Unending inductive/deductive methods in
Complex Social systems
Complexity and method in the Social Sciences,
University of Warwick,
Prof Peter M. Allen
Complex Systems Research Centre,
Cranfield University
Complexity and Understanding
• In the natural sciences, for some problems we can do
repeatable experiments repeatedly - prediction. Can find robust
‘laws’. Induction. Can deduce specific behaviour - Deduction
• In social/biological systems repeatable experiments are difficult
because of memory, changing environment, learning entities. In
everyday life we learn pragmatically rules that work well
enough not to kill us immediately.
• Is ‘understanding’ about ‘describing what is going on and what
will happen’? Is explanation about prediction? Or can knowledge
(dice) ‘explain’ lack of prediction? Do system dynamics explain
or describe? Or show which assumptions hold?
• Even apparently similar People/Entities in evolved systems have
internal levels that are marked by heterogeneous histories, and
differing responses.
12 May, 2014
Complex Systems Management
Page 2
Science and Social Systems:
• Are there underlying natural laws for social science?
• Or is social science more about ‘describing’ and noting
regularities? Constructing classifications and taxonomies? Clash
of physiological or evolutionary (cladistic) markers.
• What can complexity science do for social science? Social
systems are evolving, complex systems – with cooperative and
competitive interactions over multiple time scales
• Full of pragmatically learning individuals/groups with diverse
interpretive frameworks and histories. All responding to
individual/group reward/punishment signals locally.
12 May, 2014
Complex Systems Management
Page 3
My Youthful Illusions:
My (very reasonable) starting point
• Physics must be right (and me
too)
And life is therefore:
Input
• A “system” is just a set of
interacting components (possibly
agents), whose behaviour can be
predicted providing we can define
their behavioural rules, and their
interactions
• Problems, including social
systems, can be modelled,
understood and hence “solved”
12 May, 2014
Output
But it is a view with
No Uncertainty –
No Learning!!!
NOT A SOCIAL SYSTEM!
Complex Systems Management
Page 4
Modelling in Ecology, Economics or Societies…
Newton’s Laws work for the
Planets: But what
about evolved Systems?
• Develop the equations of
population dynamics
Species
Growth and
decline
• Measure birth and death
rates, cash flows, sales,
growth rates….
dx1
 F ( x1 , x2 ...xi ,....b1 , m1 ,....).x1
dt
******
Firm Types
Growth and
decline
dxi
 F ' ( x1 , x2 ...xi ,....bi , mi ,....).xi
dt
dxi 1
 F '' ( x1 , x2 ...xi 1 ,....bi 1 , mi 1 ,....).xi 1
dt
12 May, 2014
Complex Systems Management
Page 5
Modelling an Ecosystem or an Economy:
• We can construct and calibrate dynamical equations that
describe the growth and decline different species or types of
firm in the system
• The nodes are ‘species’ in biology or chemistry (population
Dynamics) or types of (economic) activities in human systems
with many parallel supply chains and distribution systems..
C
L
P
I
B
Y
K
H
O
A
X
X
J
J
M
E
E
D
M
D
G
N
F
Computer Model of Interacting
Populations/firms – birth, growth, death
rates Interactions.
12 May, 2014
Run Computer
Forward
Computer Model simplifies
down to a few species/types
Complex Systems Management
Page 6
Modelling an Ecosystem or an Economy:
• We can construct and calibrate dynamical equations that
describe the growth and decline different species or types of
But
in Reality this does not happen!
firm in the
system
• The nodesWhy?
are ‘species’
in biology
or chemistry
(population
Because
each ‘type’
in reality
Dynamics) or types of (economic) activities in human systems
has micro-diversity and the vulnerable
with many parallel supply chains and distribution systems..
go first, others try harder and so the
parameters CHANGE!
C
L
P
I
B
Y
K
The average behaviour is calculated
from the individuals - individuals
are not directed by the average!!!!!
The model has Run
differential
survival –
Computer
Computer Model simplifies
Forward
Computer Model of Interacting
down to a few species/types
But
NOT
behavioural
plasticity!!!
Populations/firms – birth, growth, death
H
O
A
X
X
J
J
M
E
E
D
M
D
G
N
F
rates Interactions.
12 May, 2014
Complex Systems Management
Page 7
Assumptions to go from Reality to a Mechanical Model?
