change processes

35

Rockefeller College of Public Affairs and Policy University at Albany

Behaviour over time

Overview

36

Rockefeller College of Public Affairs and Policy University at Albany

Policies

Higher level of security

Improve the safety and

security culture

share and learn from incident B/W Orgs. & W/n Orgs

Annual awareness campaigns measures

security culture

Increase the amount of

incident reports from IRT & CEO

Be open about security incidents

Establish common risk perception

improve incident reporting

Establish CSRS

Increase knowledge information

sharing across industry

Cooperation between IRTs in

different organisations

Prevent out of sight out of mind mentality

Build IDS to systematically

gather information

Create formal CSIRTS

Increase CSIRTS authority

a warning system for the communication

network

Training to close knowledge gap

Perform Pilots do intrusion tests

and spread success stories

Identify ‘Best practice’ other countries other industries

Invest in survivability Solutions

Perform HazOp of the e-

operation solution of 2010

Invest in Vulnerability

Detection

Balance work & RD of CSIRT

Risk assessment in change processes

(continuously)

Common security

requirement on products and

services

Understand the erosion of

compliance

Establish common system incident

report management

-common report in –publish incidents

in industry

Establish Best /Good Practice Against Insider

Threat

Establish a security Quality

process improvement

Monitor/Measure Risk Change

(Auditing)

Create collaborate

arena & culture for information

sharing on-shore / off-shore

Prioritization by group members

Added day two

Continuous training

37

Rockefeller College of Public Affairs and Policy University at Albany

A Tiny Model Capturing the Problem Dynamics

38

Rockefeller College of Public Affairs and Policy University at Albany

A Tiny Model Capturing the Problem Dynamics

39

Rockefeller College of Public Affairs and Policy University at Albany

A Tiny Model Capturing the Problem Dynamics

40

Rockefeller College of Public Affairs and Policy University at Albany

• Traditional capacity (blue) phasing out • New capacity (red) rising, peaking and declining • Mature capacity (green) slowly rising • Cost per barrel (black) declining to a new low • Risk (grey) rising, peaking, and declining • …all just what the problem description called for

Behavior of this Tiny Beginning Model

But vastly oversimplified. Serious group modeling was needed.

41

Rockefeller College of Public Affairs and Policy University at Albany

Group Modeling Work in Process

42

Rockefeller College of Public Affairs and Policy University at Albany

After Much Client and Team Work – “Hydro1”

43

Rockefeller College of Public Affairs and Policy University at Albany

Policy Simulations with Hydro1

44

Rockefeller College of Public Affairs and Policy University at Albany

When It Works, Why?

• Engagement • Mental models • Complexity • Alignment • Refutability • Empowerment