1 Assume that I can define a system with a BOUNDARY
around it separating it from the ENVIRONMENT
2 Assume that I can CLASSIFY the system’s internal
elements involved in the question being asked (X, Y, Z
etc.). Over time some types have disappeared and new
ones have occurred. Evolution!
3 Assume that the behaviour of X, Y and Z will be given by
that of their average STEREOTYPES. (No microdiversity –no adaptive behaviours, or local knowledge)
4.
Assume events occur at
their AVERAGE rates (No
luck or local circumstance)
Non-Linear Dynamics
5.
12 May, 2014
4.
Assume a stationary distribution
Self-Organized Criticality
Assume Equilibrium
Complex Systems Management
Page 8
From REALITY to UNDERSTANDING?
Successive Assumptions
Complexity
Boundary
Classification
Simplicity
Stationaity –
Self-Organized Criticality
Power laws, sand piles,
Stationarity –
Qualitative Research
Structural Stability Fixed
Variables
Reality
Practice
Firm, income, city Sizes..
Attractors
Time
Stationarity
Quantity
Equilibrium
X
Equilibrium
Z
X
Soft
Systems
Not Science
Heuristics
Intuition
Literature,
History,
Descriptions....
Y
Z
Evolutionary Models
CAS
Y
Probabilistic
Non-linear
Dynamics
Average
Dynamics
Structural Evolution
Organizational Change
New Variables
Dynamic, non-Gaussian
Emergence, Innovation Probabilities, Master Equation,
Learning Multi-Agent
Multi-Agent Models
Models
Fixed Variables but
Creativity + Selection Different spontaneous regimes
Or configurations
Strategy
Price
X
Z
Y
Mechanical
Non-Linear
Dynamics
Deterministic System
Dynamics, Chaos, .
From REALITY to UNDERSTANDING?
Successive Assumptions
Complexity
Boundary
Classification
Reality
Practice
Not Science
Heuristics
Intuition
Literature,
History,
Descriptions....
Stationaity –
Self-Organized Criticality
Power laws, sand piles,
Stationarity –
Qualitative Research
Structural Stability Fixed
Variables
Firm, income, city Sizes..
Attractors
Time
Stationarity
Quantity
Equilibrium
X
COMPLEXITY
Y
CAS
Average
Dynamics
Probabilistic
Non-linear
Dynamics
Structural Evolution
Organizational Change
New Variables
Dynamic, non-Gaussian
Emergence, Innovation Probabilities, Master Equation,
Learning Multi-Agent
Multi-Agent Models
Models
Fixed Variables but
Creativity + Selection Different spontaneous regimes
Or configurations
Strategy
Equilibrium
Z
Z
Y
Understanding?
Evolutionary Models
X
Soft
Systems
Simplicity
Contingency
X
Price
Operations
Z
Y
Mechanical
Non-Linear
Dynamics
Deterministic System
Dynamics, Chaos, .
From REALITY to UNDERSTANDING?
Successive Assumptions
Complexity
Boundary
Classification
Reality
Practice
Not Science
Heuristics
Intuition
Literature,
History,
Descriptions....
Stationaity –
Self-Organized Criticality
Power laws, sand piles,
Stationarity –
Qualitative Research
Structural Stability Fixed
Variables
Firm, income, city Sizes..
Attractors
Time
Stationarity
Quantity
Equilibrium
X
COMPLEXITY
Y
CAS
Average
Dynamics
Probabilistic
Non-linear
Dynamics
Medium
Term
Structural Evolution
Organizational Change
Long
Term
New Variables
Dynamic, non-Gaussian
Emergence, Innovation Probabilities, Master Equation,
Learning Multi-Agent
Multi-Agent Models
Models
Fixed Variables but
Creativity + Selection Different spontaneous regimes
Or configurations
Strategy
Equilibrium
Z
Z
Y
Understanding?
Evolutionary Models
X
Soft
Systems
Simplicity
Contingency
X
Price
Operations
Z
Short Term
Y
Mechanical
Non-Linear
Dynamics
Deterministic System
Dynamics, Chaos, .
Evolution is a Multi-Level Phenomenon: Origami
• Emergent Features
• Emergent Variables
• Emergent Dimensions
• Emergent Functionality
• Physics deals with the PAPER
1986 Ev of Multifunctionalism in Ezymes, J McGlade and P Allen,
Can J of Fisheries and Aquatic Sciences, vol 43, No 5
12 May, 2014
Complex Systems Management
Page 12
Evolution is a Multi-Level Phenomenon: Origami
• Emergent Features
• Emergent Variables
• Emergent Dimensions
• Emergent Functionality
• Physics deals with the PAPER
BUT, “Selection” can occur on the basis of the
EMERGENT properties and features of the objects!
- They also form a higher level system!
1986 Ev of Multifunctionalism in Ezymes, J McGlade and P Allen,
Can J of Fisheries and Aquatic Sciences, vol 43, No 5
12 May, 2014
Complex Systems Management
Page 13
Ames and Hall – The Philosophical translation of the
Dao
• Instead of proposing a stable Confucian hierarchy, the Dao De Jing
says:
• “That which can be spoken of is not eternal. That which is named
is not the eternal name. Our new thoughts shape how we think
and act and how we are presently disposed to think and act
disciplines our novel thoughts. It is the underdetermined nature
of the world that makes it, like a bottomless goblet, inexhaustibly
capacious!”
• So, if this brilliant evocation of Evolutionary Complexity was
known 2500 years ago, what has science got to add?
12 May, 2014
Complex Systems Management
Page 14
Can COMPLEXITY help more than the Dao de Jing?
• 2500 years later we can build ‘interpretive
frameworks’ – MODELS! Even if mechanical models
are ‘wrong’ – less than the truth - they can still be
worth building.
• Intelligence can reside in the modeller and/or in the
model. Complexity models will always contain noise,
randomness and non-linearities that can lead to
structural evolution!
• These are not predictions but are experimental
explorations – attempts to imagine possible futures
12 May, 2014
Complex Systems Management
Page 15
What can Complexity ‘do’ for Social Science?
Complexity Science
Non-Equilibrium
Dynamics
Evolution
Attractors
Bifurcations
Chaos – Edge?
Co-evolution
Imperfect Learning
……
Old Science
Equilibrium
Stability
Optimality
Rationality
Perfect
Knowledge
Metaphors?
Confucian?
Daoist?
Social Science
People
Preferences
Values, ethics, morals
Beliefs
Knowledge
Culture
Psychology
Rewards/Risks
Decisions
Descriptions
Surveys,
semi-structured
Data, Statistics
Models
……
On-Going Evolution?
12 May, 2014
Complex Systems Management
Page 16
Induction – Deduction and Back!
Science is about finding general Rules that
allow us to predict particular cases.
Building Interpretive Frameworks
Reinforcement
Deduction
Induction
General
Particular
But, in an evolving world this is an unending process!
Micro-Diversity leads to new behaviours and Learning.
Learning changes the world, requiring more learning!
12 May, 2014
Complex Systems Management
Page 17
Social Science or just muddling through?:
Our interpretive framework results
from our experiences – which are guided
by our interpretive framework!
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
World
NO SCIENCE!
12 May, 2014
Modify, Update
(not unique)
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
Complex Systems Management
Continue
Modify Beliefs
Values given
By beliefs
Individual
My Skull
Do our
experiments
stabilize or
destabilize our
understanding?
Page 18
But, who am I?
Pragmatism is probably all we have. But are we honest?
The Honest
Scientist
12 May, 2014
I was not wrong:
Loss of face, insecurity,
pig-headed…..
Complex Systems Management
What do you
think, Master?
Page 19
But the “world” is also others:
-Incoherence can only end in coherence
-Open exchange can lead to emergent collective capabilities
- Agreement might not be ‘the truth’.
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Modify, Update
(not unique)
World
Aims, Goals
Values
Actions, Experiments
(Noisy, Probabilistic)
Ideas confirmed
Expectation
Deny/Confirm
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Modify, Update
(not unique)
World
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
World
Modify, Update
(not unique)
Modify, Update
(not unique)
World
Expectation
Deny/Confirm
Aims, Goals
Values
Aims, Goals
Values
Ideas confirmed
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Ideas confirmed
Expectation
Deny/Confirm
World
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
World
Modify, Update
(not unique)
Modify, Update
(not unique)
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
12 May, 2014
Complex Systems Management
Page 20
But there are groups, and networks….
Schumpeterians…
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Modify, Update
(not unique)
World
Aims, Goals
Values
Actions, Experiments
(Noisy, Probabilistic)
Ideas confirmed
Expectation
Deny/Confirm
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Modify, Update
(not unique)
World
Pragmatists
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Modify, Update
(not unique)
World
Modify, Update
(not unique)
World
Actions, Experiments
(Noisy, Probabilistic)
Actions, Experiments
(Noisy, Probabilistic)
World
Modify, Update
(not unique)
Aims, Goals
Values
Actions, Experiments
(Noisy, Probabilistic)
Ideas confirmed
Expectation
Deny/Confirm
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Ideas confirmed
Expectation
Deny/Confirm
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Aims, Goals
Values
Aims, Goals
Values
Modify, Update
(not unique)
World
Ideas confirmed
Ideas confirmed
Expectation
Deny/Confirm
Expectation
Deny/Confirm
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
World
Modify, Update
(not unique)
World
Expectation
Deny/Confirm
Aims, Goals
Values
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Aims, Goals
Values
Ideas confirmed
Modify, Update
(not unique)
Actions, Experiments
(Noisy, Probabilistic)
Aims, Goals
Values
Ideas confirmed
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Expectation
Deny/Confirm
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
World
Modify, Update
(not unique)
World
Aims, Goals
Values
Modify, Update
(not unique)
Ideas confirmed
Expectation
Deny/Confirm
Aims, Goals
Values
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Ideas confirmed
Expectation
Deny/Confirm
World
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
World
Modify, Update
(not unique)
Modify, Update
(not unique)
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
Actions, Experiments
(Noisy, Probabilistic)
Economic
Theorists
Modify, Update
(not unique)
World
Flat Earthers
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Actions, Experiments
(Noisy, Probabilistic)
Ideas confirmed
Beliefs, model
“Knowledge”
Modify, Update
(not unique)
World
Actions, Experiments
(Noisy, Probabilistic)
Beliefs, model
“Knowledge”
World
Modify, Update
(not unique)
Modify, Update
(not unique)
Decision, Choice
(not unique)
Expectation
Deny/Confirm
Aims, Goals
Values
Aims, Goals
Values
Ideas confirmed
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Ideas confirmed
Modify, Update
(not unique)
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
World
Modify, Update
(not unique)
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
Ideas confirmed
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Expectation
Deny/Confirm
World
Collective performance will
Lead to retention of ‘successful’
Group identities and ethical behaviour
Modify, Update
(not unique)
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Modify, Update
(not unique)
Expectation
Deny/Confirm
12 May, 2014
Expectation
Deny/Confirm
Aims, Goals
Values
Actions, Experiments
(Noisy, Probabilistic)
Ideas confirmed
World
Modify, Update
(not unique)
World
Modify, Update
(not unique)
World
Aims, Goals
Values
Aims, Goals
Values
Ideas confirmed
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Actions, Experiments
(Noisy, Probabilistic)
Expectation
Deny/Confirm
Decision, Choice
(not unique)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Modify, Update
(not unique)
Aims, Goals
Values
Ideas confirmed
World
Actions, Experiments
(Noisy, Probabilistic)
Beliefs, model
“Knowledge”
Actions, Experiments
(Noisy, Probabilistic)
Expectation
Deny/Confirm
Beliefs, model
“Knowledge”
World
Aims, Goals
Values
Ideas confirmed
Expectation
Deny/Confirm
Decision, Choice
(not unique)
World
Expectation
Deny/Confirm
Decision, Choice
(not unique)
Modify, Update
(not unique)
World
Actions, Experiments
(Noisy, Probabilistic)
Expectation
Deny/Confirm
Actions, Experiments
(Noisy, Probabilistic)
Decision, Choice
(not unique)
Beliefs, model
“Knowledge”
Aims, Goals
Values
Aims, Goals
Values
Ideas confirmed
Complex Systems Management
Page 21
Complexity-Agent Based Urban and Regional Planning:
ISBN : 0-203-99001-3
ISBN 90-5699-071-3
Great Read!
Brussels, Detroit, USA, Senegal, Rhone Valley, Marina Baixa, Argolid, West Bengal, Nepal,….
Guy Engelen & Roger White
12 May, 2014
Complex Systems Management
Page 22
Complexity Applied to Socio-Technical Systems:
Darwin’s Finches
Predicting Diversity
Canadian Fisheries
Rhone Valley
Marina Baixa
Argolid Valley
12 May, 2014
Complex Systems Management
Senegal
Page 23
Complexity, Markets, and Organizational Structure:
Agent 1
Revenue
Total Jobs
Customer Agents
Customers w ith
Sales Staff
a product
Production Staff
SALES
+
Net
Inputs+Wages
STOCK
Interaction
Type 1
Production
- Fixed Costs
Potential Market
Profit
COSTS
QUALITY
PRICE
Relative
Attractivity
Attributes of
Customer
Type 2
Demand
Attributes
of Supply
Type 3
PROFIT MARGIN
OTHER
FIRM 1
FIRMS
POTENTIAL
CONSUMERS
Other Agents
Strategy on Profit Margin,
Quality, R&D, Design….
Economic Markets
Manufacturing Practices
and Techniques
Aerospace
Supply Chain
practices
12 May, 2014
Complex Systems Research Centre
Page 24
But REFLEXIVITY occurs in Agent Based Models:
What’s in
It for me?
Plato’s Cave
Data
Model
Modeller
12 May, 2014
Complex Systems Management
Page 25
But REFLEXIVITY occurs in Agent Based Models:
What’s in
It for me?
Plato’s Cave
Data
Model
Modeller
12 May, 2014
Complex Systems Management
Page 26
But REFLEXIVITY occurs in Agent Based Models:
What’s in
It for me?
Plato’s Cave
Data
Model
Such a Model could be a tool for building
Social Cohesion – e.g. Climate Change but
Only if you already have a Community!
If you haven’t got a Community
You can’t get one easily….
Modeller
12 May, 2014
Complex Systems Management
Page 27
1: Complexity, Knowledge and Prediction:
• Traditional Science depends on Repeatable Experiments –
essentially Closed Systems – and provides Popperian
Knowledge. The behaviour of the elements under study is
not changed by their experiences. Molecules don’t get
bored or angry and have a poor sense of humour!
• Social Science must deal with evolutionary “Creative
Destruction”. Repeatable experiments not really possible.
New things emerge, new structures form with emergent
features and capabilities, while other things disappear.
• ‘Understanding’ is about facing what cannot be understood. The
‘micro-MESS’ is the engine of resilience and future evolution –
the opaque core of evolutionary complexity.
12 May, 2014
Complex Systems Management
Page 28
2 Evolution: Limits to Knowledge and Explanation?
• When we look at a system/network/organization all that
we see are things/behaviours that happen to have been
created, minus the non-viable and the unlucky!
• This means that what we will find does not necessarily
‘make sense’ and many elements may have no role or
function. We cannot necessarily understand ‘better’ by
analysing more data.
• This affects the nature of ‘explanation’ since we cannot
always attribute functionality to elements. It also makes
the outcomes of innovations unpredictable
• Models with noise, randomness and non-linearities can tell
us things that we didn’t know and weren’t in the data!
12 May, 2014
Complex Systems Management
Page 29
3: Reflexivity – Why the future is not what it was
• In Social Science we have REFLEXIVITY. Different agents
behave according to their own acquired interpretive
frameworks, which continue to change with their ‘resulting’
experiences and changing the variables and model qualitatively.
• And when agents see the outcomes of a model, they may
modify their internal representations and hence their behaviour.
This would then INVALIDATE the model.
• We could INCLUDE the changes in participating agents’
representations as part of the model. (Machiavelli) Or include
agents anticipating other agents ‘learning’ in their internal
representations. Machiavelli SQUARED!!! Etc.
• This is not a solved problem!
12 May, 2014
Complex Systems Management
Page 30
4: Creative Destruction & Complexity for ever!
• Complexity and Evolution involve changing systems of
changing elements. Both Qualitative and Quantitative
changes happen as structural instabilities (new variables)
of the dynamics occur. Qualitative Research is important
• Our interpretive frameworks (Models) do not make
predictions about the world, but about themselves. We
have an unending deductive/inductive loop that is the
instrument of our exploration/reflection. A model may warn us of
possible problems (e.g. climate change, limits to growth) and increased
“credibility” may make evasive action MORE likely.
• Our understanding and interpretive frameworks are
just part of the Evolving Complex World.
12 May, 2014
Complex Systems Management
Page 31
A Good Book!
12 May, 2014
Complex Systems Management
Page 32
Recent Festschrift:
Emergent Publications
ISBN 978-1-938158-13-1
An interesting collection of very
diverse papers written by some
of the friends who have worked
with me over the years.
http://www.amazon.com/The-SocialFace-ComplexityScience/dp/193815813X
12 May, 2014
Complex Systems Management
Page 33